KR102599363B1 - Ai energy reduction and demand prediction system for user based - Google Patents

Abstract

The present invention relates to a user-based AI energy saving and demand forecasting system. More specifically, it allows users to determine and use the energy they use in their residences, and collects such data to determine the overall demand for heat energy, electric energy, etc. It is about a user-based AI energy saving and demand forecasting system that enables prediction.
According to the user-based AI energy saving and demand forecasting system formed as a preferred embodiment of the present invention, the target value selected for saving energy to be used in the future can be used as predictive data without using the energy pattern used by the user. If the user chooses to save energy, not only can he receive separate benefits, but he can also save money by reducing the amount used, and even if the user selects the eco mode to save energy, the controller's artificial intelligence Due to the control device, users can be provided with the most comfortable environment, and it is possible to predict energy demand from users by directly determining energy consumption, rather than predicting energy demand using conventional energy usage, thereby increasing the national energy level. Effects such as bringing about new changes in policy occur.

Description

User-based AI energy reduction and demand prediction system {AI ENERGY REDUCTION AND DEMAND PREDICTION SYSTEM FOR USER BASED}

The present invention relates to a user-based AI energy saving and demand forecasting system. More specifically, the user determines the amount of energy he or she wants to use, collects this amount of energy, and provides information on the overall energy use such as heat energy and electric energy. It is about a user-based AI energy saving and demand forecasting system that enables demand forecasting.

The central government or local governments must predict the overall amount of energy used by the people and prepare in advance to ensure a smooth supply of energy.

In other words, in the case of thermal energy, a somewhat predictable level of thermal energy must be prepared for district heating, and in the case of electrical energy, the amount of power generation must be adjusted so that there is no power outage even if the use of air conditioners increases rapidly and electric energy is suddenly consumed.

In modern society, as urbanization progresses, the concentration of population increases. As population density increases, more complex buildings are built, and these complex buildings use a large amount of energy.

These residential buildings use energy such as heat energy, electric energy, and water energy, and this energy must be provided for a comfortable residential environment.

In terms of energy use in the complex building, there are unit-specific facilities installed for each unit household and common facilities installed in common.

A variety of technologies and control methods are provided to manage the energy of conventional multi-unit buildings. In the Republic of Korea Patent Office Publication No. 2014-0141923, 'Design method for optimizing building energy system under complex conditions' is disclosed. Published Patent Publication No. 2018-0138463 provides 'a cloud server and method for optimally controlling the cooling system of a target building based on AI, and a cooling system control device.'

In the 'Energy Management System, Energy Management Method, and Energy Demand Prediction Method' (see Figure 1) of the above-described conventional published patent publication, a community is formed, and building energy usage data and the community are collected, including a pricing policy for the demand response capacity of the community. It is a configuration that includes input data for the acceptance response incentive policy.

Figure 2 is composed of a unit household device, environmental management system, meter, and an integrated system (AI server) that collects, judges, and controls data related to the units applied to conventional complex buildings, and the integrated system includes artificial intelligence based on big data. Intelligence (AI) is formed to enable deep learning and control.

Korea Intellectual Property Office Registered Patent Publication No. 1908515 provides an energy demand prediction method for buildings under the title ‘Energy management system, energy management method, and energy demand prediction method.’

(Patent Document 1) KR10-2014-0141923 A

(Patent Document 2) KR10-2018-0138463 A

(Patent Document 3) KR10-2016-0136965 A

(Patent Document 4) KR10-1908515 B1

However, conventional energy management systems and energy demand forecasting methods had the following problems.

(1) In the past, the method of predicting demand for energy usage was to collect and statistics the energy usage used by users to predict energy usage, making it difficult to accurately or reliably predict energy demand.

(2) Even if users individually save energy, there is no special benefit, and it is impossible to plan exactly how to save energy.

(3) If energy use is unconditionally reduced in order to save energy, users will not be able to live in a comfortable environment.

In order to solve the above problems, the present invention provides each unit household device including a total heat exchanger, air purifier, and kitchen hood, a hot water distributor, a differential pressure flow control valve, a water leak detection sensor, and a PTS sensor, and hot water, domestic hot water, gas, It is connected to a meter that measures the amount of heating used, and a temperature sensor, humidity sensor, fine dust concentration sensor, and carbon dioxide concentration measurement sensor are formed inside. A PID control device is formed, and the data can be deep-learned using an artificial intelligence neural network. A control device is formed, a controller capable of connecting to a network is installed in each household, and each controller, user terminal, and main server are connected to the Internet,

A registration step in which the main server registers each controller;

A tracking step in which the main server determines whether each controller is using one of the automatic control mode, manual control mode, or eco control mode and tracks each controller in which the eco control mode is set;

a classification step of classifying energy-using devices attached to each controller tracked by the main server by energy usage;

a target value setting step in which the main server receives a target energy saving amount (target value) from each controller after the classification step;

a notification step of notifying the user terminal of benefits that can be provided when the main server saves energy based on the target value;

It is characterized in that it consists of a prediction step in which the main server predicts the corresponding energy demand that can be used in each controller.

