WO2026027508A1

WO2026027508A1 - Controller for a heat pump-based water heating and/or cooling system

Info

Publication number: WO2026027508A1
Application number: PCT/EP2025/071744
Authority: WO
Inventors: Aurélien MASSEIN
Original assignee: BDR Thermea Group BV
Current assignee: BDR Thermea Group BV
Priority date: 2024-08-02
Filing date: 2025-07-29
Publication date: 2026-02-05
Anticipated expiration: 2027-02-02
Also published as: EP4686885A1

Abstract

An implementation of the disclosed technology provides a controller (100, 210) configured to control at least a hot water system, the hot water system (200) includes a heat pump (202), an optional electrical backup thermal source, in particular heat source, (208) and/or an optional hydraulic backup heat source. The controller includes a thermal comfort determination module (302) configured to predict a thermal comfort satisfaction level based and/or trained on at least one of: a predicted water temperature at the point of use, a predicted availability of water during expected usage periods, historical user feedback or usage patterns,or system response time to a water demand; and an energy optimization input module (304) configured to predict an energy-related objective level based and/or trained on at least one of: a predicted activation duration and/or frequency of the heat pump (202); the operational status and contribution of the optional electrical backup thermal source, in particular heat source, (208) and/or optional hydraulic backup heat source; historical usage patterns and demand forecasts; external temperature, in particular room temperature or outdoor temperature, and system thermal losses; or at least one energy tariff, CO₂ footprint, or cost model. The controller is configured to: store instructions in a memory (110); access data and execute instructions via a processor (102); communicate with the hot water system via a communication interface (140); and receive a prediction of the thermal comfort satisfaction level from the thermal comfort determination module (302); receive a prediction of the energy-related objective level from the energy optimization input module (304); perform an optimization operation (320) that determines a control strategy for the hot water system (200), the optimization being based on a balanced consideration of a thermal comfort objective and at least one energy-related objective, comprising cost, environmental impact, or system efficiency; and that determine commands to operate the hot water system (200) based, at least in part, on a result of the optimization operation wherein the commands in particular comprise scheduling, source selection, or control command. The invention further relates to a hot water system (200), a method for controlling it, a data processing device, a computer program product, a computer-readable data carrier, and a use of the controller (100, 210).

Description

CONTROLLER FOR A HEAT PUMP-BASED WATER HEATING AND/OR

COOLING SYSTEM

TECHNICAL FIELD

[1] The disclosed technology relates to a controller for a hot water system. Moreover, the invention relates to a hot water system, a method for controlling the hot water system, a data processing device, a computer program product, a computer readable data carrier and a use of the controller.

BACKGROUND

[2] Water heating and/or cooling systems are means of producing hot or cool water for heating and/or cooling applications, and may also be used in the production of domestic and/or sanitary hot water. It will be understood that while much of the discussion here will center on heat pump water heaters that are used for the production of domestic hot water, similar systems may be used for other heating and/or cooling applications, and the discussion of heat pump water heaters is intended merely to illustrate the technology. Similar systems and methods may be applied to a wide range of applications in which water or other liquids are heated and/or cooled for use in a wide range of heating and/or cooling applications.

[3] A heat pump water heater is a water heater with at least one heat pump as a water heating element. A heat pump water heater may have other associated heating means such as an electric resistive element or heat exchangers associated with a hot water circuit. Backup heating, such as a gas-powered boiler, pellet boiler, thermal solar panel, and/or other heating systems may also be used.

[4] Heat pump water heaters need an available energy/heat source to transfer heat to the water to be heated. Different sources can be used. The type of source can be used to differentiate and designate different types of heat pump water heaters. [5] Ambient air heat pump water heaters use ambient air as an energy source. Ambient air entering and leaving the heat pump is drawn in and returned to the volume of air available at the installation site. The place of installation is often an unheated, frost-free room such as a utility room in a building (e.g., a cellar, garage, attic, etc.), though controlled temperature rooms may also be used. Ambient air heat pumps provide a simplified product compared to other heat pump types. This is because, e.g., the ambient air used has a limited and positive temperature range. In addition, the pressure losses in the air flow are low as the air is pulled in and discharged in the same space as the product location.

[6] Ducted air heat pump water heaters use outside air as an energy source - i.e. , air that is drawn in and/or discharged from and/or to the outside. This type of heat pump water heater offers greater flexibility in installation modes and allows the user to choose a configuration that provides for comfort throughout the year. For example, it is possible to choose to discharge the air from the heat pump to the outside when the outside temperature is below the comfort temperature of the room. Alternatively, one can choose to recirculate the air from the heat pump at the installation site when this provides comfort. This type of heat pump water heater may be susceptible to significant pressure drops due to lengths, bends, and height differences in the ducts. A heat pump system could also be split, with an outside portion (e.g., an outdoor unit (ODU) with a mono-block or split-block heat pump) and an indoor portion (e.g., an indoor unit (IDU) with the tank or heating/cooling circuit).

[7] Extracted air heat pump water heaters use air extracted from a ventilation network of the installation building as an energy source. This air has a relatively constant and high temperature, as it comes from living areas of a dwelling. In general, this air may be particularly humid and can contain significant levels of dust or other debris, because it may come from damp rooms, such as bathrooms or kitchens. Heat pumps using this type of air must be able to operate with a relatively low air flow rate (that of the building's ventilation system). A fan may be included in the heat pump water heater. Alternatively, the fan of the ventilation system of the installation building may be used, removing the need for a fan in the heat pump water heater. [8] Water or ground source heat pump water heaters use an open (e.g., for underground water pumping) or closed (e.g., for soil) water circuit as an energy source. The water circuit may be the return of a heating circuit, a geothermal circuit or any other closed or open water circuit.

[9] For all above-mentioned heat pump water heater types, the heat pump includes at least one closed refrigerant circuit. The refrigerant circuit includes a first heat exchanger with the source medium (an evaporator), a compressor, a second heat exchanger with the destination medium (a condenser), in particular water or other liquids (e.g., for use in heating and/or cooling applications, and/or domestic hot water), and an expansion device.

[10] Control of heat pump water heating and/or cooling systems can present a number of challenges. Because it can take some time to heat or cool water using a heat pump, it may be difficult to ensure that there is sufficient hot or cool water to meet demands. For example, when the demand for hot water becomes greater than can be readily supplied, systems may fall back on other heat sources, such as electric resistive heating or a backup gas boiler. While these sources may be faster at producing hot water than a heat pump, they may be less energy efficient and/or more expensive or environmentally harmful to operate.

[11] Environmental conditions can also cause control challenges. For example, heat pump water heating systems may be more efficient when the source medium (e.g., outside air, ambient air) is warm than when it is cold. In systems that use outside air, the heat pump may not be able to operate when the outside air temperature is too cold. As with meeting demands for hot and/or cool water, the energy efficiency of heat pumps will vary with environmental conditions, and the use of other heat sources, such as electric resistive heating or a backup gas boiler may make sense, depending on environmental conditions.

[12] Control systems for heat pump water heating and/or cooling systems, therefore, need to be able to take into account environmental conditions, anticipated hot or cool water demand, energy consumption, defrost cycles, and other factors to deliver a desired level of comfort and a reasonable level of energy usage. In addition to the challenges that are imposed by these conditions, users of such systems may have differing demands on how control is to be handled. For example, some users may place a greater emphasis on energy savings, while others may place a greater emphasis on comfort.

[13] WO2023183576A1 addresses the problem of inefficient and inflexible control systems for heating fluids, particularly in applications such as heat pumps. Traditional systems often rely on static or reactive control strategies that fail to account for dynamic variables like real-time energy prices, solar energy availability, and greenhouse gas (GHG) intensity. This results in suboptimal energy usage, increased operational costs, and a higher environmental footprint. WO2023183576A1 proposes a predictive control method and system that enhances the efficiency and responsiveness of heat transfer to or from a fluid. The system receives a range of input data, including user preferences, energy pricing, solar forecasts, and GHG intensity metrics. Based on this data, WO2023183576A1 uses a predictive controller to determine optimal settings — such as fluid temperature set points and compressor configurations — for a heat pump system. The fluid is then heated in accordance with these settings to meet user comfort requirements while maximizing economic and environmental efficiency. By integrating predictive analytics and real-time data, WO2023183576A1 enables proactive and adaptive control of fluid heating systems, offering a more sustainable and cost-effective solution. The architecture of WO2023183576A1 is built around an integrated predictive control system that combines forecasting, optimization, and reactive adjustment into a unified framework. At its core, WO2023183576A1 includes a resource usage prediction capability that estimates future energy consumption, solar energy availability, and greenhouse gas (GHG) intensity. These forecasts are not handled by isolated modules but are embedded within the control logic itself, allowing the system to anticipate and plan for optimal operation over a future time horizon. This predictive layer is tightly coupled with an optimization engine that balances multiple objectives — such as minimizing energy cost, reducing environmental impact, and maintaining user comfort. The optimization is holistic, meaning it considers all available data simultaneously rather than treating each input in isolation. The system continuously updates its predictions and control decisions as new data becomes available, ensuring that it remains aligned with real-time conditions. In addition to its predictive capabilities, WO2023183576A1 incorporates a reactive control mechanism. This allows the system to respond immediately to unexpected changes, such as a sudden spike in energy prices or an unplanned demand for hot water. These reactive adjustments override or refine the predictive plan to maintain performance and user satisfaction. Overall, the architecture of WO2023183576A1 is characterized by its tight integration of forecasting, optimization, and real-time responsiveness. It is designed to operate fluidly in dynamic environments, where external conditions like energy markets and renewable availability can shift rapidly. This makes it particularly well-suited for applications like smart heat pump systems, where both efficiency and adaptability are critical.

[14] US20230019836A1 addresses the problem that traditional building management systems often perform inefficiently in new buildings due to the absence of historical operational data. Without such data, these systems struggle to generate effective control settings, resulting in poor energy efficiency and suboptimal occupant comfort. US20230019836A1 proposes a solution in the form of an intelligent building energy management system that uses a building model — representing the physical structure of the building — and historical weather data to simulate predicted energy usage. Based on this simulation, a predictive model is trained to estimate optimal equipment operating settings for systems such as HVAC, lighting, and air quality control. These settings are then used to manage environmental conditions within the building. US20230019836A1 further describes how the predictive model is continuously retrained using real-time sensor data and user feedback on comfort levels. This enables the system to adapt to actual building behaviour and improve its performance over time. The patent also includes a user interface that allows users to graphically define the building model, view optimal settings, and visualize projected energy savings and electric load comparisons. Through this adaptive, data-driven approach, US20230019836A1 aims to enhance both energy efficiency and occupant comfort in building environments.

[15] US10378805B2 is directed to a control system for heating fluids, such as domestic hot water, with the objective of improving energy efficiency and ensuring that the fluid reaches a desired target temperature at a specified future time. The system is based on a model predictive control (MPC) approach. In order to achieve this goal, US10378805B2 discloses a controller that uses a predictive thermal model to simulate the behavior of the heating system over a future time horizon. The controller receives a target temperature and a target time, and calculates a sequence of intermediate temperature setpoints that guide the system toward the desired condition. The optimization process considers energy cost, deviation from the target temperature, and system constraints. The control logic operates in a receding horizon fashion, where the optimization is repeated at each control interval using updated measurements. The system is particularly suited for configurations involving resistive heating elements and intermediate fluid loops. The controller adjusts the operation of the heating system to minimize energy consumption while achieving the target temperature at the desired time. The system may include a user interface through which a user can specify a desired temperature and a time by which that temperature should be reached. The controller uses this input to determine a control strategy that minimizes energy use while satisfying the user’s request. The predictive model may incorporate thermal characteristics of the heating system, including the fluid loop and heating element, and may be refined using historical data to improve accuracy. The optimization is performed using a cost function that includes terms for energy consumption and deviation from the target temperature.

SUMMARY

[16] Based on the above, it is an object to provide a controller for a hot water system, particularly a hot water system using a heat pump as a primary heat and/or cool source, that is able to balance thermal comfort and energy consumption. In other words, it is an object to enable a controller configuration and strategy for hot water systems that balances thermal comfort and energy consumption in a personalized, efficient, and user-friendly manner — without requiring technical intervention.

[17] It is a further object to provide a controller which address the technical problem of how to reconcile user comfort requirements with energy efficiency constraints in the operation of the hot water system. This problem arises particularly in today’s dynamic conditions where user expectations for water temperature and availability must be met under varying environmental conditions and energy cost structures.

[18]

[19] It should also be understood that, as used herein, the term “hot water system” refers to systems configured to provide hot water for domestic use or space heating. In some embodiments, the system may also be capable of delivering water at ambient temperature or lower than ambient temperature or be adapted for additional thermal applications, provided such sunctionality is supported by the system architecture. For example, a hot water system may provide domestic hot water, hot water for heating applications, such as room heating, and/or cool water for cooling applications, such as room cooling. This capability is particularly relevant where the system includes a reversible heat pump or a similar bidirectional thermal device. Such a device may include, for example, a refrigeration cycle unit with reversible flow paths, or a hydronic module equipped with both heating and cooling exchangers, along with control logic capable of switching between thermal modes. These configurations enable the system to operate flexibly in response to environmental conditions and user demand. The ability to support cooling or other thermal modes is optional and depends on the specific implementation.

[20] The object is solved by a controller configured to control at least a hot water system, the hot water system comprising a heat pump and an optional electrical backup thermal source, in particular heat source, and/or an optional hydraulic backup heat source, the controller comprising:

• a memory;

• a processor configured to access data and execute instructions stored in the memory; and • a communication interface communicatively coupled to the processor and to the hot water system, the communication interface configured to communicate commands and data between the controller and the hot water system.

