HK1237396B - Variable air volume modeling for an hvac system - Google Patents

Description

Variable air volume modeling for HVAC systems

Cross-references to other applications

This application claims the benefit of U.S. Provisional Patent Application No. 62/120,218, filed February 24, 2015, which is incorporated herein by reference to the extent permitted by law.

Technical Field

The present embodiments generally relate to industrial process heating, ventilation, and air conditioning (HVAC) systems.

Background Art

To distribute air in an HVAC system, an air distribution system, including air handling units, moves air between the space to be ventilated and the equipment. The air handling units include fans that move air between zones, rooms, or other areas local to an occupied space.

Air handling is controlled by one or more controllers, such as those in a panel. Using a set of rules, the controller causes the air handling units to provide more or less flow. For example, feedback from a temperature sensor is used to increase or decrease fan speed to drive the temperature within a set point range. Due to poor design, wear, or other reasons, air handling units may not operate optimally or may not operate adequately. Rule-based control may not identify incorrect operation through error reporting. For more complex air distribution systems with multiple interconnected air handling units, rule-based control may not be able to handle interference between air handling units.

Summary of the Invention

Using information provided by the air handling unit's controller or controllers, the remote server uses a heuristic model to determine the air handling unit's settings. Rather than simply using rules for each air handling unit, a model-based solution determines the settings. The model is used to optimize air distribution operations. In additional or alternative embodiments, measurements are collected and used to derive analytics. The measurements may include data not otherwise used for rule-based control of the air handling units. Analytics are used to forecast demand, as input to modeling, to identify problems, and/or to identify opportunities.

In a first aspect, a control system for heating, ventilation, and air conditioning (HVAC) is provided. An air handling unit (AHU) includes a plurality of sensors, the plurality of sensors comprising a group consisting of temperature, relative humidity, fan speed, pressure, input power, and fan flow from a variable speed drive of a fan of the AHU. A network is connected to a controller of the AHU. A memory is configured to store measurements from the AHU sensors and a heuristic model of the AHU. A cloud server is remote from the AHU and connected to the network. The cloud server is configured to receive the measurements, identify operating parameters for the AHU, solve the heuristic model using the measurements, and output the operating parameters.

In a second aspect, a method for modeling heating, ventilation, and air conditioning (HVAC) is provided. A server optimizes a model of air handling in the HVAC system based on sensor measurements. The server determines settings for the air handling in the HVAC system based on the optimized model and transmits the settings to the HVAC system.

In a third aspect, a method for performing analysis in a heating, ventilation, and air conditioning (HVAC) system is provided. The operation of an air handling unit in an HVAC system is measured. The measurements include fan speed, pressure, power input, and flow rate. The measurements are sent from the measurement unit to a processor. The processor analyzes the operation of the air handling unit based on a combination of two or more of the fan speed, pressure, power input, or flow rate. Based on the results of the analysis, issues or opportunities with the air handling unit are presented on a display.

Other systems, methods, and/or features of the present embodiments will become apparent to one skilled in the art upon examination of the following drawings and detailed description. All such additional systems, methods, features, and advantages are intended to be included within this description, be within the scope of the invention, and be protected by the following claims. Additional features of the disclosed embodiments are described in, and will be apparent from, the following detailed description and drawings.

BRIEF DESCRIPTION OF THE DRAWINGS

The components in the drawings are not necessarily drawn to scale, emphasis instead being placed on clearly illustrating the principles of the present disclosure. In the drawings, like reference numerals designate corresponding components throughout the different views.

FIG1 is a block diagram of an embodiment of an air handling control system in an HVAC system;

FIG2 illustrates an exemplary control system with an air handling unit;

FIG3 shows a graph of information used for operational efficiency analysis;

Figure 4 is an example traffic diversity graph;

Figure 5 shows example trends in diversity over time;

FIG6 shows an example distribution of air flow;

FIG7 illustrates a diagram of multiple bins as a function of maximum flow percentage according to one example;

FIG8 is an example diagram of the maximum flow percentage varying with the box;

FIG9 is a diagram of one example of box interconnections in an air distribution system;

FIG10 is an example of a graph showing the number of times the actuator is repositioned as a function of the box;

FIG11 shows the determination of critical and maximum flow regions and pressures over time;

Figure 12 shows the integration of performance analysis for air handling, equipment, and controlled areas;

FIG13 is an embodiment of a method for HVAC control using a heuristic model; and

FIG. 14 illustrates one embodiment of a method for HVAC analysis.

DETAILED DESCRIPTION

Variable air volume control is described below, but can be applied to other HVAC processes. A model-based approach to variable air volume can be a cost-effective implementation. This can result in a high-value solution that improves operational performance.

In one aspect, statistical processing, physics-based modeling, typical optimization methods, or learning algorithms are provided for online optimization. HVAC is controlled based on optimization in the cloud from a remote server. A controller or multiple controllers of the HVAC system are connected to a remote server via a network, such as an intranet or the Internet. The server provides a cloud service for controlling the HVAC system. As the HVAC system operates, various measurements are provided to the server from sensors and/or controllers. These measurements indicate characteristics of the HVAC system. The server can apply analytics to observe operating points (e.g., actual operating points and set points) to establish the behavior of the HVAC system through modeling. Optimal set points, operations, or other controls for the HVAC system are determined and provided from the server to the HVAC system. Predictions of future behavior can be made based on current behavior and used to plan maintenance, establish controls immediately to change expected performance or avoid undesirable situations, or otherwise reset the operation of the HVAC system.

Analysis of operations is used as feedback for modeling. Alternatively or in addition, analysis is used to indicate trends, efficiencies, problems, or opportunities. Given various measurements, air handling operations can be analyzed to identify design changes, more optimal occupancy distribution, diagnostic information, maintenance information, or other information that can be used to monitor air distribution.