The controller further comprises: a thermal comfort determination module configured to predict a thermal comfort satisfaction level delivered by the hot water system based and/or trained on at least one of:

• a predicted water temperature, in particular of a predefined water temperature, at the point of use,

• a predicted availability of water, in particular water of a predefined temperature, during expected usage periods,

• historical user feedback or usage patterns, or

• system response time to water demand, in particular a predefined temperature water demand; and an energy optimization input module configured to predict an energy consumption level of the hot water system based and/or trained on at least one of:

• a predicted activation duration and frequency of the heat pump;

• the operational status and contribution of the optional electrical backup thermal source, in particular heat source, and/or optional hydraulic backup heat source;

• historical usage patterns and demand forecasts;

• ambient temperature and system thermal losses;

• at least one energy tariff, CO₂ footprint, and/or at least one cost model associated with a heat source.

The memory stores instructions that when executed by the processor cause the processor to:

• receive a prediction of the thermal comfort satisfaction level from the thermal comfort determination module;

• receive a prediction of the energy consumption level from the energy optimization input module; perform an optimization operation that determines a control strategy for the hot water system, the optimization being based on a balanced consideration of:

• a thermal comfort objective, and

• at least one energy-related objective, comprising cost, environmental impact, or system efficiency; wherein the optimization operation determines whether an adjustment of the operation of the hot water system is required and, if so, defines such adjustment based at least in part on the predicted thermal comfort satisfaction level and the predicted energy-related objective level, such that the adjustments satisfy a balanced combination of the thermal comfort and energy-related objectives; and that determine commands to operate the hot water system based and/or trained on a result of the optimization operation, in particular the commands to operate comprising at least one of scheduling, source selection, or control command.

[21] In some implementations, In some implementations, the scheduling and source selection decisions determined by the controller may account for:

• at least one time-of-day preference, such as morning or evening comfort profiles;

• availability of renewable or lower-cost energy sources, including solar, ambient air, or grid-supplied electricity during off-peak periods; and

• at least one system performance constraint, such as maximum heating capacity, thermal inertia, or response time limitations of the heat pump or backup sources.

This embodiment enables the controller to adapt its control strategy to user routines, energy availability, and system limitations, thereby improving both comfort delivery and energy efficiency.

[22] In some implementations, the disclosed technology provides a controller configured to control at least a hot water system, the hot water system comprising a heat pump and an optional electrical backup thermal source, in particular heat source, and/or an optional hydraulic backup heat source. The controller includes: a memory; a processor configured to access data and instructions from the memory; and a communication interface communicatively coupled to the processor and to the hot water system, the communication interface configured to communicate commands and data between the controller and the hot water system. The controller further includes a thermal comfort determination module configured to predict a thermal comfort satisfaction level delivered by the hot water system and an energy optimization input module configured to predict an energy-related objective level of the hot water system. The memory further includes instructions that when executed by the processor cause the processor to: receive a prediction of the thermal comfort satisfaction level from the thermal comfort determination module; receive a prediction of the energy-related objective level from the energy optimization input module; perform an optimization operation based, at least in part, on the prediction of the thermal comfort satisfaction level and the prediction of the energy-related objective level; and determine commands to operate the hot water system based, at least in part, on a result of the optimization operation.

The term “energy-related objective level” within the meaning of this application is a predictive value representing the expected energy performance of a hot water system under specific operating conditions. It may reflect energy consumption, cost, efficiency, or environmental impact, and is used as an input to optimization operations that balance energy use with thermal comfort.

[23] The controller within the meaning of this application is a computing device or a system comprising at least a processor, memory, and communication interface, configured to manage the operation of a hot water system.

[24] The hot water system within the meaning of this application refers to system configured to supply heated water, comprising at least a heat pump and optionally one or more backup heat sources, such as electrical or hydraulic sources.

[25] The term heat pump within the meaning of this application refers to a device configured to transfer thermal energy from a lower-temperature source to a higher-temperature sink, typically using a refrigeration cycle. In the context of a hot water system, a heat pump extracts heat from an external source — such as ambient air, ground, or water — and transfers that heat to water stored or circulated within the system. The heat pump may operate in one or more modes, including heating, cooling, or reversible operation, and may include components such as a compressor, evaporator, condenser, expansion valve, and associated control circuitry. The heat pump is configured to operate independently or in conjunction with auxiliary heating sources to achieve desired thermal output.

[26] The term backup heater within the meaning of this reference refers to a resistive or other electrically powered heating element used to supplement or replace the heat pump under certain conditions.

[27] The term hydraulic backup heat source within the meaning of this reference refers to a heat source that uses fluid-based thermal energy transfer, such as from a central heating system or district heating network.

[28] The thermal comfort satisfaction level is a value that represents how effectively the hot water system is expected to meet user-defined comfort criteria. These criteria include water temperature at the point of use, availability during anticipated usage periods, and responsiveness to demand. The value is generated by a thermal comfort determination module using predictive models based on system operating parameters, environmental conditions, historical usage patterns, and optionally user feedback. The thermal comfort satisfaction level may be expressed as a numerical score, percentage, or categorical value, and serves as an input for optimizing the operation of the hot water system. For example, if a user typically showers at 7:00 AM and prefers water at 42°C, the module may predict a satisfaction level of 95% when the system preheats water to 43°C by 6:50 AM. If the water is only 39°C at that time, the predicted satisfaction level may drop to 60%, indicating a need to adjust the heating schedule.

[29] The thermal comfort determination module refers to a software and/or hardware component configured to generate the thermal comfort satisfaction level based on system parameters, user preferences, and historical usage data.

[30] The energy optimization input module refers to a module configured to estimate or predict the energy usage of the hot water system under specific operating conditions.

[31] An optimization operation within this application means a computational process that evaluates multiple predicted outcomes (e.g., comfort and energy use) to determine the most favorable operating parameters for the hot water system. So, as used herein, the term “optimization operation” refers to a computational process executed by the controller to determine a control strategy for the hot water system. The operation is configured to balance at least a thermal comfort objective and one or more energy-related objectives, such as cost, environmental impact, or system efficiency. The optimization operation may be implemented using a variety of techniques, including but not limited to:

• Model predictive control (MPC), where future system behavior is predicted and control actions are selected to minimize a cost function over a prediction horizon;

• Rule-based or heuristic logic, where control decisions are derived from predefined rules, thresholds, or expert knowledge;

• Machine learning-based policies, where control strategies are inferred from models trained on historical data, user feedback, or system performance;

• Adaptive or feedback-based control, where the system continuously adjusts its behavior in response to real-time measurements or deviations from expected outcomes.

The optimization operation may involve explicit evaluation of multiple control options or may rely on implicit decision-making mechanisms embedded in the control logic or learned models.

[32] As used herein, the term “weighted consideration” or “weighted combination” refers to a method of evaluating multiple objectives — such as thermal comfort and energy-related goals — by assigning relative importance values (weights) to each objective. These weights influence the outcome of the optimization operation by prioritizing one objective over another depending on system configuration, user preferences, or contextual factors. The weights may be:

• Static: preconfigured by the manufacturer or system designer;

• User-defined: adjustable by the end user through a user interface;

• Adaptive: dynamically adjusted by the controller based on historical usage patterns, feedback, or external signals (e.g., energy tariffs or carbon intensity). The weighted combination may be implemented explicitly, for example, through a cost function of the form:

J ^1 ' f comfort T ^2 ' fenergy where wi and W2 are weights, and fcomfort, fenergy are functions representing the respective objectives. Additionally or alternatively, the weighting may be embedded in a learned policy and/or heuristic rule set that implicitly balances the objectives. In other words, depending on the implementation, the optimization operation may rely on a learned policy and/or a heuristic rule. For example, a learned policy may be trained on historical optimization outcomes, while a heuristic rule may encode expert knowledge or predefined thresholds. In other words, the decision-making logic may be embedded in a learned policy, a heuristic rule, or a combination thereof. An advantage of this flexible architecture is that it allows the controller to adapt to different deployment contexts — ranging from data- rich environments where machine learning can be applied, to constrained systems where rule-based logic is more appropriate — while still enabling consistent optimization of thermal comfort and energy-related objectives.

[33] As used herein, scheduling and source selection decisions refer to control actions determined by the controller to manage:

• Scheduling: the timing of heating operations, including when to activate the heat pump or backup sources, and when to preheat or delay heating based on predicted demand, energy availability, or tariff windows.

• Source selection: the choice of which heat source(s) to activate (e.g., heat pump, electrical backup, hydraulic backup) based on efficiency, cost, environmental impact, or availability.

These decisions may be:

• Rule-based, using predefined logic (e.g., “use heat pump unless ambient temperature is below a predefined value”);

• Predictive, using forecasted demand and energy prices;

• Real-time, responding to immediate system conditions or user inputs.

The decisions are derived from the result of the optimization operation, which may be either: • Based on: a deterministic output from the optimization logic (e.g., a control schedule);

• T rained on: a learned policy or model that has been trained using historical optimization outcomes or simulated scenarios.

Depending on the implementation, control decisions may be based on deterministic outputs from the optimization logic — such as control schedules or rule-based thresholds — or trained on historical optimization outcomes using machine learning models. For example, a control schedule may be generated directly from the optimization operation, while a learned policy may be trained to approximate optimal decisions under varying conditions. In other words, the decision-making logic may be embedded in a learned policy, a heuristic rule, or a combination thereof. Training may also include learning from user-defined or system-defined rules, control parameter values, or historical setting changes. For example, the controller may adapt rule thresholds based on observed outcomes, adjust optimization weights based on user behavior, or infer preferred schedules from repeated manual interventions. This allows the system to evolve over time and better align with user preferences and operational efficiency.

[34] A communication interface within the meaning of this application is a hardware and/or software interface that enables data exchange between the controller components of the hot water system, such as sensors, actuators, heat sources, or auxiliary modules. This interface may support unidirectional or bidirectional communication and may be implemented using wired or wireless protocols, depending on the system configuration.

[35] A control command within the meaning of this application refers to instructions generated by the controller to adjust the operation of the hot water system. Examples include activating or deactivating the heat pump, switching between primary and backup heat sources, modifying temperature setpoints, or initiating a specific control mode. These commands may be determined based on optimization logic, user preferences, or learned behavior, thereby enabling a dynamic balance between thermal comfort and energy efficiency.

[36] The balance between thermal comfort and energy consumption in the claimed controller is achieved through a structured optimization process that integrates predictive inputs from two distinct modules: the thermal comfort determination module and the energy optimization input module.

[37] The thermal comfort determination module predicts a comfort satisfaction level, which reflects how well the hot water system is expected to meet the user's comfort expectations - such as delivering water at a preferred temperature and time. Simultaneously, the energy optimization input module predicts the energy usage associated with operating the system under various conditions. These two predictions are then fed into the optimization operation, which evaluates different control strategies to find one that best satisfies both objectives: maximizing comfort while minimizing energy use. The optimization is configured to allow for trade-offs that can be influenced by user preferences. For example, a user might prioritize comfort in the morning and energy savings in the evening. The system can adjust accordingly by weighting the comfort and energy predictions differently in the optimization process. Alternatively, the system may treat thermal comfort and energy consumption goals as fixed or absolute - for example, by assigning constant weights to each objective - thereby supporting a fixed-weight optimization configuration that enables faster execution and may be preferable for less complex system implementations, while remaining consistent with the functional structure and architecture of the invention. By structuring the controller in this way, the claim enables a dynamic and user-sensitive balance. The controller according to the invention thus does not rely on static rules or fixed thresholds but instead uses predictive modeling and optimization to adapt to changing conditions and preferences. This approach helps resolve the subjectivity inherent in comfort-energy trade-offs and supports user-friendly customization without requiring manual tuning of technical parameters. In some configurations, the controller can be configured to apply fixed weights to the predictive outputs, enabling faster execution and making it suitable for less complex system implementations — both approaches being compatible with the described control strategy.

[38] The controller according to the invention thus solves the problem by introducing a system architecture in which the controller receives predictions from both the thermal comfort module and the energy consumption module. These predictions are then used in an optimization operation that determines how the hot water system should be operated. Because the optimization is based on both comfort and energy use, and because the comfort level can be influenced by user- defined preferences or thresholds, the system inherently supports subjective customization.

[39] As a further advantage, the user does not need to manually adjust complex parameters. Instead, the system interprets user preferences — either directly, through explicit settings, or indirectly, through learned behavior such as usage patterns and feedback — and integrates them into the optimization logic. These preferences guide the outcome by influencing how the system weighs thermal comfort against energy consumption during the optimization process. For example, a user who consistently prefers higher water temperatures in the morning may have their comfort prioritized during that time, while energy savings may be favored at other times. This satisfies the objective of allowing the user to influence the comfort-energy trade-off in a manner that is easy to use, without requiring technical expertise, manual tuning of system parameters, or the need to refer to a long or complex manual. In other words, the controller provides a structured, automated way to balance comfort and efficiency, while allowing user preferences to guide the outcome, directly addressing the dual goals of personalization and usability.

[40] In some implementations, the controller further includes an environmental determination module configured to predict environmental conditions outside of the hot water system. An environmental determination module within the meaning of this application is a subsystem designed to predict external environmental conditions such as weather, temperature, or humidity. These predictions are not merely observational; they are proactively integrated into the system’s optimization logic. The environmental determination module according to the invention has the additional advantage of enabling the system to anticipate and respond to future environmental conditions, such as changes in weather, temperature, or humidity, rather than simply reacting to current data. This predictive capability allows the system to optimize its performance proactively, improving energy efficiency and maintaining user comfort more effectively. By integrating these forecasts into its control logic, the system can make smarter decisions about when and how to operate, ultimately leading to more sustainable and cost-effective outcomes. In other words, with access to predicted environmental data, the controller can be configured to carry out an optimization operation that adjusts its behavior to improve overall performance. This optimization may include modifying heating schedules, reducing energy consumption during warmer periods, or preheating water in anticipation of colder weather. Rather than relying solely on real-time data, the system further advantageously incorporates forecasted environmental conditions along with user preferences and historical usage patterns to make informed decisions. This architecture allows the system to operate even more proactively, further enhancing both energy efficiency and user comfort by anticipating needs and responding in advance.