Figure 1 illustrates one embodiment of a control system for HVAC. The system uses a remote or cloud server with heuristic modeling to determine settings for air handling in the HVAC system. The heuristic modeling uses physics-based or machine learning models to provide settings based on current and/or current and past air handling operation. Alternatively or in addition, the control system performs analytics that can be used for maintenance, predictive operations, diagnostics, or other indications of problems or opportunities in air distribution.

The control system implements one or both of the methods in Figures 13 and 14. Other methods may be implemented.

The control system includes air handling units 12, a network 22, a cloud server 24, and a computer 30. Additional, different, or fewer components may be provided. For example, any number of air handling units, such as tens or hundreds, may be provided. As another example, the computer 30 may be implemented as part of the controller 14 of the air handling unit 12, rather than being implemented as a standalone device, or may not be provided. In another example, different HVAC air handling systems having corresponding air handling units 12 are connected to the cloud server 24 via the network 22.

FIG2 illustrates an embodiment of the control system of FIG1 , wherein the heuristic model is a physics-based model designed to minimize energy consumption within a floor or zone within building 10. Data from previous operations is collected to represent past behavior. In the example shown, controller 14 receives data from sensors, calculates data from other information, and/or uses the data to control or operate air handling unit 12. T represents temperature, RH represents relative humidity, V represents volume flow, P represents input power, Q represents heat, and m represents mass flow. The subscript oa represents outside air, the subscript ra represents return air, the subscript chws represents chilled water supply, the subscript hwr represents chilled water return, the subscript sa represents supply air, and the subscript fan represents fan 18.

In the example of FIG2 , controller 14 provides various operational data of air handling units 12 of building 10 to server 24. Metrics and/or loads of temperature, volume, return handling, mass flow, fan pressure, or other information are measured and applied through physics-based modeling of the HVAC system. Controller 14 or server 24 may collect or use additional, different, or fewer values.

The server 24 calculates or provides a heat calculated based on solar heat, internal heat, weather, and any other heat sources. The heat is calculated for the volume or area of the air handling unit 12. Indications of the operation of the air handling unit 12 other than heat can be used. Based on a physical model, the model is adapted to the specific air handling unit using heat and past settings or measurements. The solution may include a cost function, such as minimizing energy through calculus of model parameter changes (e.g., searching for a combination of input parameters that produces the calculated heat while minimizing energy consumption). Other costs may be used. Based on the model that solves for minimum energy, set points for the supply air temperature T _sa , the cooling water supply temperature T _chws , the cooling water return temperature T _hwr , the cooling water mass flow rate m _chws , and the return water mass flow rate m _hWs are determined and provided to the controller 14 for further use.

Returning to Fig. 1, air handling unit 12 is any air handling unit known now or developed later, for residential, industrial or office use.Air handling unit 12 comprises return air input, fresh air input, air mixing section, filter, cooling coil, heating coil, damper or actuator, attenuator, exhaust port and one or more fans 18.Other, different or fewer components can be provided.For example, air handling unit 12 is a box with damper and fan 18 without filter, mixing section, attenuator, heating coil and/or cooling coil.The heating coil and cooling coil are connected by pipes, for the supply and return of cooling and heated water.Heating and/or cooling without water can be provided.

The fan 18 used for air distribution is any fan used to force air into an area. The fan 18 includes blades and a motor. Any blades can be used. Any motor can be used. In one embodiment, the motor is a variable drive, such as a variable frequency drive (VFD). In response to a control signal, such as a frequency, duty cycle, amplitude, or other signal characteristic, the motor controls the speed of the fan 18. Variations in the speed of the fan 18 can result in greater or lesser airflow. Alternatively or in addition, an actuator controls a damper for increasing and/or decreasing airflow. Variations in airflow caused by the fan 18 can be used to more closely regulate the temperature downstream of the fan in the air distribution.

The air handling unit 12 includes one or more controllers 14. Controller 14 is a field panel, processor, computer, application specific integrated circuit, field programmable gate array, analog circuit, digital circuit, or other controller. A single controller 12 is shown, but different controllers may be arranged. For example, different controllers may be provided for different components (e.g., a controller for fan 18 may be different from a controller for a damper, heating coil, or cooling coil). Distributed controllers may communicate for interactive control, may be controlled by a master controller, and/or may operate independently of other controllers.

The memory 16 may be a random access memory (RAM), a read-only memory (ROM), a removable medium, flash memory, solid-state memory, or other memory. The memory 16 stores set points, sensor values, control information, and/or instructions for control by the controller 14. For example, the memory 16 may be a non-transitory computer-readable storage medium for storing instructions. When the physical controller 14 executes the instructions, the control discussed herein is performed.

In another example, the memory 16 stores data acquired from the sensors 20, set points, or other operational metrics of the air handling unit 12. The stored data is used to control the operation of the air handling unit 12, such as temperature measurements for comparison with temperature set points and fan speed settings for providing conditioned air to occupied spaces. The stored data may also include information not used in rule-based control of the air handling unit 12, such as measurements of pressure and/or input power.

Sensors 20 are used to measure temperature, relative humidity, fan speed, pressure, input power, and fan flow. Additional, different, or fewer types of sensors 20 may be used. Solid-state or other sensors may be used. For example, the input power sensor is a power meter for the variable speed drive fan 18 and/or the entire air handling unit 12. One or more types of sensors may be simulated. For example, the fan speed uses a control value (e.g., frequency) for controlling the fan 18 rather than a measured speed. An expected speed is used to account for the speed setting. As another example, fan flow is estimated based on fan speed. Pressure may be modeled based on fan speed and based on modeled or pre-measured impedance from the damper and downstream impedance sources.