[41] In some implementations, the memory includes instructions that when executed by the processor cause the processor to receive predicted environmental conditions from the environmental determination module and to perform the optimization operation further based, at least in part, on the predicted environmental conditions.

[42] In some implementations, the controller further includes an information module configured to provide operation history information and/or product information on the hot water system. In some implementations, the memory includes instructions that when executed by the processor cause the processor to receive operation history information and/or product information from the information module and perform the optimization operation further based, at least in part, on the operation history information and/or product information.

[43] In some implementations, the optimization operation is influenced by a single parameter set by a user, the single parameter providing a user-selected balance between the thermal comfort satisfaction level and the energy-related objective level in the optimization operation. This single parameter provides an easy to use means for the user to provide the controller with their preferences with respect to comfort and energy consumption. It is an advantage of the controller according to the embodiment that the controller advantageously further enables the user to influence the balance between thermal comfort and energy consumption through a single, intuitive parameter. This parameter allows the user to express a preference without requiring detailed configuration, thereby simplifying interaction with the system while still enabling personalized optimization of hot water delivery.

[44] In some implementations, the single parameter can be a user- determined parameter which may be a fixed setting at design time to reflect a predefined balance between comfort and energy consumption. Additionally, this parameter may serve as a basic condition during an initial operation phase, such as a starter mode following installation or commissioning, where the controller is configured to operate using simplified predictive logic and predefined defaults. This starter mode is particularly beneficial for first-time users, offering a low- complexity, low-risk configuration that ensures reliable performance without requiring advanced input, while still enabling meaningful balancing of thermal comfort and energy consumption. The controller may transition from this mode to more sophisticated configurations based on elapsed time, user interaction, or system learning. Furthermore, the starter mode may be re-engaged as a fallback configuration in case of system reset or user preference.

[45] In some implementations, the controller may be configured to use a single user-determined parameter to influence the trade-off between thermal comfort and energy consumption. This parameter may be normalized to a value between 0 and 1 , where a value of 0 gives maximum importance to thermal comfort satisfaction level, and a value of 1 gives maximum importance to energy- related objective level (e.g., minimizing energy use or cost). The predicted values for the thermal comfort satisfaction level and the energy-related objective level may also be normalized to values between 0 and 1. For example, a thermal comfort satisfaction level of 0 may represent minimal comfort, while a value of 1 may represent optimal or “perfect” comfort. Similarly, an energy-related objective level of 0 may represent maximum energy use or cost, while a value of 1 may represent minimal energy use or cost. An advantage of this approach is that it enables intuitive configuration for users, especially in early usage phases, by reducing complexity to a single normalized input while still allowing meaningful control over system behavior. This configuration enables the controller to pursue concurrent optimization objectives, dynamically balancing thermal comfort and energy efficiency based on a unified parameter and shared predictive inputs, thereby allowing energy users to be operated as efficiently as possible.

[46] In such a configuration, the controller may apply a fitness function that combines the normalized predictions using the user-determined parameter. One example of such a function is: fitness (%) = Comfort ■ (1 — x) + Consumption ■ x where “x” is the user-determined parameter, “Comfort” is the predicted thermal comfort satisfaction level, and “Consumption” is the predicted energy-related objective level. This function further enables the controller to evaluate control strategies based on a weighted combination of the two predicted outcomes. A higher value of x shifts the optimization toward energy efficiency, while a lower value prioritizes thermal comfort. In this case, the optimization operation is influenced by a single user-defined parameter that provides a tunable balance between thermal comfort satisfaction and energy consumption. A lower value of x prioritizes comfort, while a higher value emphasizes energy efficiency. The comfort satisfaction level may be represented as a numerical score, such as a similarity metric or deviation from a predefined satisfaction threshold. The optimization may also account for predicted time-to-comfort when selecting between a heat pump and one or more backup heat sources. This approach provides a number of advantages. First, it allows the user to influence the optimization outcome through a single intuitive setting, without requiring technical knowledge or manual tuning of complex system parameters. Second, it supports a structured and transparent optimization process that can be adapted to different user preferences and usage contexts. Third, the use of normalized values and a parameterized fitness function enables consistent behavior across different installations and system configurations. Finally, while the linear fitness function described above is one useful implementation, other fitness functions may also be used. The user-determined parameter may be applied in non-linear or multi- dimensional formulations, and need not always serve as a direct mathematical weight. This formulation further enables concurrent optimization of thermal comfort and energy efficiency, allowing the controller to operate energyconsuming components as particularly efficiently while respecting user-defined comfort preferences.

In some implementations, the controller may incorporate a machine learning model, such as a neural network, decision tree, support vector machine, or reinforcement learning agent, to predict thermal comfort satisfaction and energy- related objective levels. The model may be pre-trained or continuously updated based on operational data. Its output may be used as input to an optimization operation, which determines control actions based on a fitness function or similar evaluation metric. An advantage of this approach is that it enables adaptive and data-driven control, allowing the system to respond intelligently to changing conditions and user preferences while maintaining efficient operation and comfort balance.

[47] In some implementations, the controller is configured to comprise a predictive module that estimates future thermal demand and energy consumption. These estimates are based on historical usage data, environmental conditions, and user-defined comfort parameters. The controller evaluates the expected heating or cooling load over a defined time horizon and determines an operational strategy that balances energy efficiency with thermal comfort. To do so, the controller considers the performance characteristics of the primary heat pump system as well as any available auxiliary heating means, such as electric resistive elements or hydraulic systems including gas boilers or thermal solar panels. For example, the controller may anticipate the activation of any present electrical and/or hydraulic backup heat sources when it predicts that the primary heat pump system will be unable to meet the thermal demand or when predefined energy efficiency thresholds are expected to be exceeded. This implementation allows the system to proactively adjust its operation in anticipation of changing conditions, which enhances overall energy efficiency and reduces dependence on reactive control strategies. [48] In some implementations, the thermal comfort satisfaction level and the energy-related objective level are both provided in terms of energy. While thermal comfort is usually expressed in terms of delivered temperature, ability to deliver a desired volume of water at a desired temperature at a desired time, or other definitions, by expressing thermal comfort in terms of energy, it may be easier to compare and balance thermal comfort with energy consumption.

[49] In some implementations, the thermal comfort determination module, and/or the energy optimization input module, and/or the environmental determination module uses an artificial intelligence model, more particularly a machine learning model, to generate its predictions.

[50] In some implementations, the machine learning model used by the thermal comfort determination module and/or the energy optimization input module is trained using data on the operation history and/or product information of the hot water system, and/or using data on the environmental conditions outside of the hot water system.

[51] In some implementations, the machine learning model is continuously refined using data from operation of the hot water system. In some implementations, the machine learning model is refined based on occurrence of: a time condition; and/or a duration condition; and/or a processor usage condition, such as processor usage falling below a threshold value; and/or a product usage condition, such as product usage falling below a threshold value; an electricity and/or energy cost condition, such as electricity and/or energy cost falling below a threshold; and/or an event condition, such as an event notification received through the communication interface; and/or a condition related to the thermal comfort satisfaction level, such as the thermal comfort satisfaction level falling below a threshold value; and/or a condition related to the energy-related objective level, such as the energy-related objective level exceeding a threshold value; and/or a condition related to an environmental condition outside of the hot water system; and/or a deviation between the previous prediction result of the machine learning model and the current prediction result from the operation history of the hot water system. [52] In some implementations, the artificial intelligence model of the thermal comfort determination module is configured to predict the thermal comfort satisfaction level, based, at least in part, on the following inputs: the operation history of the hot water system; and/or the product information of the hot water system; and/or the actual or predicted environmental conditions outside of the hot water system; and/or commands to operate the hot water system.

[53] In some implementations, the artificial intelligence model of the energy optimization input module is configured to predict the energy-related objective level, based, at least in part, on the following inputs: the operation history of the hot water system; and/or the product information of the hot water system; and/or the actual or predicted environmental conditions outside of the hot water system; and/or commands to operate the hot water system. In some implementations, the artificial intelligence model of the environmental determination module is configured to predict environmental conditions outside of the hot water system, based, at least in part, on the following inputs: the operation history of the hot water system; and/or the product information of the hot water system.

[54] In some implementations, the optimization operation uses an artificial intelligence model, more particularly a machine learning model. In some implementations, the optimization operation uses an artificial intelligence model, more particularly a machine learning model.

[55] In some implementations, the thermal comfort determination module, the energy optimization input module, and/or the environmental determination module may be configured to use an artificial intelligence model, more particularly a machine learning model, to generate its predictions. The model may be trained on historical system performance data (e.g., including hot water usage patterns), user feedback on comfort levels, environmental sensor data (e.g., including ambient temperature data), and operational parameters of the hot water system. An advantage of this approach is that it enables the system to learn from real- world usage behavior and environmental conditions, thereby improving the accuracy of its predictions and enhancing the responsiveness and efficiency of the control strategy. [56] In some implementations, at least portions of the controller are cloudbased. For example, any or all of the thermal comfort determination module, the energy optimization input module, the environmental determination module, the information module, and/or the optimization operation may be cloud-based or may be handled by a local controller. Similarly, the memory or processing associated with any of these modules, the optimization operation, and/or any other functions of the controller may be handled locally, or may be handled on a cloud-based system. Similarly, a physical or a virtual machine may be used to implement any of the functions of the controller, either locally or on a cloud-based system.

[57] In some implementations, the thermal comfort satisfaction level is considered achieved when at least one of the following conditions is met: a. The predicted water temperature at the point of use satisfies a user-defined comfort threshold for a predefined minimum duration, wherein the “user- defined comfort threshold” refers to a temperature value specified by the user that represents either a minimum or maximum acceptable water temperature for comfort, depending on whether the system is operating in heating or cooling mode, and the “predefined minimum duration” refers to a fixed or configurable time interval during which the predicted water temperature must remain within the comfort threshold to be considered satisfactory. b. The system is able to deliver water at a requested temperature within a predefined response time after a usage request, wherein the “predefined response time” refers to the maximum allowable time between a user- initiated water request and the system delivering water at or above the comfort threshold, and the “usage request” refers to an event indicating user intent to access water, such as opening a tap, triggering a sensor, or issuing a command via a control interface. c. The predicted availability of water at the requested temperature matches or exceeds expected usage patterns, wherein “expected usage patterns” refer to anticipated times and volumes of water demand derived from user behavior, schedules, or learned system data. d. User feedback indicates satisfaction with recent water delivery events, wherein “user feedback” refers to explicit or implicit input from the user regarding satisfaction with water delivery, which may include manual ratings, app-based responses, or inferred signals such as repeated manual overrides. e. The estimated runtime and thermal energy exchange of the heat pump under current or forecasted operating conditions, wherein “thermal energy exchange” refers to the predicted amount of energy transferred to or extracted from the water by the heat pump, and “forecasted operating conditions” include ambient temperature, humidity, and system load. f. the anticipated activation of any present electrical and/or hydraulic backup thermal source to meet a thermal comfort threshold, wherein the thermal source may provide heating or cooling depending on the system mode.

This implementation has the further advantage that it enables the system to dynamically adapt to user needs and behavioral trends, improving comfort delivery without requiring constant manual configuration. By incorporating both predictive modeling and real-time feedback, the controller can fine-tune its operation to maintain high satisfaction levels while optimizing energy use. This results in a more responsive, efficient, and user-aligned hot water system.

[58] In some implementations, the energy-related objective level is predicted based on at least one of: a. The estimated runtime and energy consumption of the heat pump under current or forecasted operating conditions, in particular external temperature, in particular room temperature and/or outdoor temperature, humidity, time-of-day, or energy tariff periods, which may influence system behavior and energy usage. b. The anticipated activation and/or availability of any present electrical and/or hydraulic backup thermal source, in particular heat source and/or renewable source, wherein anticipated activation refers to a predicted need for supplementary heating based on system limitations, demand surges, and/or energy optimization strategies. c. The thermal load required to meet predicted water demand, wherein thermal load refers to the amount of energy required to raise the temperature of a given volume of water to a desired comfort level, accounting for factors such as inlet water temperature, usage volume, and timing. d. System efficiency metrics derived from historical performance data, wherein system efficiency metrics refer to performance indicators such as coefficient of performance (COP), seasonal coefficient of performance (SCOP), energy factor (EF), or other learned or calculated values that reflect the energy efficiency of the hot water system over time, and may be derived from sensor data, usage logs, or predictive models.

This implementation has the further advantage that it enables the controller to anticipate energy demand with further enhanced precision, allowing for proactive adjustments that reduce unnecessary energy consumption. By incorporating for example real-time and/or historical data, the system can optimize operation schedules, minimize reliance on less efficient backup heat sources, and align energy usage with favourable tariff periods or renewable energy availability. This not only improves overall system efficiency but also supports cost savings and environmental sustainability, while maintaining user comfort as a primary objective.

[59] In some implementations, the controller (100, 210) is configured such that the predicted thermal comfort satisfaction level is considered valid when a predicted thermal load and/or system response fall within predefined operational tolerances.

The term “predicted thermal load” refers to the estimated amount of energy required to meet anticipated hot water demand, based on factors such as inlet water temperature, target comfort temperature, usage volume, and timing.

The term “system response” refers to the ability of the hot water system to react to a usage request within expected performance parameters, including but not limited to heat-up time, flow rate, and delivery stability.

The term “predefined operational tolerances” refers to acceptable ranges or thresholds for thermal and temporal performance metrics, which may be configured by the user, learned from historical system behavior, or derived from manufacturer specifications.

This implementation has the further advantage that it enables the controller to validate thermal comfort predictions against real-world system behavior, ensuring that optimization decisions are grounded in operational feasibility. By enforcing tolerance checks, the system can avoid over-optimistic or unrealistic predictions, thereby improving reliability, reducing the risk of underperformance, and enhancing user trust in automated control decisions. This validation mechanism also supports adaptive learning by flagging deviations that may indicate system degradation, sensor drift, or changing usage patterns.