One or more sensors of each type are provided. Temperature sensors may be provided for return air, fresh air, outdoor air, outlet air, fans, and/or other locations. Similarly, relative humidity sensors may be provided for return air and fresh air. In other embodiments, sensors located remote from the air handling unit 12 are used, such as those obtaining outdoor temperature and relative humidity from a weather station or source via network 22. Where sensor 20 is not used for rule-based control, sensor 20 may be retrofitted or added.

Based on sensors 20 or other sources, controller 14 collects or gathers data from or for air handling units 12 and/or air distribution systems. In the absence of modeling by server 24, the collected data is also used to operate air handling units 12 and/or may be collected for other purposes other than rule-based operation of air handling units 12. In one example, the measured data for a given air handling unit 12 is fan revolutions per minute (rpm) and fan flow, fan total pressure, and power input (in kilowatts) from the variable speed drive of the fan. Other metrics may be used.

The network 22 connects to the server 24, the computer 30, and the controller 14 of the air handling unit 12. The network 22 is a local area network, wide area network, enterprise network, cellular network, intranet, Internet, wireless, wired, or other network used for TCP/IP or other communications. The network 22 can be used for various purposes unrelated to HVAC or can be a network dedicated to HVAC. Although a single network 22 is shown, a combination of multiple networks can be used. The network 22 provides communication between the air handling unit 12, the server 24, and/or the computer 30.

Server 24 is a processor, computer, card, or other server for processing and communicating with air handling unit 12. Server 24 is remote from air handling unit 12, such as in a different building, city, county, state, or country. Alternatively, server 24 is a workstation in the same building, such as a workstation used for overall management of the entire HVAC system. Server 24 is not located in the same room or area as air handling unit 12. In other embodiments, a workstation is used as a controller instead of server 24.

The memory 26 of the server 24 is the same or a different type of memory as the controller's memory 16. In one embodiment, the memory 26 is a database memory or other memory that is part of the server 24 or accessible by the server.

The memory 26 is configured to store measurements from the sensors 20 and/or other data (e.g., set points and design specifications) of the air handling unit 12. The data is provided to the memory 26 via the network 22. Alternatively, the server 24 accesses the memory 16 without storing the data in a separate or server-side memory 26. The memory 26, 20 at the HVAC system or at the server 24 stores the measurements for analysis by the cloud server 24.

Data from multiple devices can be stored. For modeling, data is linked based on the HVAC system. Data from physically linked or related components of the HVAC system are marked with links. Alternatively, data is marked by source, and the memory 26 includes a schematic or other link structure for associating data from related components. Any relationship can be used in the linked data, such as a physical relationship between devices. For example, sensors 20, actuators, controllers 14, and/or rooms, areas or other building spaces for the same air handling unit are linked. As another example, variable air volume boxes (e.g., AHUs) are linked to building areas. In another example, devices used in a control loop or other control structure are linked together.

The memory 26 also stores one or more heuristic models 28. The heuristic models 28 represent the type of air handling unit, but are adapted to the specific air handling unit 12 based on data received from the air handling unit 12. The adaptation takes into account wear and tear, operation of the device within tolerances, installation effects, or other considerations that are unique to a given air handling unit 12, although of the same type. The heuristic models 28 are solved to adapt the representation of operation to actual operation. The heuristic models 28 can be used to determine the optimal or desired settings for operation, rather than using a rule-based (e.g., classical control) approach to control the HVAC system. The heuristic models 28 are used to determine set points, which are then used by the controller 14 in rule-based control.

Adaptation relies on data used in rule-based control, including measurements with or without data not used as part of rule-based control (e.g., other measurements, design specifications, and/or other information). For example, pressure is not used by controller 14 to control the air handling unit in any feedback loop without input from the model. Pressure is measured or calculated and used to adapt the model.

In this online optimization method, the server 24 performs heuristic operations to determine the values for rule-based control. In other embodiments, the local controller 14 or the controller of the HVAC system performs the heuristic operations.

An example heuristic model 28 is a model based on physics. Physics is used to represent the interrelationships between components. The operation or behavior of the air handling unit 12 or other air distribution is modeled using a set of equations, matrices, lookup tables of related variables, or a combination thereof. The air entering and exiting is represented as having volume or mass flow, temperature, pressure, and/or other characteristics. These variables are used to correlate the internal operation of the input air that produces the output air. Since the internal operation depends on the device (e.g., fan 18) specific to the air handling unit 12, the modeling can be adjusted based on the measurement of the operation from the controller 14 to consider the specific device. For example, the efficiency of the fan is included in the model. For that specific fan 18, the efficiency of producing the output air is determined based on the input air measured for a given air handling unit 12. Modeling is performed for the characteristics of the fan 18 for solving the effect of any damper on the fan 18, so the damper characteristic is another variable or multiple variables that are adapted to the air handling unit 12.

In one embodiment, the physics-based model is a demand flow model. For example, the air handling unit 12 is modeled as described in U.S. Published Patent Application No. 2011/0301766 for controlling chilled water in a chilled water system or hot water distribution system. Demand flow considerations are used to model the interrelationships of various components for controlling the air handling unit 12. Demand flow reduces or eliminates low delta temperatures (T) and improves efficiency. Demand flow utilizes variable flow to address low Delta T and significantly improves efficiency. When Delta T is at or near a component's design Delta T, variable flow under demand flow maintains that component's Delta T. Therefore, demand flow significantly improves operating efficiency, thereby saving energy costs. The increased efficiency provided by demand flow can also reduce pollution. Furthermore, unlike conventional variable or other pumping technologies, demand flow can also increase component life expectancy by operating these components at or near their specified inlet and outlet temperatures, or design Delta T.