[60] In some implementations, the optimization operation comprises a safeguard mechanism that ensures minimum thermal comfort is maintained. The thermal comfort threshold is defined as a predefined or user-configurable limit that guarantees hot water delivery for at least basic hygiene or domestic use, regardless of energy optimization outcomes.

The term “minimum thermal comfort” refers to a baseline level of hot water availability and temperature that is sufficient to support essential activities such as handwashing, showering, or dishwashing. This threshold may be fixed by system defaults or adjusted by the user based on personal or cultural expectations of comfort.

This implementation has the further advantage that it prevents the system from compromising user well-being in pursuit of energy savings. By embedding a safeguard mechanism, the controller ensures that even under aggressive optimization scenarios — such as during high energy tariffs or limited renewable availability — basic comfort needs are always met. This not only enhances user trust and system reliability but also supports compliance with health and safety standards in residential or shared living environments.

[61] In some implementations, the controller ensures support for at least basic hygiene needs by maintaining the availability of hot water at or above 40 °C for a cumulative duration of a pre-defined time, such as at least 10 minutes per day. This threshold is intended to guarantee the minimum thermal energy required to enable essential personal and household hygiene activities, such as handwashing, dishwashing, or short-duration personal cleaning. This service level is maintained regardless of energy optimization outcomes, ensuring that user health and hygiene are not compromised even under constrained energy conditions.

[62] In some implementations, the optimization operation comprises a filtering mechanism that compares predicted energy availability with required energy demand during expected water usage events, and adjusts system operation accordingly.

The term “predicted energy availability” refers to the estimated amount of energy accessible to the system during a given time window, based on factors such as grid supply forecasts, renewable energy production estimates (e.g., solar or ambient sources), and tariff schedules.

The term “required energy demand” refers to the amount of energy needed to meet anticipated water usage, as determined by predicted thermal load, user comfort thresholds, and system efficiency.

The term “expected water usage events” refers to time periods during which water demand is likely to occur, based on historical usage patterns, user-defined schedules, or learned behavior.

This implementation has the further advantage that it enables the system to proactively align energy consumption with availability, thereby reducing reliance on high-cost or carbon-intensive energy sources. By filtering and adjusting system operation in real time, the controller can prioritize energy-efficient heating windows, avoid unnecessary standby losses, and maintain comfort without overprovisioning. This contributes to both operational sustainability and economic efficiency, especially in environments with variable energy pricing or intermittent renewable supply.

[63] In some implementations, the term “energy availability” refers to the presence of usable energy from one or more sources that the system can access to perform heating or cooling operations. This may include grid availability, such as the presence of electrical power from a public or private utility; renewable energy availability, such as thermal or electrical energy derived from solar, ambient air, or geothermal sources; and stored energy, such as energy retained in thermal storage tanks or battery systems. The controller may assess energy availability either in real time or based on forecasts, and may use this information to prioritize energy sources or adjust operational strategies accordingly.

[64] In some implementations, the optimization operation comprises a scheduling strategy that anticipates temperature setpoints based on user-defined comfort profiles, in particular day and night comfort profiles.

The term “user-defined comfort profiles” refers to configurable temperature preferences that reflect the user’s desired hot water availability and comfort levels during specific time intervals. In particular, “user-defined day and night comfort profiles” refer to preferences associated with periods such as morning routines, daytime absence, evening use, or overnight standby. These profiles may be manually set by the user or learned over time based on observed usage behaviour.

The scheduling strategy may incorporate these profiles to proactively adjust heating cycles, ensuring that hot water is available when needed while minimizing energy use during low-demand periods.

This implementation has the further advantage that it enables the system to align hot water production with actual user routines, thereby improving both comfort and energy efficiency. By anticipating demand rather than reacting to it, the controller can reduce unnecessary standby heating, avoid peak energy costs, and ensure that hot water is delivered precisely when expected. This contributes to a more intelligent and user-adaptive control strategy that balances convenience with sustainability.

[65] In some implementations, the scheduling strategy dynamically adjusts heating or cooling operations based on learned user behaviour, environmental conditions, and/or historical usage patterns. Rather than relying solely on fixed time schedules, the controller thus continuously refines its operational plan by analysing when and how thermal energy is typically required. This may include identifying recurring usage windows, such as morning or evening routines, and preconditioning the system accordingly to ensure comfort and efficiency. The scheduling logic may also incorporate external factors such as weather forecasts, energy availability, or tariff structures to optimize timing. As a result, the system becomes increasingly responsive to actual demand overtime, improving both user satisfaction and energy performance without requiring manual reconfiguration.

[66] In some implementations, the optimization operation comprises delaying or advancing system activation based on external conditions, in particular external temperature forecasts, in particular room temperature and outdoor temperature, and the thermal inertia of the system to reduce energy consumption while maintaining acceptable comfort levels.

The term “external temperature forecasts” refers to predicted environmental temperature values over a future time horizon, which may be obtained from external weather services, on-site sensors, or building management systems.

The term “thermal inertia of the system” refers to the system’s capacity to retain and gradually release thermal energy over time, which depends on factors such as tank insulation, water volume, material properties, and prior heating cycles.

This implementation has the further advantage that it allows the controller to shift energy-intensive operations to periods of lower ambient heat loss or higher energy availability, thereby reducing peak load and improving overall system efficiency. By leveraging thermal inertia, the system can preheat water in advance of demand or delay heating without compromising user comfort, enabling smarter load balancing and better integration with renewable energy sources or dynamic pricing models.

[67] In some implementations, the optimization operation comprises a balanced consideration of thermal comfort and energy-related objectives, the balance being dynamically adjusted based on user preferences, system learning, or external conditions.

The term “balanced consideration” refers to a control strategy in which the controller evaluates both thermal comfort and energy-related objectives concurrently, assigning dynamic importance to each based on contextual factors. Unlike fixed-weight optimization, balanced consideration allows the controller to adaptively prioritize comfort or efficiency depending on user-defined preferences (e.g., comfort priority during morning routines), learned behavior (e.g., recurring usage patterns), or external signals (e.g., energy tariffs, weather forecasts, or grid constraints). This implementation has the further advantage that it enables the controller to respond intelligently to evolving user expectations and environmental conditions without requiring manual reconfiguration. By dynamically adjusting the balance between comfort and efficiency, the system can maintain optimal performance across diverse usage scenarios, improving both user satisfaction and operational sustainability.

[68] In some implementations, the optimization operation comprises a weighted balancing of thermal comfort and energy-related objectives. The term “weighted balancing” refers to a decision-making process in which multiple objectives — such as maintaining user comfort and minimizing energy consumption — are assigned relative importance values (weights), which influence the outcome of the optimization.

These weights may be dynamically adjusted for example based on:

• “User preferences,” such as prioritization of comfort versus cost savings;

• “System learning,” which includes adaptive behaviour derived from historical usage and performance data (for example hot water usage patterns); or

• “External conditions,” such as energy tariffs, weather forecasts, or grid constraints.

The optimization objective can be mathematically expressed as:

Objective = w_c ■ C + w_E ■ (1 — E)

Wherein C represents the thermal comfort satisfaction level, for example in the form of a numerical score, E represents the energy-related cost or impact score, and wc and WE are the respective weights assigned to comfort and energy objectives. The weights may be normalized such that WC+WE = 1 , and the system seeks to maximize the overall objective. This implementation has the further advantage that it enables the controller to flexibly adapt to changing user needs and environmental contexts without requiring manual reconfiguration. By continuously recalibrating the balance between comfort and efficiency, the system can deliver personalized performance while optimizing for cost, sustainability, or operational constraints. The inclusion of a formalized objective function also facilitates implementation in software-based control systems and supports transparency in decision-making logic. This dynamic responsiveness further enhances both user satisfaction and system intelligence over time.

This implementation has the further advantage that it enables the controller to flexibly adapt to changing user needs and environmental contexts without requiring manual reconfiguration. By continuously recalibrating the balance between comfort and efficiency, the system can deliver personalized performance while optimizing for cost, sustainability, or operational constraints. The inclusion of a formalized objective function also facilitates implementation in software-based control systems and supports transparency in decision-making logic. This dynamic responsiveness further enhances both user satisfaction and system intelligence over time.

[69] In some implementations, the optimization operation comprises selecting or prioritizing between the heat pump and one or more backup heat sources of the hot water system, based on a tradeoff between energy efficiency and time-to-comfort. The term “time-to-comfort” refers to the predicted time required to deliver hot water at or above a predefined comfort temperature at the point of use. This prediction may be based on system response characteristics, current water temperature, heating capacity, and thermal inertia. The selection or prioritization process may consider the relative energy efficiency of each available heat source, including the heat pump and any electrical or hydraulic backup systems, and determine which source or combination of sources can meet the comfort requirement within the shortest acceptable time while minimizing energy consumption. This implementation has the further advantage that it enables the system to dynamically balance responsiveness and efficiency, ensuring that hot water is delivered promptly when needed without defaulting to high-energy backup sources unless necessary. By quantifying and comparing time-to-comfort across available heating options, the controller can make intelligent, context-aware decisions that improve user satisfaction while reducing operational costs and environmental impact.

[70] In some implementations, the optimization operation logic employs a set of weights that influence the prioritization of competing objectives, such as thermal comfort, energy efficiency, and cost. These weights may be user- configurable, system-learned, or both. For example, the system may dynamically adjust the weighting factors based on user preferences, observed behavioral patterns, or external conditions such as ambient temperature forecasts or time-of- use energy tariffs. This allows the controller to adapt its decision-making strategy over time, aligning system behavior with evolving user needs and environmental constraints.

[71] In some implementations, the optimization operation comprises selecting or prioritizing between the heat pump and one or more backup thermal sources, in particular heat sources, of the hot water system, based on a tradeoff between energy efficiency and time-to-comfort as a component of thermal comfort, wherein time-to-comfort refers to the predicted time required to deliver water at a predefined comfort temperature at the point of use.

This prediction may be based on system response characteristics, current water temperature, heating capacity, and thermal inertia. The selection or prioritization process may consider the relative energy efficiency of each available thermal source, including the heat pump and any electrical or hydraulic backup systems, and determine which source or combination of sources can meet the comfort requirement within the shortest acceptable time while minimizing energy consumption.

This implementation has the further advantage that it enables the system to dynamically balance responsiveness and efficiency, ensuring that hot water is delivered promptly when needed without defaulting to high-energy backup sources unless necessary. By quantifying and comparing time-to-comfort across available heating options, the controller can make intelligent, context-aware decisions that improve user satisfaction while reducing operational costs and environmental impact.

[72] In some implementations, the controller may employ a weighted cost function or decision rule to dynamically balance the competing objectives of thermal comfort and energy efficiency within the optimization operation. This tradeoff mechanism is designed to adapt to varying user demand profiles and operational contexts. When user demand is high — such as during peak usage periods or when immediate hot water delivery is anticipated — the controller increases the weight assigned to minimizing time-to-comfort. This results in system behavior that prioritizes rapid thermal response, potentially by activating auxiliary heating elements or adjusting compressor operation to accelerate water heating. Conversely, when demand is low, delayed, or uncertain, the controller shifts emphasis toward minimizing energy consumption. In such cases, the system may operate at lower power levels, defer heating cycles, or exploit ambient thermal conditions to improve efficiency. The weighting factors may be predefined, user-configurable, or dynamically adjusted based on predictive analytics, historical usage patterns, or external signals such as time-of-use tariffs. This adaptive control strategy enables the system to intelligently navigate the tradeoff space between comfort and efficiency, ensuring optimal performance across a range of real-world scenarios.

[73] In some implementations, the thermal comfort satisfaction level and/or the energy-related consumption level is represented as a numerical score.

The term “numerical score” refers to a scalar value that quantifies the degree to which the predicted or actual hot water delivery aligns with user-defined comfort expectations. This score may be computed using predictive models, sensor data, or historical feedback, and is typically normalized to a defined range (e.g., 0 to 1 or 0 to 100) to facilitate comparison, thresholding, and trend analysis.

This implementation has the further advantage that it enables the system to express comfort outcomes in a structured, machine-readable format. This facilitates integration with optimization algorithms, supports real-time decisionmaking, and allows for continuous monitoring and improvement of comfort delivery performance.

[74] In some implementations, the score representing the thermal comfort satisfaction level corresponds to a similarity metric or deviation between the predicted thermal comfort satisfaction level and a predefined or expected satisfaction threshold.

The term “similarity metric” refers to a mathematical function — such as cosine similarity, Euclidean distance, or absolute deviation — that quantifies how closely the predicted comfort score matches a target value. The “predefined or expected satisfaction threshold” may be a static value configured by the user or dynamically learned from historical comfort ratings, behavioral patterns, or contextual factors such as time of day or usage frequency. This implementation has the further advantage that it allows the system to detect and respond to deviations from expected comfort levels with high precision. By continuously comparing predicted outcomes to target thresholds, the controller can trigger corrective actions, refine predictive models, and maintain a high standard of user satisfaction even under varying operating conditions.

[75] In some implementations, the controller is further configured to use predictive models to estimate thermal comfort satisfaction, energy consumption, environmental conditions, or other relevant parameters. These predictions may be used to optimize control decisions in accordance with user preferences. This enables the system to anticipate user needs and system behavior, thereby improving the balance between comfort and energy efficiency without requiring manual intervention or static rule definitions.