Whether it's cooling or heating demand or loads, demand flow provides higher efficiency by operating components in a synchronized manner. In one or more embodiments, this is achieved by controlling pumps and/or fans to maintain Delta T at specific components or points. Generally, demand flow operates on individual components to maintain Delta T at specific components or points. Control of the individual pumps or motors (and flow rates) in this manner creates synchronized operation. This synchronized operation balances flow rates, significantly reducing or eliminating low Delta T and the associated inefficiencies.

This same demand flow model can be used as the physics-based model.Other physics-based models can be used.

In order to adapt a general physics-based model to a specific air handling unit, the determination of the behavior and/or settings for the controls is treated as an optimization problem. Operation is optimized using neighborhood search, local minimization, other processes, or a combination thereof. Measurements are used. Other data, such as design specifications and analysis, can be used. Input measurements or data are used as boundary conditions in the solution. By changing one or more variables of the components of the air handling unit 12, the model is changed to provide the measured output. A combination of characteristics modeled in the air handling unit is determined that provides an output based on the time-varying inputs.

Once adapted, the physics-based model can be used to determine the optimal settings for controlling the air handling unit. The calculus of variance is used. Inputs are varied to find differences in output using the adapted model. The adapted model is then used to determine the combination of inputs that produces the output that best meets a criterion or criteria, such as minimizing energy consumption.

In an alternative embodiment, heuristic model 28 is a machine learning classifier. Using data from the same air handling unit and/or a large number of air handling units of the same type, the classifier can be trained using machine learning to output optimized or desired settings for given input features. The training data is annotated to provide substantially realistic or actual outputs for various examples. Machine learning learns from input features to predict outputs. Any input features can be used, such as measurements from the air handling unit. Design specifications or other data can also be used, or alternatively. The trained classifier simulates the behavior of the air handling unit 12.

A classifier is trained to predict outcomes, such as output air properties, given measured and other inputs. The classifier can predict a range of outputs for a given range of input setpoints. By iteratively testing different settings, the machine learning classifier can be used to find the one that produces the desired outcome, such as minimizing energy consumption.

Machine learning uses classification and merging, probability distributions, neural networks, support vector machines, or other processes to learn to predict the expected output given input features. Any machine learning method can be used, such as a genetic algorithm that applies any rejection/hold criteria and/or neural network. In one embodiment, machine learning is implemented using a genetic algorithm combined with a neural network. Other machine learning methods can be used, such as probabilistic boosted trees or Bayesian networks.

In an alternative embodiment, training data is collected from a representative HVAC system with known optimal settings. Data is collected for the system using different settings. Machine learning is applied to train a classifier that receives input measurements for a given HVAC system and outputs the optimal settings. The classifier outputs optimized control settings for the given HVAC system based on the machine-learned classifier. Other machine learning methods may be used.

In one embodiment, a machine learning classifier uses online or ongoing learning. The classifier is trained entirely or at least in part on the operation of a specific air handling unit. As the behavior of the HVAC system is determined, that behavior is used using online or ongoing machine learning. As changes in operation are collected as a feedback mechanism, further machine learning specific to the HVAC system is performed. As additional data is collected based on different settings and results, the feedback obtained is used to learn additional probabilities and/or distributions in an ongoing manner. The machine learning continues to learn the behavior of the given HVAC system. The learned behavior is used to determine settings for HVAC controls that can be optimized. The machine learning provides probability distributions or other statistics that can be used to control the operation of the HVAC system. The achievable operations are learned to optimally control the HVAC system.

Cloud server 24 is configured to receive measurement and/or other data for, about, or from air handling unit 12. The data is received in response to a query by loading from memory 26 or storage 16, by pushing from controller 14, or other processing.

For modeling, the cloud server 24 can be configured to derive one or more characteristics from the measurements. Any analysis discussed herein can be derived. For example, the diversity of fan operation can be calculated. Other characteristics can be mass flow based on fan speed and pressure measurements. The model itself can allow fan speed and pressure to be input without deriving mass flow, or derive mass flow and use it as an input feature in the model. The server 24 collects measurements, calculates derived characteristics, and/or looks up characteristics or specifications. Information is collected as input for adapting the model 28 and/or as input for modeling the output of a given input using the adapted model 28.

The server 24 is configured to identify operating parameters of the air handling unit 12 by solving a heuristic model using the measurements and other data. An operating parameter is a set point or other setting that can be controlled by the controller 14. An operating parameter is a variable of the air handling unit 12, and the value of the variable can be used to change or affect the operation of the air handling unit 12.

The server 24 loads the model 28 from the memory 26 and solves for a given or specific air handling unit 12. The solution adapts the heuristic model 28 to the air handling unit 12. Measured, derived characteristics, and/or other data are used to represent boundary conditions or inputs for the model 28. The server 24 uses the adapted model 28 to determine adjustments or values for the variables in the model 28 to optimally control the air handling unit 12.

For a machine learning classifier, input features are obtained. By applying the input features to the classifier, the classifier solves for a specific air handling unit 12. The resulting output represents the expected output given the input features. By iteratively adjusting inputs such as those associated with fan operation, a probability distribution or other settings is output. The classifier can be trained to output the optimal setting to use (e.g., to reduce energy consumption), or the output probability distribution can be used to select a setting regardless of cost given a variance in the inputs.

For physics-based models, server 24 solves through iterative optimization. Optimization methods, such as neighborhood search and/or local minimization, are applied to the model. In this iterative approach, energy consumption can be minimized using model optimization. Alternatively, variance calculus can be used as part of minimizing energy consumption using adaptive model 28. The resulting control values are determined and applied. Depending on the application, additional metrics are provided to further optimize using the physics-based model.

The server 24 is configured to output operating parameters or control parameters such as set points to be used by the controller 14 . These parameters are provided on the network 22 .