[76] In some implementations, the controller is further configured to comprise a model refinement module, which is a software and/or hardware component that updates internal models used for prediction and control, based on new data collected during operation. These models, referred to as predictive models, are mathematical or algorithmic representations of system behavior used to forecast future states such as temperature, energy use, or comfort levels. The controller uses these refined predictive models to optimize system performance by balancing thermal comfort and energy consumption. Thermal comfort, as used throughout this application, means a subjective measure of user satisfaction with the temperature and humidity conditions in a space, often influenced by personal preferences and environmental factors. Energy consumption refers to the amount of electrical or thermal energy used by the system over a given period. The optimization process also considers environmental conditions, which include external or internal factors such as ambient temperature, humidity, or solar gain that influence system performance. The controller is further configured to adapt its behavior to installation-specific and user-specific requirements. Installationspecific means tailored to the physical configuration, insulation, and usage patterns of a particular system installation, while user-specific refers to adaptation based on the preferences, schedules, and comfort expectations of individual users. This embodiment enables the controller to deliver a personalized and efficient operation, improving both comfort and energy efficiency in a dynamic and responsive manner.

[77] In some implementations, the controller is further configured to be operatively connected to a communication interface that facilitates data exchange with one or more components of the hot water system.

The communication interface may include hardware and/or software elements and support various communication protocols, such as wired (e.g., Modbus, CAN) or wireless (e.g., Wi-Fi, Zigbee, Bluetooth) technologies. This interface enables the controller to receive sensor data, transmit control commands, and coordinate the operation of system components including heat pumps, backup heaters, valves, and actuators. In certain configurations, the communication interface may also support bidirectional communication to enable feedback-based control, diagnostics, or remote updates.

[78] In some implementations, the controller is configured to generate control commands that adjust the operation of the hot water system based on optimization logic, user preferences, or learned behavior.

These control commands may include instructions to activate or deactivate heat sources, switch between heating modes, modulate thermal energy output or energy consumption, or modify temperature setpoints.

They may also include instructions to initiate specific operating modes, such as energy-saving or comfort-priority modes.

The commands may be issued periodically, in response to events, or based on predictive models. Over time, the controller may refine these commands by learning from historical usage patterns, manual interventions, or system performance feedback.

[79] In some implementations, the disclosed technology provides a hot water system including a heat pump and an optional backup electrical heat source and/or an optional hydraulic backup heat source, the hot water system controlled by the controller of the disclosed technology. [80] In some implementations, the hot water system is configured for use in cooling applications.

[81] In some implementations, the hot water system the hot water system is further configured to operate the hot water system in a cooling mode, wherein the system extracts thermal energy from the delivery point to deliver water at or below external temperature, such that the hot water system delivers water at a temperature at or lower than ambient temperature for cooling applications.

[82] According to another aspect of the invention, a method is provided for controlling a hot water system comprising a heat pump and optionally an electrical and/or hydraulic backup thermal source, in particular heat source. The method comprises: a. storing, in a memory, instructions executable by a processor; b. accessing, by the processor, data and instructions from the memory; c. receiving, via a communication interface communicatively coupled to the processor and the hot water system, data from the hot water system; d. predicting, by a thermal comfort determination module, a thermal comfort satisfaction level delivered by the hot water system based and/or trained on at least one of:

• a predicted water temperature at the point of use;

• a predicted availability of water during expected usage periods;

• historical user feedback or usage patterns;

• system response time to a water demand; e. predicting, by an energy optimization input module, an energy-related objective level of the hot water system based and/or trained on at least one of:

• a predicted activation duration and/or frequency of the heat pump;

• historical usage patterns and demand forecasts;

• external temperature, in particular room temperature or outdoor temperature, and system thermal losses; • at least one energy tariff, C0₂ footprint, and/or at least one cost model associated with a heat source or thermal exchange system; f. performing, by the processor, an optimization operation to determine a control strategy for the hot water system, the optimization being based on a weighted consideration of:

• a thermal comfort objective; and

• at least one energy-related objective, including cost, environmental impact, or system efficiency; wherein the optimization operation determines whether an adjustment of the operation of the hot water system is required and if so, defines such adjustment based at least in part on the predicted thermal comfort satisfaction level and the predicted energy-related objective level, such that the adjustments satisfy a weighted combination of the thermal comfort and energy-related objectives; g. determining, by the processor, commands to operate the hot water system based and/or trained on a result of the optimization operation, the commands comprising at least one of scheduling, source selection, or control command.

[83] According to another aspect of the invention, a data processing device is provided, comprising means for carrying out the method described above for controlling a hot water system.

[84] According to another aspect of the invention, a computer program product is provided, comprising instructions to cause the hot water system to execute the steps of the method described above.

[85] According to another aspect of the invention, a computer-readable data carrier is provided, having stored thereon the computer program product comprising instructions to execute the method for controlling the hot water system.

[86] According to another aspect of the invention, a data carrier signal is provided, carrying the computer program product comprising instructions to execute the method for controlling the hot water system.

[87] According to another aspect of the invention, the controller is used for controlling a hot water system as described above. In this use, the controller executes an optimization operation based on predicted thermal comfort satisfaction and energy-related objective levels, and determines control commands for the hot water system based on the result of the optimization. The control commands comprise scheduling and source selection decisions that account for time-of-day preferences, availability of renewable or lower-cost energy sources, and system performance constraints.

[88] In the context of the present specification, unless expressly provided otherwise, the words “first”, “second”, “third”, etc. have been used as adjectives only for the purpose of allowing for distinction between the nouns that they modify from one another, and not for the purpose of describing any particular relationship between those nouns.

[89] In the context of the present specification, unless expressly provided otherwise, directions indicated by terms such as “top”, “bottom”, “upper”, “lower”, “above”, “below”, etc., are used in their usual sense - i.e., relative to a gravitational direction or axis.

[90] Additional and/or alternative features, aspects and advantages of implementations of the present technology will become apparent from the following description, the accompanying drawings and the appended claims.

BRIEF DESCRIPTION OF THE DRAWINGS

[91] In the figures, the subject-matter of the disclosure is schematically shown, wherein identical or similarly acting elements are usually provided with the same reference signs.

[92] FIG. 1 is a block diagram of an example controller that could be used in some implementations of an energy management system.

[93] FIG. 2 is a block diagram of an example hot water system.

[94] FIG. 3 is a block diagram showing the structure of a control system for a hot water system, in accordance with the described technology.

[95] FIG. 4 is a block diagram illustrating the training and refinement of a machine learning model for use in modules of the control system for a hot water system, in accordance with the described technology. [96] FIG. 5 illustrates a method for controlling an example hot water system.

DETAILED DESCRIPTION

[97] The examples and conditional language recited herein are principally intended to aid the reader in understanding the principles of the present technology and not to limit its scope to such specifically recited examples and conditions. It will be appreciated that those skilled in the art may devise various arrangements that, although not explicitly described or shown herein, nonetheless embody the principles of the present technology.

[98] Furthermore, as an aid to understanding, the following description may describe relatively simplified implementations of the present technology. As persons skilled in the art would understand, various implementations of the present technology may be of a greater complexity.

[99] In some cases, what are believed to be helpful examples of modifications to the present technology may also be set forth. This is done merely as an aid to understanding, and, again, not to define the scope or set forth the bounds of the present technology. These modifications are not an exhaustive list, and a person skilled in the art may make other modifications while nonetheless remaining within the scope of the present technology. Further, where no examples of modifications have been set forth, it should not be interpreted that no modifications are possible and/or that what is described is the sole manner of implementing that element of the present technology.

[100] Moreover, all statements herein reciting principles, aspects, and implementations of the present technology, as well as specific examples thereof, are intended to encompass both structural and functional equivalents thereof, whether they are currently known or developed in the future. Thus, for example, it will be appreciated by those skilled in the art that any block diagrams herein represent conceptual views of illustrative systems embodying the principles of the present technology. [101] With these fundamentals in place, we will now consider some nonlimiting examples to illustrate various implementations of aspects of the present disclosure.

[102] Controller

[103] FIG. 1 depicts an example controller 100, which may be any type of computer system or embedded controller. It will be recognized that some or all the components of the controller 100 may be virtualized and/or cloud-based. As depicted, the controller 100 may include one or more processors 102, a memory 110, a storage interface 120, and a communication interface 140. These system components may be interconnected via a bus 150, which may include one or more internal and/or external buses (not shown) (e.g. a PCI bus, universal serial bus, IEEE 1394 “Firewire” bus, SCSI bus, Serial-ATA bus, etc.), to which these various hardware components may be electronically coupled.

[104] The memory 110, which may be a random-access memory or any other type of memory, may contain data 112, an optional operating system 114 (it will be understood that not all controllers require an operating system), and a program 116. The data 112 may be any data that serves as input to or output from any program in the controller 100. The operating system 114, which is optional, since not all control applications require use of an operating system, may be an operating system such as MICROSOFT WINDOWS, FreeRTOS, an operating system based on the LINUX kernel (e.g., UBUNTU, DEBIAN, FEDORA, ARCH, SUSE, etc.), or any other operating system suitable for use on a computer system or microcontroller. The program 116 may be any program or set of programs that include program instructions that may be executed by the processor to control actions taken by the controller 100. In particular, the program 116 may include program instructions that, when executed by the processor, cause the processor to carry out one or more of the methods described below.

[105] The storage interface 120 may be used to connect storage devices, such as the depicted storage device 125, to the controller 100. The storage device 125 may be a solid-state drive using an integrated circuit assembly to store data persistently. Alternatively, the storage device 125 may be a hard drive using any of a variety of types of magnetic storage media to store and retrieve digital data. As another alternative, the storage device 125 may be an optical drive, or a card reader that receives a removable non-volatile semiconductor memory card. As still another alternative, the storage interface 120 may provide a universal serial bus connection to which the storage device 125 may be hot-pluggable, and the storage device 125 may be a flash memory device (e.g., a USB thumb drive). In some implementations, in which storage of data is not necessary, the storage interface 120 and storage device 125 may optionally be omitted.

[106] In some implementations, the controller 100 may use well-known virtual memory addressing techniques that allow the programs of the controller 100 to behave as if they have access to a large, contiguous address space instead of access to multiple, smaller storage spaces, such as the memory 110 and the storage device 125. Therefore, while the data 112, the optional operating system 114, and the programs 116 are depicted as residing in the memory 110, those skilled in the art will recognize that these items may not necessarily be wholly contained in the memory 110 at the same time.

[107] The one or more processors 102 may include one or more microprocessors and/or other integrated circuits able to execute program instructions stored in the memory 110. When the controller 100 starts up, the processor(s) 102 may initially execute program instructions of a boot routine and/or the program instructions that make up the operating system 114.

[108] The communication interface 140 may be used to communicatively connect the controller 100 to other controllers, computer systems, or still other devices (not shown) via a communication channel 160. The communication channel 160 may be a serial or parallel connection, a wired, wireless, mesh or cellular network, or any other type of communication channel or combination of channels. Data and/or program instructions may be sent to the controller 100 as signals via the communication channel. The communication interface 140 may include a combination of hardware and software that enables communications on the communication channel 160. The software in the communication interface 140 may include software that uses one or more communication protocols to communicate over the communication channel 160, including and not limited to, network protocols such as TCP/IP (Transmission Control Protocol/lnternet Protocol) or Modbus.

[109] It will be understood that the depicted controller 100 is merely an example, and that the technology disclosed herein may be used with a wide variety of other controllers or computer systems, or still other computing devices having different configurations.

[110] Hot Water System

[111] FIG. 2 depicts an example hot water system 200. The hot water system 200 is generally used to provide domestic hot water to a domestic hot water installation, such as is used in a residential dwelling, but may be used in other heating and/or cooling applications, such as room heating or room cooling. However, it will be understood that the hot water system 200 may be used in other types of buildings, such as apartment buildings or other multi-dwelling buildings, office buildings, or any other type of building at which hot water systems are installed and/or controlled. It will also be understood that although referred to as a “hot water system”, the hot water system 200 may also be used in cooling applications, in which the water is cooled, e.g., for use in room cooling.

[112] As depicted, the hot water system 200 is a heat pump-based system with an optional electric resistive heating element providing backup heating. The system includes a heat pump 202 (the construction and operation of such a heat pump being well-known) and a hot water tank 204, which supplies hot water to a domestic hot water output 220, which may include (for example) hot water taps (not shown), showers (not shown), and other outlets or appliances that consume domestic hot water (not shown). The domestic hot water output 220 may mix the hot water from the hot water tank 204 with cold water to produce a range of water temperatures for domestic use.

[113] A heat exchanger 206 in the hot water tank 204 transfers heat from primary water (or other heating fluids, such as a water-glycol mixture) warmed by the heat pump 202 to the sanitary domestic hot water in the hot water tank 204. An electric resistive heating element 208 may optionally be included in the hot water tank 204 to serve as backup heating, during periods in which the heat pump 202 may be unable to provide sufficient heating to meet the demand for domestic hot water. Other backup systems, such as electric, oil, hydrogen, wood, and/or gas-powered boilers, solar thermal systems, additional heat pumps, and/or other known heating systems may optionally be used in addition to the heat pump 202 to heat the primary water (or other heating fluids) that heat the domestic hot water in the hot water tank 204. These additional heat sources may be referred to herein as hydraulic heat sources or hydraulic backup heat sources.

[114] A controller 210, which may be similar to the example controller 100, described above, controls the heat pump 202, the optional electric resistive heating element 208, any optional hydraulic heat sources (not shown), and/or other devices and actuators that control the flow and mixing of domestic hot water through the hot water system 200. The controller 210 may be connected to various sensors (not shown) that provide information on, e.g., water temperatures at various points in the system, ambient air temperature, and other measurements that may be used (as described below) to control the hot water system 200. The controller 210 may also be connected to a cloud-based system 250, which may provide access to databases, updates, commands from users, remote commands or information, external services, cloud-based storage, or other information sources and services that may be accessed, e.g., over the Internet. For example, the cloud-based system 250 may be used by the controller 210 to access weather predictions that may be used by the controller 210. In some implementations, most of the control of the hot water system 200 may be carried out in the cloud-based system 250, with the controller 210 serving primarily to communicate commands to the hot water system 200, and as a backup controller if there are communication problems that prevent communications between the cloud-based system 250 and the controller 210.