In one example, the air handling unit 12 consumes energy. Various operational metrics are collected from the HVAC system associated with the air handling unit 12. Other information can be derived from the measurements. This collected information, specific to a given HVAC system, is used by the server 24 to provide a set of control parameters to be used in the operation of the HVAC system. The cloud-based server 24 uses a heuristic approach rather than a strict rule-based optimization to determine the control parameters of the air handling unit 12. Machine learning and/or an iterative adaptation process are applied to determine the values of the control parameters of the air handling unit 12. Load, temperature, cost, mass flow, and/or other information can be used in the analysis. The set point provided to the controller 14 can be any set point, such as damper position or fan speed. By using a model of the HVAC system, the optimal fan operation for a given cost function is determined.

In addition to or as an alternative to using heuristic models for optimal control, the server 24 or computer 30 analyzes the operation of the air handling units 12. The analysis can be performed for individual air handling units 12 or for an entire HVAC system (e.g., dozens of air handling units on a floor or throughout a building). Data analysis is provided for monitoring ongoing operations, predicting maintenance or future problems, identifying design flaws, determining deviations from design or specifications, or identifying opportunities (e.g., by reallocating areas for more efficient operation).

The computer 30 is a workstation, personal computer, tablet computer, smartphone, or other processing device. The computer 30 receives output from the air handling unit 12 and/or the cloud server 24. For example, the computer 30 is an HVAC workstation that manages or monitors the operation of the HVAC system including the air handling unit 12. As another example, the computer 30 is a personal computer or a server.

Computer 30 includes a display 32 for displaying analysis information, such as graphs, values, or recommended remedies. Display 30 may additionally or alternatively display measurements, data, settings, or model information.

Figures 3 through 11 illustrate example analyses. Data from or about air handling unit 12 is used to analyze the operation of air handling unit 12. Other examples may be provided. By processing data used for control and/or data acquired but not used for control, issues or opportunities with the HVAC system and/or air handling unit 12 are identified.

FIG3 illustrates an example of total fan pressure as a function of fan flow. Three curves are shown representing pressure as a function of flow at different fan speeds (revolutions per minute (rpm)). There is a fluctuation region where two different fan flows (cfm) produce the same total fan pressure. Fluctuation regions can create operational confusion where a rule-based system may operate in an undesirable manner. The two solid lines illustrate the operating range with maximum efficiency. The dashed line represents the typical operating range for a given fan 18. Some operations fall outside the most efficient operating range. Line 36 represents the curve for the demand flow system. The demand flow curve provides ideal operation for the fan 18 based on modeling. The example of FIG3 is for a given fan 18. Other fans 18 may have different efficiency ranges, fluctuation regions, or other operations.

For analysis, the performance of fan 18 can be determined. For example, a demand flow curve 36 for fan 18 can be determined. Efficiency can be calculated as the ratio of power delivered by the fan to power input to the fan. If fan 18 efficiency decreases, fan 18 maintenance can be scheduled. Forward detection and diagnostics can be performed to improve efficiency.

If the fan 18 is operated at maximum capacity for a sufficient number of times, the fan 18 can be scheduled for replacement with a larger fan 18. The capacity of the fan can be predicted. The operation of the fan 18 can be tracked over time to identify trends in the capacity or amount of use, such as through regression analysis. If the fan 18 is expected to reach capacity within a given time, maintenance or replacement can be scheduled before that time.

Energy consumption can be determined so that savings can be measured and verified. Energy can be calculated by multiplying the product of flow and pressure by the efficiency.

The analysis relies on fan flow, fan total pressure, power input (eg, power input in kW from a variable speed drive), and fan speed. Additional, different, or fewer measurements and corresponding sensors may be used.

FIG4 and FIG5 illustrate another example of an analysis determined by computer 30 or server 24. The analysis is diversity, which may indicate a mismatch between fan load and area or occupancy. Diversity is a measure of operational variability. FIG4 illustrates energy savings, or the percentage of maximum potential savings, as a function of average fan flow diversity. For fan flow diversity, the fan flow setpoint and maximum design flow of fan 18 are measured or acquired. Additional, different, or fewer operating parameters may be used.

In the example of FIG4 , there is diversity in fan flow, but the diversity of the other 6 operating parameters (e.g., speed, capacity, efficiency, or other) can be determined. This example is the diversity of a given fan 18, but the diversity of a region within any time period can be used in other embodiments. A region or fan 18 with low diversity can indicate that the fan 18 or multiple fans are too large. To save costs, a smaller fan 18 can be used, or the region can be reallocated to more efficiently distribute fan power. In another example, a region with low diversity can indicate the area that is most suitable for placing a new employee or relocating an employee. Placing another employee in a region or area with high fan diversity will be more likely to drive the fan 18 to operate under load.

In the example of FIG5 , a trend or change in zone diversity over time is determined as a percentage of the design maximum fan flow rate. The operation of the fans (e.g., single versus dual) can be controlled to achieve optimal or learned diversity. Zone diversity trends can be analyzed for space utilization and used for planning. For example, an increase in diversity may indicate a need for larger fans. For another example, a decrease in diversity may indicate that one fan of a dual-fan air handling unit can be powered off or not used.

FIG6 illustrates another example of analysis. The cloud server 24 or computer 30 determines the sensitivity of the air handling unit 12 to outdoor conditions based on the relationship between the outdoor conditions and the operation of the air handling unit 12. For example, the fan flow set point, the fan maximum design flow, the outdoor temperature, the outdoor wet bulb/relative humidity, and/or other data may be measured. Additional, different, or fewer data types may be measured.