[115] The controller 210 may communicate with various system components in a variety of ways. For example, the controller 210 may communicate wirelessly with sensors, actuators, hydraulic heat sources, and/or other system components using a wireless communication protocol, such as WIFI, Bluetooth, or Zigbee. In some implementations, they may communicate over a wired bus, using protocols such as OpenTherm. Some devices, such as sensors, may be directly connected to the controller 210. For some devices, including older boilers, etc., the controller 210 may be limited to controlling the device using on/off signals. In some implementations, the controller 210 may use numerous types of communication, depending on the devices that are being controlled.

[116] In some implementations, the heat pump 202 may be built within a single housing (not shown) that also houses the hot water tank 204, the controller 210, and the optional electric resistive heating element 208, to provide a self- contained water heating appliance. In some implementations, such a self- contained water heating appliance may also include an optional backup hydraulic heat source, such as a boiler within the single housing or as an attachment to the housing. In some implementations, the heat pump 202 may more directly heat the domestic hot water in the hot water tank 204, without use of primary water or the heat exchanger 206 (e.g., the water in the hot water tank 204 is heated through heat exchange with the refrigerant circuit of the heat pump 202).

[117] It will be understood that although the example hot water system 200 shown in FIG. 2 is configured to provide domestic hot water, other configurations could be used to provide hot water for heating and/or cooling applications (e.g., through radiators, underfloor or ceiling heating or cooling loops, etc.). Such systems may include, e.g., various load circuits and pumps, to provide hot or cool water to heating or cooling zones in a house. Other hot water systems 200 may provide hot water for both heating/cooling and domestic hot water uses. Such systems may include load circuits that include domestic heating, such as through radiators, and at least one load circuit that includes a hot water tank for domestic hot water, such as is shown in FIG. 2. While the technology is discussed primarily in terms of control for providing domestic hot water, it will be understood that substantially the same optimization techniques, balancing thermal comfort satisfaction with energy consumption, could be used in domestic heating systems, and/or in systems that provide both domestic heating/cooling and domestic hot water.

[118] Intelligent Control

[119] FIG. 3 shows the structure of an “intelligent” control system for a hot water system such as is shown in FIG. 2. The modules and other elements that make up the intelligent control system described with reference to FIG. 3 may be operated as a program on the controller 210 shown in FIG. 2, may be operated on the cloud-based system 250, may be operated on separate controllers, which may be hardware-based, virtual, or cloud-based (not shown in FIG. 2), or may be distributed between any combination of these.

[120] As shown in FIG. 3, the intelligent control system 300 includes a thermal comfort determination module 302, an energy optimization input module 304, an environmental determination module 306, an information module 308, and a single user-determined parameter 310. An optimizer 320 uses information and predictions provided by these modules to generate instructions to control the hot water system 200 to achieve a balance of thermal comfort and energy consumption influenced by the preferences of the system’s user, as specified using the single user-determined parameter 310.

[121] The intelligent control system 300 is structured as several determination modules and an optimization system that attempts to balance thermal comfort with energy consumption using information provided by these determination modules. Generally, thermal comfort and energy consumption are linked and dependant. It is necessary to consume energy to produce thermal comfort, but there may be flexibility with respect to how and when to consume energy to produce heat (e.g., in terms of which heat producers to use, performance for heat pumps based on weather conditions, prices for various types of energy, CO2 impact, etc.). Thus, the aim of the intelligent control system 300 is to: balance thermal comfort and energy consumption to meet a user’s objective and preferences; determine which heating device to use; and determine how to use the available heat producers, depending on their energy consumption (taking into account their energy performances, their energy costs, weather conditions that may affect their performance, and hot water usages).

[122] The intelligent control system 300 is driven by the optimizer 320, which attempts to determine “optimal” commands for controlling the hot water system 200. The degree to which a set of commands is “optimal” depends on balancing thermal comfort and energy consumption, as described below. This involves generating a prediction of the thermal comfort satisfaction level and the energy- related objective level that will result from any given set of commands, by sending potential commands, as well as other inputs to the thermal comfort determination module 302, to obtain a prediction of the thermal comfort satisfaction level that will result from the commands, and to the energy optimization input module 304, to obtain a prediction of the energy-related objective level that will result from the commands. A function that balances these predictions may then be applied to determine the “fitness” of the potential commands. This fitness function is influenced by the single user-determined parameter 310, which specifies, in a single parameter, the how much importance the user wants to give to thermal comfort satisfaction level vs. energy-related objective level in determining this balance. The optimization can proceed by testing commands to find a set of commands that provides the best fitness score according to this fitness function.

[123] In some implementations, the single user-determined parameter 310 may be normalized to a value between 0 and 1 , where a value of 0 gives maximum importance to thermal comfort satisfaction level, and a value of 1 gives maximum importance to energy-related objective level (i.e., consuming as little energy, or energy cost as possible). The predicted values for the thermal comfort satisfaction level provided by the thermal comfort determination module 302 and the energy- related objective level provided by the energy optimization input module 304 may also be normalized to valued between 0 and 1 . For example, a value of 0 for the thermal comfort satisfaction level would represent minimal thermal comfort, while a value of 1 for thermal comfort satisfaction level would represent “perfect” thermal comfort satisfaction. Similarly, an energy-related objective level of 0 would represent maximum energy inefficiency or cost, while an energy-related objective level of 1 would represent minimal energy inefficiency or cost. In such a system, one possible fitness function would be: fitness(x) = Comfort * (1 — x) + Consumption * x

Where: x is the single user-determined parameter 310, Comfort is the predicted thermal comfort satisfaction level, and Consumption is the predicted energy- related objective level. While this provides a useful example of a fitness function for use by the optimizer 320, other fitness functions could be used. It will also be understood that the single user-determined parameter 310 may be used in a variety of ways in such fitness functions, and need not always be used as a mathematical weight in the function.

[124] The single user-determined parameter 310 provides an easy to use means for a user of the system to choose and orient the behavior and aims of the intelligent control system 300, to determine the desired balance between energy consumption/savings and thermal comfort. This convenient single parameter simplifies the configuration/parametrization of the product control for the user, and also quantitatively expresses the desired balance between these two concurrent objectives.

[125] In some implementations, both thermal comfort and energy consumption may be expressed in terms of energy, so that they may be more readily compared and balanced. In some implementations, both levels can be converted to a notion relative to an energy, such as an energy cost. This may provide measures that are practical for a user who is interested in obtaining a balance between thermal comfort and energy cost. Expressing a notion such as thermal comfort in terms of energy is discussed further below.

[126] The optimizer 320 may use any known optimization techniques or algorithms to generate commands for the hot water system 200. For example, techniques including, e.g., simplex methods, evolutionary algorithms, genetic algorithms, greedy algorithms, heuristic and meta-heuristic methods, and/or rulebased algorithms may be used. For example, a “greedy” or iterative optimization algorithm may iteratively generate command sets, generate predicted thermal comfort satisfaction levels and predicted energy-related objective levels for those command sets to use the fitness function to evaluate the command sets, and then alter a command set in the next iteration in a way that is expected to improve the fitness of command set in the next iteration. This will lead to finding at least local optima for the command sets when the alterations are no longer improving the fitness.

[127] In addition to these (or other) known optimization techniques or algorithms, known artificial intelligence and/or machine learning techniques may be used to generate command sets. For example, an artificial neural network may be trained to accept weather conditions, operation conditions, and product information on the hot water system as inputs, and generate commands as outputs. The fitness function and prediction modules could be used both to train the artificial neural network, and to refine its training as the system is used. Other artificial intelligence and/or machine learning methods (supervised and/or unsupervised), such as reinforcement learning, hidden Markov models, linear or non-linear regression, random trees, support vector machines, and/or Kalman filters could also be applied to find command sets that provide a good “fitness”. As with artificial neural networks, some of these artificial intelligence and/or machine learning models may be trained, and may improve over time as that training is refined during use of the model.

[128] In some implementations, the optimizer 320 may access predictions relating to environmental conditions such as weather information and predictions from the environmental determination module 306. This information may be passed on from the optimizer 320 to other modules, since the environmental information may affect both thermal comfort determinations and energy consumption determinations. Additionally, the information may be used by the optimizer 320 itself, for generating commands, as input to an artificial intelligence or machine learning model, and/or for training or refining such artificial intelligence or machine learning models.

[129] The optimizer 320 may also access product information about the hot water system and operation history information on the hot water system from the information module 308. The product information may include information such as the tank volume, heat pump power, maximum flow rated, location of the installation, etc. Operation history information may include recorded history information on air temperatures, water inlet temperatures, water tank temperatures, water outlet temperatures, outlet flow rates, ambient temperatures, thermal tank energy, consumed energy, producer injected energy, which heating devices were used, the commands that were used, etc. This information may be passed from the optimizer 320 to the other modules, since this product and historic information may be used to predict both thermal comfort satisfaction levels and energy-related objective levels. Additionally, the optimizer 320 may itself make use of this information, to generate commands, as input to an artificial intelligence or machine learning model, and/or for training or refining such artificial intelligence or machine learning models.

[130] It will be understood that the intelligent control system 300 shown in FIG 3 represents only one possible organization of the modules that make up the control system 300, and that other possible organizations of these modules may be used. For example, in some implementations, the optimizer 320 may use only a thermal comfort determination module 302 and an energy optimization input module 304, with information on the system and environmental conditions being considered in these modules, rather than being handled by the optimizer 320. For example, the thermal comfort determination module 302 and/or the energy optimization input module 304 may access information on environmental conditions, product information, and/or operation history directly. In implementations in which these modules are based on artificial intelligence and/or machine learning models, predictions of environmental conditions, product information, and/or operation history may be built into the model, e.g., through training.

[131] The thermal comfort determination module 302 predicts a thermal comfort satisfaction level delivered by the hot water system or water heating device. Thermal comfort is often understood in terms of the delivered temperature, but may also be expressed as the ability of the system to deliver a desired volume of water at a desired temperature (heated or cooled, for example based on a predefined temperature or a temperature defined by the user), at a desired time. This is conventionally expressed in terms of “V40”, which is the volume of water at 40° C that can be delivered by a system by mixing all water above 40° C with the inlet/domestic cold water. Using this metric, one can easily understand how much hot water can be delivered by the system. This V40 measure is conventionally used, e.g., by consumers to determine if a system is able to meet the hot water demands of a household. It should be noted that calculation of the V40 artificially "dilutes" water above 40° C so that it reaches 40° C. For example, a thermodynamic water heater delivering 200L of water at 55° C with a domestic cold temperature (DOW) of 10° C has a V40 of 300L. The value of 40° C is used because it corresponds to a standard maximum operating temperature in a domestic water system, though for purposes of comfort and other uses, other temperatures could be used (e.g., V50, using 50° C).

[132] While V40 is a useful measure of comfort, it may not be adequate for purposes of predicting a comfort satisfaction level. Such a calculation needs to take into account that different amounts of hot water may be needed at different times, or that different minimal hot water temperature (e.g. a V50 is not comparable with a V40) may be needed at different times. For example, during certain morning periods, when much of a household is showering, or during evening periods when dishes are being washed, greater amounts of hot water may be needed than during mid-day periods, when members of the household are away at work, or during periods in the middle of the night, when household members are asleep.

[133] Additionally, because the disclosed technology is directed to balancing thermal comfort satisfaction with energy consumption, it is useful to express thermal comfort in terms of energy, rather than in terms of a temperature or in terms of a volume of water at a particular temperature. The water volume at a particular temperature (compared to a base temperature) can be expressed as an amount of energy. Thus, thermal comfort may be understood as the ability to deliver the desired energy of hot water at the desired time.

[134] For purposes of the thermal comfort determination module 302, there is perfect thermal comfort if all present and future actual usages of hot water are thermally satisfied - i.e., the user is able to draw the amount of thermal energy from the tank to provide the desired volume of water at the desired temperature up to a given flowrate. The closer the system is able to meet this full amount of thermal energy, the higher the thermal comfort satisfaction level. Because the perceived level of comfort is inexact, a high but imperfect thermal comfort satisfaction level may be good enough to satisfy most users.

[135] Based on the above, to predict the thermal comfort satisfaction level at the present or at a future time, the patterns of hot water usage may be used or predicted, as well as the time to provide the needed thermal energy to the hot water tank (which will depend on, e.g., the water heating that is used, loss of heat from the tank, the ambient temperature of heat pump’s source medium (air, water), etc.). The patterns of hot water usage will depend on the household in which the system is installed. These patterns of usage can be learned over time by, e.g., an artificial intelligence or machine learning-based model. The ability of the system to provide particular amounts of thermal energy as hot water over a particular period of time will depend on factors such as product information (e.g., tank volume, heat pump power), environmental conditions outside of the hot water system (e.g., ambient temperature, outdoor temperature, cold water temperature), and commands that have been sent to, e.g., set up operational parameters of the heating device(s), such as the power at which the heat pump is set, the availability and power at which any backup heat sources are set, and so on.

[136] Thus, the thermal comfort determination module 302 may be understood to take inputs including operation history information, product information, external environmental conditions, and commands to operate the hot water system. The thermal comfort determination module 302 then uses these inputs to predict a present or future thermal comfort satisfaction level that will be achieved using these commands. This prediction may be made using an artificial intelligence or machine learning model that has been trained on, e.g. usage patterns and/or other information, such as product information, operation history information, environmental conditions, etc. While an artificial intelligence and/or machine learning-based approach to predicting the thermal comfort satisfaction level may be used, the thermal comfort determination module 302 may also make use of more conventional models of the performance of heating systems.