The fan flow or other operating parameters are associated with the outdoor temperature or other outdoor data. The analysis shows the relationship between the outdoor temperature and the fan flow or the extent to which the fan flow is affected by the outdoor temperature. The analysis uses the correlation, probability distribution or other analysis of the measured data to perform load analysis (e.g., ventilation versus outdoor driving load) and/or reduce ventilation based on occupancy. Thus, areas or fans 18 that are particularly affected by outdoor conditions can be identified and solutions can be determined. For example, fixing window glass, adding or using sunshades, or adding insulation. By isolating susceptible fans 18, solutions can be located (e.g., by fan or area), thereby reducing costs. Analyzing the correlation over time can indicate an appropriate solution, such as a high correlation as a driving load during the morning when sunlight is present. Occupancy can be determined based on the analysis. The load or fan flow can be reduced by changing the occupancy, thereby avoiding high loads in areas susceptible to outdoor conditions.

Figure 6 shows load or fan flow (cfm) as a function of a probability distribution associated with temperature. The probability distribution is the probability distribution of temperature over time. For example, 5% of the time, the temperature is within the range of 90-95 degrees Fahrenheit. Figure 6 relates fan flow to the temperature distribution. Figure 6 shows that there is a 90% chance (the range between the vertical lines) that the hourly average flow rate will be between 20,200 and 22,000 cfm. Some example statistics that can be used include the minimum (e.g., 18,100 cfm), maximum (e.g., 22,700 cfm), mean (e.g., 21,000 cfm), and standard deviation (e.g., 570 cfm) of the temperature and average airflow rate for each fan.

Figures 7-10 illustrate example analyses including dynamic flow or dynamic pressure performance. Figure 7 shows a graph of the number of boxes (i.e., fans 18 or air handling units) (y-axis) as a function of the maximum flow percentage for a given time. Figure 8 shows the change in maximum flow percentage with the box. Figures 7 and 8 are shown over a given time. Changes over time can be determined. Figure 9 shows the physical relationship of the air distribution boxes. In this example, 55 boxes are provided. Each point is a connection between a given box and an air distribution node or area. For example, box 34 is connected to nodes or boxes 1, 9, and 33, such as receiving air from those nodes. The flow or pressure of a box is based at least in part on the flow or pressure in the associated or connected box. A given box may require a specific pressure, so the box connected upstream provides that pressure. Figure 10 shows the number of times the actuator is repositioned over a given time period as a function of the number of boxes.

Tank flow setpoints can be used to analyze dynamic flow distribution and/or dynamic pressure requirements per tank based on the tank flow setpoints. This allows identification of critical and/or degraded areas. Degraded areas may include tanks with the highest pressure requirements. By identifying and determining the cause of the highest pressure requirements, energy can be saved. Critical areas may be areas with higher pressure requirements. By identifying critical areas, solutions can be determined and energy can be saved. Differences in the flow of each tank or multiple tanks, which are closer or farther from the maximum flow, can be identified and used to modify air distribution to prevent wear and tear on the tanks.

In one embodiment, the tank flow setpoint and tank actuator repositioning are measured. Degraded or critical regions are determined based on the number of actuator repositions, tank pressure, and tank flow. Any function combining these variables can be used. The distribution of tank pressure requirements, tank flow requirements, and actuator repositioning can be analyzed to predict service requirements for proactive service and preventative failures to reduce downtime, avoid comfort complaints, and/or extend equipment life (maximum pressure operation can shorten life).

FIG11 shows another example of analyzing pressure, critical zones, and maximum flow areas over time. The same bin relationships as shown in FIG9 are used. For each sampling time (e.g., every 2-3 hours), the zone with the critical pressure is identified (e.g., bin 15 at 8:00 AM), the zone or zones with the maximum flow are identified (e.g., bins 17, 56, and 26 at 12:00 AM), and the pressure of the critical zone is identified (e.g., 1.50 at 8:00 AM). The pressure range or other information can be identified for each bin, or any critical or degraded bin. In a combined zone operation, the cloud server 24 or computer 30 determines the relationship or combined zones based on an HVAC system configuration such as that shown in FIG9. Determine the dynamic pressure, flow, or other requirements over time.

Pressure, and/or dynamic pressure in critical zones, and/or flow in critical zones can be used to predict areas of maximum flow. Regression analysis can be used to establish trends over any time period. Comfort analysis can be predicted. Lower pressures can indicate comfort, while higher pressures can indicate airflow that does not meet occupant needs. Troubleshooting of design flaws or other problems can be provided. Box sizes and maximum flow values can be estimated to identify areas requiring larger boxes or fans 18. The maximum flow of each box can be compared to the design specifications of the HVAC system or to the box design to identify boxes operating near their maximum values.

In another example of dynamic flow and/or pressure analysis, zones or boxes are identified that are associated with temperature, flow, or both temperature and flow. Dynamic flow/pressure and comfort analysis can be performed on the intercepted zones. Air distribution nodes corresponding to air distribution variable air volume boxes associated with the zones are found. Box flow setpoints, zone temperatures, temperature setpoints, actual flow, and actuator reposition counts are measured. Pressure and flow combined with temperature are analyzed, such as recording the box at maximum flow and looking up the temperature measured for that zone to determine if the box is adequately controlling temperature. Each zone can be analyzed for energy balance application (i.e., flow sufficient to meet temperature setpoints if performance trends indicate a correct balance) and/or flow and comfort. This analysis provides value for dynamic first-principles positive detection of zones (i.e., zone failures due to not meeting temperature setpoints or exhibiting some error) and diagnostics, detection of dynamic trends (e.g., morning versus afternoon trends), and/or provision of predictive services to improve comfort and energy performance.

Figure 12 illustrates the integration of supply and demand-side information in performance analysis. The analysis performed by the cloud server 24 is associated with the supply side of energy distribution (such as the cooling and heating equipment 15), on the one hand, and with the demand side (such as the area or space side of the building 10), on the other hand. By using communication to and/or from any of the air handling units 12, the areas of the building 10, and/or the equipment 15, the air handling model can be expanded to include analysis, thus serving as the center of global execution.

In one embodiment, outdoor sensors, zone sensors or space sensors for a portion of a building supplied with air by an air handling unit 12, sensors for the air handling unit, and/or equipment sensors provide information used in modeling and/or analysis. For example, for outdoor and zone information, the temperature, relative humidity, and volume of the outdoor air drawn into or supplied by the air handling unit are measured. For the equipment, the supply and return cooling water and hot water temperatures, as well as the supply and return mass flow rates, are measured. For the air handling unit 12, fan power, the temperature of the supplied air, and the relative humidity of the supplied air are measured. Additional, different, or fewer measurements and corresponding sensors may be provided.

The model can be expanded to include different sources of information. Physical quantities of equipment and/or areas can be added. The values of these variables can be included as features used in machine learning and subsequent applications using the machine learning model. The heuristic model can then be used for analysis and/or control. The information provided can be used for any analysis, such as those discussed herein or others. By integrating the values of additional variables external to the air handling unit 12, additional analyses can be provided to assist with maintenance, design, planning, and/or operation.

In one embodiment, the analysis provided by the cloud server 24 includes analysis of regional and/or spatial performance variables to trigger an investigation of equipment performance variables, and vice versa. In another embodiment, the analysis provided by the cloud server 24 analyzes regional and/or spatial performance variables to trigger an investigation of air handling performance variables, and vice versa. The investigation may be an analysis using a model (such as a heuristic model) or other analysis. The analysis of the relevant systems or variables may result in different actions (such as model-based, automatic changes), may result in the output of information used by designers or others (such as a maintenance plan), and/or may generate warnings or instructions for possible improvements.

FIG13 illustrates one embodiment of a method for HVAC modeling. The method may be implemented by the system of FIG1 , the system of FIG2 , the controller 14 , the server 24 , the computer 30 , or other devices. For example, the server performs all actions. Execution relies on communication with the air handling unit 12 and/or the controller 14 .

The method is performed in the order shown or in another order. For example, action 58 is performed before action 56. Additional, different or fewer actions may be provided. For example, action 58 is not performed.

In action 52, the server 24 optimizes the model 28 of the air handling in the HVAC system based on the measurements from the sensors 20. The sensors 20 and other data sources provide information used by the server 24. The server 24 adapts the heuristic model 28 to the data, such as adapting a physics-based model. A neighborhood search, local minimization, or other iterative method is used to adapt the model to the data and/or to determine the optimal output based on the adapted model. Alternatively, the server 24 models the air handling using a machine learning classifier. Given a set of inputs, the data is input to the classifier to provide a learned result. The machine learning classifier performs classification based on the measurements and other input data.

In action 54, the server 24 determines the settings of the air treatment in the HVAC system. The model 28 is optimized to match the operation of the air treatment. The model 28 is used to determine the settings of one or more operating parameters of the air treatment, such as determining a set point for fan speed. The fan speed that is likely to provide the desired comfort level while minimizing cost can be determined. For example, the model is used to determine the difference in output results as a function of the difference in input or multiple inputs. The settings of one or more controlled settings are determined by applying a cost function, such as providing the desired comfort level with the minimum energy input. In one embodiment, all settings are determined together using the model 28 to minimize the cost function.

In act 56, the settings are transmitted to the HVAC system, such as by transmitting the settings to the controller 14 via the network 22. Any transmission format may be used. The transmission provides the settings that the controller 14 uses to operate the HVAC system.

In action 58, server 24 calculates an analysis that indicates problems or opportunities with the HVAC system's air handling. Any of the analyses discussed herein or other analyses may be calculated. The analysis provides information that indicates problems or opportunities. For example, the analysis may indicate trends that may lead to air handling failure or inadequate operation. As another example, the analysis may indicate an area or areas that can be used by occupants without reconfiguring the HVAC system.

The analysis is output on display 32. The user may use the analysis to make decisions such as maintenance planning, cost savings verification, occupancy scheduling, replacement, reallocation, reconfiguration, or other remediation.

FIG14 illustrates one embodiment of a method for analyzing HVAC. The method may be implemented by the system of FIG1 , the system of FIG2 , the controller 14 , the server 24 , the computer 30 , or other devices. For example, the controller 14 performs action 60 using the sensor 20 and then performs action 62 . The server performs actions 64 , 66 , and 68 . Execution relies on communication with the air handling unit 12 and/or the controller 14 .

The method is performed in the order shown or in another order. For example, action 68 is performed before action 64 and/or 66. Additional, different, or fewer actions may be provided. For example, action 68 may not be performed.

In act 60 , sensors 20 measure the operation of the air handling system in the HVAC system. For example, fan speed, pressure, power input, and/or flow rate may be measured. Other operating parameters may be measured. Controller 14 may collect or store additional information such as set points or design specifications (e.g., maximum mass flow).

In act 62, the measurement and/or other data is transmitted to a processor, such as a processor in server 24 or computer 30. Transmission refers to the transmission of any data at one time or over a period of time. Subsequent transmissions may only transmit data that has changed. Transmission may be wired or wireless, either directly or over a network. In one embodiment, the transmission is performed by server 24 accessing or searching the data in memory.

In action 64, a processor (e.g., server 24 or computer 30) analyzes the operation of the air handling system. Measurements and/or other data, such as fan speed, pressure, power input, and/or flow rate, are analyzed. The analysis provides remedial recommendations, charts, graphs, data, or other information that the operator can use for various purposes. This information may indicate problems or opportunities. This information may be used to highlight problems or opportunities, or information may be provided and the user may rely on the information to identify problems or opportunities.

To perform the analysis, one or more statistics are calculated. Alternatively or in addition, a formula or formulas are used to calculate the analytical information. Regression or other analysis can be used to identify trends.