[137] It will be understood that the thermal comfort determination module 302 as described above produces a measure of thermal comfort satisfaction that is appropriate for hot water systems that provide domestic hot water. A hot water system that provides hot water for use in home heating systems including, e.g., radiators and underfloor or ceiling heating loops, may use somewhat different measures of thermal comfort satisfaction. Similarly, a cool water system that provides cool water for use in home cooling systems including, e.g. radiators or underfloor or ceiling cooling loops, may use somewhat different measures of thermal satisfaction. While the levels may be somewhat different, the same methods as described above may be applied. Thermal comfort will can still be measured in terms of availability of sufficient volumes of hot/cool water, at sufficient temperatures, at the correct times to meet user demand. This measure can still be expressed in terms of energy, to make it easier to balance against energy consumption. Additionally, this can still be predicted using, e.g., machine learning models, albeit with somewhat different training than would be used for a system predicting thermal comfort satisfaction levels for domestic hot water. It will also be understood that a single model that is capable of handling thermal comfort measures for multiple applications, such as domestic hot water, heating, and/or cooling could also be trained and used, applying substantially the same methods as are described above.

[138] The energy optimization input module 304 predicts a present or future energy-related objective level for the hot water system. The energy consumption will depend on the characteristics of the heat pump, external conditions that affect the efficiency of the heat pump, such as weather or ambient temperature, the mix of backup heating sources that are used, such as an electric resistive heating element, a gas boiler, etc. It should be noted that the energy consumption may come in various forms. For example, a heat pump and an electric resistive heating element will use electricity, while a gas boiler will use gas (e.g., propane) to generate heat. In some implementations, energy costs are used to combine the energy consumption of these various forms of energy into a single energy consumption value. This has the advantage of providing energy consumption values that reflect the cost of the energy, so that optimization of energy consumption effectively optimizes cost.

[139] Thus, the energy optimization input module 304 takes inputs such as the operation history of the hot water system, product information on the hot water system, actual or predicted environmental conditions outside of the hot water system, and/or commands to operate the hot water system. Based on these inputs, the energy optimization input module 304 predicts a current or future energy consumption for use in optimization.

[140] Because of the various factors that go into predicting the present and/or future energy consumption, in some implementations, an artificial intelligence or machine learning model may be used, trained on past energy consumption of the system, given the operation history of the hot water system, product information on the hot water system, actual or predicted environmental conditions outside of the hot water system, and/or commands to operate the hot water system. Conventional models of the energy performance of heating systems may also be used for predicting the energy-related objective level.

[141] It will also be understood that the energy-related objective level predicted by the energy optimization input module 304 may represent slightly varying measures of energy consumption. For example, in some implementations, the energy-related objective level may represent energy efficiency. In some implementations, the energy-related objective level may represent energy cost or energy savings.

[142] The environmental determination module 306 predicts present or future environmental conditions outside of the hot water system. These environmental conditions may include local weather, such as temperature, humidity, solar irradiance, wind, and so on. These conditions may affect the operation and efficiency of various heating systems. For example, the efficiency of an air-source heat pump will depend on the temperature of the source air. If this air is being drawn from outdoors, this will depend on the outdoor temperature. Solar irradiance will affect the operation of a solar thermal system that may be used as a heat source in the system. Of course, in some systems, such as heat pump water heaters located in a basement or utility room and drawing air from the room in which they are installed, the ambient temperature of the basement or utility room may affect the efficiency of the heat pump and the heat loss from the hot water tank.

[143] For many of these environmental conditions, predictions may be obtained from cloud-based weather services, if the location of the system is known. For other environmental conditions, such as ambient temperature in a basement or utility room, the environmental conditions can be obtained by sensors connected to the hot water system or from a home automation system. The data may be collected immediately for present environmental conditions, or may be based on data collected from the site of the system over time, for future environmental condition predictions. In some cases, an artificial intelligence or machine learning model trained on the conditions surrounding the system over time may be used in the environmental determination module 306.

[144] Generally, the environmental determination module 306 may accept inputs such as the time, date, location of the hot water system, and/or other information on the operation history and/or product information on the hot water system, and may output current or predicted future environmental conditions that may affect the operation of the hot water system.

[145] The information module 308 stores and provides product information on the hot water system 200, and on its operation history. This information is stored by the information module 308 during system installation and operation, and may be provided by the information module 308 as the information is requested, e.g., by the optimizer 320 and/or by other modules of for other purposes. The product information may include information such as the tank volume, heat pump power, maximum flow rated, location of the installation, etc. Operation history information may include recorded history information on air temperatures, water inlet temperatures, water tank temperatures, water outlet temperatures, outlet flow rates, ambient temperatures, thermal tank energy, consumed energy, producer injected energy, which heating devices were used, the commands that were used, etc.

[146] It will be understood that this information may require considerable storage capacity, depending on how much information is stored by the information module 308, and for how long a period of time. In some implementations, the information module 308 may use local storage to store the information. In some implementations, the information may be stored in a cloud-based system. In some implementations, the information module 308 may itself be cloud-based, such that both the information storage and the functions of the information module 308 are provided in the cloud.

[147] As described above, the thermal comfort determination module 302, the energy optimization input module 304, and the environmental determination module 306 may use artificial intelligence and/or machine learning models to make their predictions. For example, to determine a thermal comfort satisfaction level in the thermal comfort determination module 302, an artificial neural network may be trained to accept weather conditions, operation conditions, product information on the hot water system, and proposed commands or instructions for operating the hot water system as inputs, and generate a predicted thermal comfort satisfaction level as an output. Actual measured results from applying commands may be used to train such an artificial neural network. Other artificial intelligence and/or machine learning methods (supervised and/or unsupervised), such as reinforcement learning, hidden Markov models, linear or non-linear regression, random trees, support vector machines, and/or Kalman filters could also be applied to predict a thermal comfort satisfaction level, an energy-related objective level, or environmental conditions. As with artificial neural networks, some of these artificial intelligence and/or machine learning models may be trained, and may improve over time as that training is refined during use of the model. It will also be understood that more conventional models may also be used, in which, e.g., a predicted energy-related objective level is based on mathematical models or recorded data regarding the performance of heating appliances used in the hot water system.

[148] Model Refinement

[149] As is discussed above, the models used to predict the thermal comfort satisfaction level, the energy-related objective level, and environmental conditions may be based on known machine learning techniques. Similarly, the optimizer that generates commands to operate the hot water system to balance thermal comfort with energy consumption may be based on known machine learning techniques. In many such machine learning systems, the system is initially trained, and then may be fine-tuned during operation. Such fine-tuning or refinement may be helpful in systems where the model may vary considerably from installation to installation. For example, while an initially trained model may suffice for predicting thermal comfort satisfaction levels in some households, there may be households that have demands or schedules that are outside of “normal” expectations. If the model is to achieve good results in such households, it may be useful to fine-tune or refine the model as it is used, so that it will adjust to the thermal comfort needs of such a household over time. [150] FIG. 4 is a diagram showing the training and refinement of the machine learning models that may be used in the determination modules and optimizer of the intelligent control system 300 of the present disclosure. As shown in FIG. 4, there are three phases to the training and refinement of these machine learning models: an initial training phase 402, an operation phase 404, and a refinement phase 406.

[151] The initial training phase 402 involves the initial training of the model. In some implementations, this may be done, e.g., by the manufacturer of the intelligent control system 300 and/or hot water system 200. For example, the manufacturer may use a large data set built up in its own testing to train a machine learning model, such as an artificial neural network, to predict a thermal comfort satisfaction level and/or an energy-related objective level, as discussed above. This training may be done once, for a “standard” installation of the system, and the resulting model may be deployed on systems either prior to or at the time of installation. While a model with such initial training may function reasonably well in a wide variety of installations, it may be possible to improve its functionality through fine-tuning or refinement.

[152] In some implementations, a system may be installed using a “generic” control system, and the initial training phase 402 may occur after the system has been installed. In such systems, the generic control system may operate the hot water system 200 during an initial period, while data are recorded to be used in the initial training phase 402. Once sufficient training data have been recorded (which may take weeks or months, depending on the installation), machine learning models may be trained using the recorded data. This training may be done locally, by a controller on the hot water system, or may be handled by a cloud-based service. Once the training of the machine learning models has been completed, the trained machine learning models may be deployed (e.g., through download from a cloud-based service), and the intelligent control system 300 may take over control of the hot water system 200. This permits the initial training to be customized for each installation of the hot water system, which could provide better performance after the initial training phase 402 is completed, and may reduce the need for refinement of the machine learning model. [153] During the operation phase 404, the intelligent control system 300 operates as described above. During operation, data may be collected to evaluate how well the modules of the system and the optimizer are operating. These data may be used to determine when fine-tuning or refinement of the machine learning models should be performed. The data may also be used in refining the machine learning models, on either a continuous basis, or when it is determined that such refinement may be beneficial.

[154] The refinement phase 406 uses known machine learning fine-tuning or refinement techniques to refine the models used by the intelligent control system 300. These refinement techniques will depend on the nature of the machine learning models that are being used. In some implementations, known incremental learning techniques, such as online gradient descent or online boosting, are used to update the models without requiring that they be completely retrained.

[155] In some implementations, refinement of the machine learning models is a continuous process - i.e., the refinement phase 406 occurs essentially as a part of normal operation of the system. Continuous model refinement is an ongoing process aimed at maintaining and improving the performance of a machine learning model after it has been deployed. This ensures that the model remains accurate and relevant as new data become available and as the underlying system or environment changes.

[156] In some implementations refinement of the machine learning model may be triggered by various events. The refinement phase 406 occurs when refinement is triggered. Examples of events that may result in the refinement phase 406 operating to refine one or more of the machine learning models employed by the intelligent control system 300 include a time condition and/or a duration condition - i.e., refinement occurs at predetermined times and/or after predetermined durations.

[157] In some implementations, refinement may be triggered by a processor usage condition, such as processor usage falling below a threshold value and/or a product usage condition, such as product usage falling below a threshold value. These conditions recognize that refinement of the machine learning models may consume considerable processing resources, so may be best undertaken during periods of low processor usage, low product usage, and/or low electricity/energy cost.

[158] In some implementations, refinement may be triggered by an event condition, such as an event notification received through the communication interface. These event notifications may be sent, for example, from a heating device, from an external control system, from a thermostat, from a user device (such as a hand-held device or smart phone), from a cloud-based service (operated, e.g., by the system manufacturer, installer, or maintainer), orfrom other authorized external sources.

[159] In some implementations, refinement may be triggered by a performance-based measure, such as a deviation between the previous prediction result of the machine learning model and the current prediction result from the operation history. Other examples of such performance-based measures that may trigger model refinement may include a condition related to the thermal comfort satisfaction level, such as the thermal comfort satisfaction level falling below a threshold value, and/or a condition related to the energy-related objective level, such as the energy-related objective level exceeding a threshold value.

[160] In some implementations, refinement of one or more of the machine learning models may be triggered by a condition related to an environmental condition, such as weather or temperature. Such a trigger may result from, e.g., discrepancies between a machine learning model’s prediction of environmental conditions and actual environmental conditions (e.g., as reported by a weather service, or by sensors).

[161] It will be understood that, although the embodiments and/or implementations presented herein have been described with reference to specific features and structures, various modifications and combinations may be made without departing from the disclosure. For example, it is contemplated that in some implementations, the features described above may be used in different arrangements, or in other combinations. The specification and drawings are, accordingly, to be regarded simply as an illustration of the discussed implementations or embodiments and their principles as defined by the appended claims, and are contemplated to cover any and all modifications, variations, combinations or equivalents that fall within the scope of the present disclosure.

[162] FIG. 5 illustrates a method for controlling a hot water system 200 comprising a heat pump 202 and optionally an electrical and/or hydraulic backup heat source. The method enables intelligent, adaptive control of the system by predicting user comfort and energy consumption, and optimizing system behaviour accordingly. The method comprises the following steps:

S101 : Storing instructions in a memory:

The controller 100, 210 stores executable instructions in the memory 110, which define the logic for predicting comfort and energy use, performing optimization, and issuing control commands.

S102: Accessing data from memory by a processor:

The at least one processor 102 retrieves the stored instructions and any relevant data required for executing the control logic.

S103: Receiving data from the hot water system 200 via a communication interface 140:

The controller 100, 210 receives real-time or recent data from the hot water system 200, including sensor readings, operational status, and environmental inputs, through the communication interface 140.

S104: Predicting a thermal comfort satisfaction level:

A thermal comfort determination module 302 predicts a thermal comfort satisfaction level based on one or more of:

• predicted water temperature at the point of use, which can for example be a heated or cooled water temperature as predefined or defined by the user,

• predicted availability of water during expected usage periods,

• historical user feedback or usage patterns, and

• system response time to a water demand. S105: Predicting an energy-related objective level:

An energy optimization input module 304 predicts the energy-related objective level of the hot water system 200 based on one or more of:

• predicted activation duration and frequency of the heat pump,

• operational status and contribution of backup heat sources,

• historical usage patterns and demand forecasts,

• external temperature, in particular room temperature or outdoor temperature, and system thermal losses, and

• at least one energy tariff, CO₂ footprint, and/or at least one cost model associated with a heat source or thermal exchange system.

S106: Performing an optimization operation 320:

The processor performs an optimization operation 320, for example via the optimizer 320, wherein the optimization operation determines a control strategy for the hot water system 200. The optimization is based on a weighted consideration of:

• a thermal comfort objective, and

• at least one energy-related objective, including cost, environmental impact, or system efficiency.

The optimization may optionally include scheduling adjustments that improve energy efficiency or reduce cost while maintaining acceptable comfort levels.