In action 66, information including any issues and/or opportunities is presented on display 32. The results of the analysis are presented to the user. Specific issues or opportunities are identified, such as indications of tanks to be replaced, areas of increased occupancy, confirmation of cost savings, correlations with outdoor conditions, trends indicating a need for maintenance or adjustments, or other information. Alternatively or in addition, charts, graphs, or data are displayed for explanation to the user.

In act 68, the operation of the air handling unit 12 is modeled using the heuristic model 28. The server 24 uses the model 28 and the measurements and other data to model the operation. The model 28 is used to determine settings for controlling the air handling unit 12. These settings are transmitted to the air handling unit 12 or the controller 14 for implementation.

Although various embodiments of the present invention have been described, it will be apparent to those skilled in the art that more embodiments and implementations are possible, and these embodiments and implementations are within the scope of the present invention. In addition, the various features, elements, and embodiments described herein may be claimed or combined in any combination or arrangement.

Claims (18)

1. A control system for heating, ventilation, and air conditioning (HVAC), the control system comprising: An air handling unit (12) has a plurality of sensors (20) which are a group consisting of a sensor for temperature, a sensor for relative humidity, a sensor for fan speed, a sensor for fan pressure, a sensor for the input power of the variable speed drive of the fan of the air handling unit, and a sensor for fan flow. A network (22) connected to the controller (14) of the air handling unit; Memory (26), configured to store measurements from the sensors of the air handling unit and a heuristic model of the air handling unit; and A cloud server (24) located away from the air handling unit and connected to the network is configured to receive the measurements in order to identify operating parameters for the air handling unit by solving the heuristic model (28) using the measurements, and to output the operating parameters. The cloud server is configured to determine diversity based on the maximum flow rate of the fan design and the fan flow rate setpoint of the fan (18), and wherein the cloud server is configured to indicate a mismatch between the fan and the space based on the diversity. 2. The control system of claim 1, wherein the heuristic model includes a machine learning classifier, and wherein the processor is configured to solve by inputting the measurement into the machine learning classifier. 3. The control system according to claim 2, wherein the machine learning classifier includes a continuously running machine learning classifier. 4. The control system of claim 1, wherein the heuristic model comprises a physics-based model, and wherein the processor is configured to solve the model through iterative optimization of the physics-based model. 5. The control system of claim 1, wherein the cloud server is configured to derive characteristics from measurements and is configured to use the characteristics to solve the heuristic model. 6. The control system according to claim 1, wherein the operating parameters include control parameters for operating the air handling unit, and the cloud server is configured to provide the control parameters to the controller. 7. The control system of claim 1, wherein the sensor measures the fan flow rate, the pressure, the fan speed and the input power, and wherein the operating parameters include the fan efficiency, demand curve, energy consumption, load or capacity trend. 8. The control system of claim 7, wherein the cloud server is configured to indicate a problem with the fan according to the operating parameters. 9. The control system of claim 1, wherein the cloud server is configured to determine the operating parameters as a relationship between outdoor conditions and the operation of the air handling unit, and wherein the cloud server is configured to indicate the sensitivity of the air handling unit to the outdoor conditions based on the relationship. 10. The control system of claim 1, wherein the air handling unit is one of a plurality of air handling units, wherein the operating parameters include the number of actuator repositionings, air handling unit pressure, and air handling unit flow rate, and wherein the cloud server is configured to identify, based on the number of actuator repositionings, air handling unit pressure, and air handling unit flow rate, a degraded region of the plurality of air handling units requiring maximum pressure or a critical region with critical pressure. 11. The control system of claim 1, wherein the cloud server is configured to determine an air handling unit in combination with temperature, flow rate, or temperature and flow rate. 12. The control system of claim 1, wherein the cloud server is configured to analyze the values of variables of a region where air is supplied by the air handling unit, and to trigger an analysis of equipment performance or air handling performance based on the analysis of the values of the variables of the region. 13. The control system of claim 1, wherein the cloud server is configured to analyze the values of variables relating to device performance, and to trigger analysis of device performance or air handling performance based on the analysis of the values of the variables relating to device performance. 14. A method for heating, ventilation, and air conditioning (HVAC), the method comprising: Measurements from sensors (20) of an air handling unit (12) are stored in a memory (26), the air handling unit having a plurality of sensors (20) consisting of a sensor for temperature, a sensor for relative humidity, a sensor for fan speed, a sensor for pressure, a sensor for the input power of the variable speed drive of the fan of the air handling unit, and a sensor for fan flow. By operating a cloud server (24) located remotely from the air handling unit and connected to a network (22), wherein the network is connected to the controller (14) of the air handling unit: Receive the measurement; The operating parameters for the air handling unit are identified by using the measurement-solving heuristic model (28); and The operating parameters are output to control the air handling unit. The cloud server is configured to determine diversity based on the maximum flow rate of the fan design and the fan flow rate setpoint of the fan (18), and wherein the cloud server is configured to indicate a mismatch between the fan and the space based on the diversity. 15. The method of claim 14, wherein the heuristic model comprises a physics-based model, and wherein the identification comprises iterative solving using nearest neighbor search or local minimization of a physics-based model. 16. The method of claim 14, wherein the heuristic model comprises a machine learning classifier, and wherein the identification comprises classifying based on the measurement using the model. 17. The method of claim 14, further comprising: The cloud server performs analysis based on a combination of two or more of the fan speed, pressure, power input, or flow rate. Based on the results of the analysis, the problems of the air handling unit are presented on the display. 18. A non-transitory computer-readable medium encoded with processor-executable instructions, wherein when executed by at least one processor (24), the at least one processor causes the processor to perform a method for heating, ventilation and air conditioning (HVAC) according to any one of claims 14-17.