S107: Determining control commands:

Based on the optimization result, the at least one processor 102 determines commands to operate the hot water system 200 based and/or trained on the result of the optimization operation, the commands comprising at least one of scheduling, source selection, or control command. REFERENCE SIGNS

100 controller

102 processors

110 memory

112 data

114 operating system

116 program

120 storage interface

125 storage device

140 communication interface

150 bus

160 communication channel

200 hot water system

202 heat pump

204 hot water tank

206 heat exchanger

208 electric resistive heating element

210 controller

220 hot water output

250 cloud-based system

300 intelligent control system

302 thermal comfort determination module

304 energy optimization input module

306 environmental determination module

308 information module

310 single user-determined parameter

320 optimizer; optimization operation

402 initial training phase

404 operation phase

406 refinement phase

S101 storing instructions in a memory 5102 accessing data from memory by a processor

5103 receiving data from the hot water system via a communication interface

5104 predicting a thermal comfort satisfaction level

5105 predicting an energy-related objective level S106 performing an optimization operationSI 07 determining control commands

Claims

CLAIMS:

1 . A controller (100, 210) configured to control at least a hot water system (200), the hot water system comprising a heat pump (202) and an optional electrical backup thermal source, in particular heat source (208), and/or an optional hydraulic backup heat source, the controller (100, 210) comprising: a memory (110); a processor (102) configured to access data and execute instructions stored in the memory (110); and a communication interface (140) communicatively coupled to the processor (102) and to the hot water system (200), the communication interface configured to communicate commands and data between the controller (100, 210) and the hot water system (200); characterized in that: the controller (100, 210) further comprises a thermal comfort determination module (302) configured to predict a thermal comfort satisfaction level delivered by the hot water system (200) based and/or trained on at least one of:

• a predicted water temperature at the point of use,

• a predicted availability of water during expected usage periods,

• historical user feedback or usage patterns, or

• system response time to a water demand; and an energy optimization input module (304) configured to predict an energy- related objective level of the hot water system based and/or trained on at least one of:

• a predicted activation duration and/or frequency of the heat pump (202);

• the operational status and contribution of the optional electrical backup thermal source, in particular heat source, (208) and/or optional hydraulic backup heat source;

• historical usage patterns and demand forecasts;

• external temperature, in particular room temperature or outdoor temperature, and system thermal losses; • at least one energy tariff, C02 footprint, and/or at least one cost model associated with a heat source or thermal exchange system; and in that the memory (110) stores instructions that when executed by the processor (102) cause the processor (102) to: receive a prediction of the thermal comfort satisfaction level from the thermal comfort determination module (302); receive a prediction of the energy-related objective level from the energy optimization input module (304); perform an optimization operation (320) to determine a control strategy for the hot water system (200), the optimization being based on a balanced consideration of thermal comfort and energy-related objectives, including cost, environmental impact, or system efficiency; wherein the optimization operation (320) determines whether an adjustment of the operation of the hot water system (200) is required and if so, defines such adjustment based at least in part on the predicted thermal comfort satisfaction level and the predicted energy-related objective level, such that the adjustments satisfy a balanced combination of the thermal comfort and energy-related objectives; and that determine commands to operate the hot water system (200) based and/or trained on a result of the optimization operation (320), in particular the commands to operate comprising at least one of scheduling, source selection, or control command.

2. The controller (100, 210) of claim 1 , further characterized in that the controller (100, 210) further comprises an environmental determination module (306) configured to predict environmental conditions outside of the hot water system (200), and in that the memory (110) comprises instructions that when executed by the processor (102) cause the processor (102) to receive predicted environmental conditions from the environmental determination module (306) and to perform the optimization operation (320) further based, at least in part, on the predicted environmental conditions.

3. The controller (100, 210) of claim 1 or claim 2, further characterized in that the controller further comprises an information module (308) configured to provide operation history information and/or product information on the hot water system, and in that the memory (110) comprises instructions that when executed by the processor (102) cause the processor (102) to receive operation history information and/or product information from the information module (308) and perform the optimization operation (320) further based, at least in part, on the operation history information and/or product information.

4. The controller (100, 210) of any one of the preceding claims, further characterized in that the optimization operation (320) is influenced by a single parameter (310) set by a user, the single parameter (310) providing a user- selected balance between the thermal comfort satisfaction level and the energy- related objective level in the optimization operation (320).

5. The controller (100, 210) of any one of the preceding claims, further characterized in that the thermal comfort satisfaction level and the energy-related objective level are both provided in terms of energy.

6. The controller (100, 210) of any one of the preceding claims further characterized in that the thermal comfort determination module (302), and/or the energy optimization input module (304), and/or the environmental determination module (306) uses an artificial intelligence model, more particularly a machine learning model, to generate its predictions.

7. The controller (100, 210) of claim 6, further characterized in that the machine learning model used by the thermal comfort determination module (302) and/or the energy optimization input module (304) is trained using data on the operation history and/or product information of the hot water system (200), and/or using data on the environmental conditions outside of the hot water system (200).

8. The controller (100, 210) of claim 6 or claim 7, further characterized in that the machine learning model is continuously refined using data from operation of the hot water system (200).

9. The controller (100, 210) of claim 6 or claim 7, further characterized in that the machine learning model is refined based on occurrence of: a time condition; and/or a duration condition; and/or a processor usage condition, such as processor usage falling below a threshold value; and/or a product usage condition, such as product usage falling below a threshold value; and/or an electricity and/or energy cost condition, such as electricity and/or energy cost falling below a threshold; and/or an event condition, such as an event notification received through the communication interface (140); and/or a condition related to the thermal comfort satisfaction level, such as the thermal comfort satisfaction level falling below a threshold value; and/or a condition related to the energy-related objective level, such as the energy-related objective level exceeding a threshold value; and/or a condition related to an environmental condition outside of the hot water system (200); and/or a deviation between a previous prediction result of the machine learning model and a current prediction result from the operation history of the hot water system (200).

10. The controller (100, 210) of any one of claims 6 to 9, further characterized in that: the artificial intelligence model of the thermal comfort determination module (302) is configured to predict the thermal comfort satisfaction level, based, at least in part, on the following inputs: the operation history of the hot water system (200); and/or the product information of the hot water system (200); and/or the actual or predicted environmental conditions outside of the hot water system (200); and/or commands to operate the hot water system (200); and/or the artificial intelligence model of the energy optimization input module (304) is configured to predict the energy-related objective level, based, at least in part, on the following inputs: the operation history of the hot water system (200); and/or the product information of the hot water system (200); and/or the actual or predicted environmental conditions outside of the hot water system (200); and/or commands to operate the hot water system (200); and/or the artificial intelligence model of the environmental determination module (306) is configured to predict environmental conditions outside of the hot water system (200), based, at least in part, on the following inputs: the operation history of the hot water system (200); and/or the product information of the hot water system (200).

11. The controller (100, 210) of any one of the preceding claims, further characterized in that the optimization operation (320) uses an artificial intelligence model, more particularly a machine learning model.

12. The controller (100, 210) of any one of the preceding claims, further characterized in that an artificial intelligence model, more particularly a machine learning model, is used to determine commands to operate the hot water system (200).

13. The controller (100, 210) of any one of the preceding claims, further characterized in that at least portions of the controller (100, 210) are cloud-based.

14. The controller (100, 210) of any one of the preceding claims, wherein the predicted thermal comfort satisfaction level is considered achieved when at least one of the following conditions is met: a. the predicted water temperature at the point of use satisfies a user- defined comfort threshold for a predefined minimum duration; b. the system is able to deliver water at a requested temperature within a predefined response time after a usage request; c. the predicted availability of water at the requested temperature matches or exceeds expected usage patterns; d. user feedback indicates satisfaction with recent water delivery events; e. the estimated runtime and thermal energy exchangeof the heat pump under current or forecasted operating conditions; f. the anticipated activation of any present electrical and/or hydraulic backup heat source to satisfy a thermal comfort threshold.

15. The controller (100, 210) of any one of the preceding claims, wherein the energy-related objective level is predicted based on at least one of: a. the estimated runtime and energy consumption of the heat pump under current or forecasted operating conditions, in particular external temperature and/or weather conditions; b. the anticipated activation and/or availability of any present electrical and/or hydraulic backup thermal source, in particular heat source and/or renewable source ; c. the thermal load required to meet predicted water demand; d. system efficiency metrics derived from historical performance data.

16. The controller (100, 210) of claim 14 or 15, wherein the predicted thermal comfort satisfaction level is considered valid when a predicted thermal load and/or system response fall within predefined operational tolerances.

17. The controller (100, 210) of any one of the preceding claims, wherein the optimization operation (320) comprises a safeguard mechanism that ensures minimum thermal comfort is maintained, or maximum thermal comfort is not exceeded in cooling applications, in particular wherein the thermal comfort threshold is defined as a predefined or user-configurable limit that guarantees water delivery for at least basic hygiene or domestic use, or cooling comfort, regardless of energy optimization outcomes, and wherein the safeguard mechanism is optionally influenced by the single user-determined parameter.

18. The controller (100, 210) of any one of the preceding claims, wherein the optimization operation comprises a filtering mechanism that compares predicted energy availability with required energy demand during expected water usage events, and adjusts system operation accordingly.

19. The controller (100, 210) of any one of the preceding claims, wherein the optimization operation comprises a scheduling strategy that anticipates temperature setpoints based on user-defined comfort profiles, in particular day and night comfort profiles.

20. The controller (100, 210) of any one of the preceding claims, wherein the optimization operation comprises delaying or advancing system activation based on external conditions, in particular external temperature forecasts, in particular room temperature and outdoor temperature, and the thermal inertia of the system to reduce energy consumption while maintaining acceptable comfort levels.

21 . The controller (100, 210) of any one of the preceding claims, wherein the optimization operation comprises a balanced consideration of thermal comfort and energy-related objectives, the balance being dynamically adjusted based on user preferences, system learning, or external conditions.

22. The controller (100, 210) of any one of the preceding claims, wherein the optimization operation comprises selecting or prioritizing between the heat pump (202) and one or more backup thermal sources, in particular heat sources, of the hot water system (200), based on a tradeoff between energy efficiency and time-to-comfort as a component of thermal comfort, wherein time-to-comfort refers to the predicted time required to deliver water at a predefined comfort temperature at the point of use.

23. The controller (100, 210) of any one of the preceding claims, wherein the thermal comfort satisfaction level and/or the energy-related consumption level is represented as a numerical score.

24. The controller (100, 210) of claim 23, wherein the score representing the thermal comfort satisfaction level corresponds to a similarity metric or deviation between the predicted thermal comfort satisfaction level and a predefined or expected satisfaction threshold.

25. The controller (100, 210) of any one of the preceding claims, wherein the controller is further configured to be operatively connected to a communication interface configured to exchange data with one or more components of the hot water system (200), the communication interface comprising hardware and/or software elements supporting wired or wireless communication protocols.

26. The controller (100, 210) of any one of the preceding claims, wherein the controller is further configured to generate control commands to adjust the operation of the hot water system, the control commands comprising instructions to activate or deactivate heat sources, switch between heating modes, modulate thermal energy output or energy consumption, or modify temperature setpoints based on optimization logic, user preferences, or learned behavior.

27. A hot water system (200) comprising a heat pump (202) and an optional backup electrical heat source (208) and/or an optional hydraulic backup heat source, the hot water system (200) comprising a controller (100, 210) configured as in any one of claims 1 to 26, the hot water system (200) being controlled by the controller (100, 210).

28. The hot water system (200) of claim 27, further characterized in that the hot water system (200) is configured for use in cooling applications.

29. The hot water system (200) of claim 27 or 28, wherein the hot water system (200) is further configured to operate in a cooling mode, wherein the system extracts thermal energy from the delivery point to deliver water at or below external temperature.

30. A method for controlling a hot water system (200), the hot water system comprising a heat pump (202) and an optional electrical backup thermal source, in particular heat source, (208) and/or an optional hydraulic backup heat source, the method comprising: a) storing, in a memory (110), instructions executable by a processor (102); b) accessing, by the processor (102), data and instructions from the memory (110); c) receiving, via a communication interface (140) communicatively coupled to the processor (102) and the hot water system (200), data from the hot water system (200); d) predicting, by a thermal comfort determination module (302), a thermal comfort satisfaction level delivered by the hot water system (200) based and/or trained on at least one of:

• a predicted water temperature at the point of use;

• a predicted availability of water during expected usage periods;

• historical user feedback or usage patterns;

• system response time to a water demand; e) predicting, by an energy optimization input module (304), an energy-related objective level of the hot water system (200) based and/or trained on at least one of: • a predicted activation duration and/or frequency of the heat pump (202);

• historical usage patterns and demand forecasts;

• external temperature, in particular room temperature or outdoor temperature, and system thermal losses;

• at least one energy tariff, CO₂ footprint, and/or at least one cost model associated with a heat source or thermal exchange system; f) performing, by the processor (102), an optimization operation (320) to determine a control strategy for the hot water system (200), the optimization being based on a weighted consideration of:

• a thermal comfort objective; and

• at least one energy-related objective, including cost, environmental impact, or system efficiency; wherein the optimization operation (320) determines whether an adjustment of the operation of the hot water system (200) is required and if so, defines such adjustment based at least in part on the predicted thermal comfort satisfaction level and the predicted energy-related objective level, such that the adjustments satisfy a weighted combination of the thermal comfort and energy-related objectives; g) determining, by the processor (102), commands to operate the hot water system (200) based and/or trained on a result of the optimization operation (320), the commands comprising at least one of scheduling, source selection, or control command.

31 . Data processing device comprising means for carrying out the method of claim 30.

32. A computer program product comprising instructions to cause the hot water system (200) of any one of claims 27 to 29 to execute the steps of the method of claim 30.

33. A computer readable data carrier having stored thereon the computer program product according to claim 32.

34. A data carrier signal carrying the computer program product according to claim 32.

35. Use of the controller (100, 210) according to any one of claims 1 to 26 for controlling a hot water system (200) according to claims 27 to 29, wherein the controller (100, 210) executes an optimization operation based on predicted thermal comfort satisfaction and energy-related consumption levels, and determines control commands for the hot water system based on the result of the optimization, the control commands comprising scheduling and source selection decisions that account for time-of-day preferences, availability of renewable or lower-cost energy sources, and system performance constraints.