According to the user-based AI energy saving and demand forecasting system formed as a preferred embodiment of the present invention, the following effects occur.

(1) The target value selected to reduce future energy use while utilizing the energy pattern used by the user can be used as predictive data.

(2) When users choose to save energy, not only can they receive separate benefits, but they can also reduce the amount used, thereby ensuring energy savings.

(3) Even if the user selects eco mode to save energy, the user can be provided with the most comfortable environment due to the controller's artificial intelligence control device.

(4) Instead of predicting energy demand using conventional energy usage, consumers can directly determine energy usage and predict energy usage in advance, which can bring about new changes in national energy policy.

1 is a conceptual diagram used in a conventional energy management system, energy management method, and energy demand forecasting method.
Figure 2 is a conceptual diagram showing a conventional unit household device, environmental management system, and meter server.
Figure 3 is a flowchart of a user-based AI energy saving and demand forecasting system formed in a preferred embodiment of the present invention.
Figure 4 is a regional demand forecasting conceptual diagram of a user-based AI energy saving and demand forecasting system formed as a preferred embodiment of the present invention.
Figure 5 is a state diagram of the artificial intelligence algorithm of the controller of the user-based AI energy saving and demand forecasting system formed in a preferred embodiment of the present invention.
Figure 6 is a flowchart of the artificial intelligence algorithm of the controller of the user-based AI energy saving and demand forecasting system formed in a preferred embodiment of the present invention.
Figure 7 is a flowchart of energy saving by a user-based AI energy saving and demand forecasting system formed in a preferred embodiment of the present invention.

Before describing specific embodiments of the present invention, the drawings of the present specification are used to more clearly explain the present invention, and the sizes and shapes of components shown in the drawings are somewhat exaggerated or exaggerated to make the description clear. It can be displayed simplified.

In addition, in order to clearly explain the present invention, descriptions of parts unrelated to the technical idea of the present invention have been omitted, and identical or similar components have been described with the same reference numerals throughout the specification.

Since the terms and symbols defined in the present invention are terms arbitrarily defined or selectively used by users, operators, and authors, these terms have meanings that are consistent with the technical idea of the present invention based on the entire contents of the specification and It should be interpreted as a concept and should not be limited to the meaning of the term itself.

The present invention provides each unit household device including a total heat exchanger, air purifier, and kitchen hood, a hot water distributor, a differential pressure flow control valve, a water leak detection sensor, and a PTS sensor, and a meter that measures the usage of hot water, hot water, gas, and heating, respectively. connected, a temperature sensor, humidity sensor, fine dust concentration sensor, and carbon dioxide concentration measurement sensor are formed inside, a PID control device is formed, a control device capable of deep learning data with an artificial intelligence neural network is formed, and a network In an environment where a controller 200 capable of being connected to is installed in each household, with each controller 200, user terminal 300, and main server 100 connected to the Internet,

A registration step in which the main server 100 registers each controller 200;

A tracking step in which the main server 100 determines whether each controller 200 is using one of the automatic control mode, manual control mode, or eco control mode and tracks each controller 200 in which the eco control mode is set. and;

A classification step of classifying energy usage devices attached to each controller 200 tracked by the main server 100 by energy usage;

A target value setting step in which the main server 100 receives a target energy saving amount (target value) from each controller 200 after the classification step;

A notification step where the main server 100 informs the user terminal 300 of benefits that can be provided when energy is reduced based on the target value;

It consists of a prediction step in which the main server 100 predicts the corresponding energy demand that can be used by each controller 200.

The main server 100 is a server of the central government, local governments, or energy producers (power plants, etc.) that requires energy demand forecasting, and the application range of each regulator 200 and the range of energy supply as needed. The prediction range can be determined.

The tracking step is a step in which the main server 100 determines whether each controller 200 uses the eco control mode. This is a step in which the target value can be set only in the eco control mode.

The target value setting step is a step in which a target energy reduction amount (target value) is input from each controller 200. The target value is to set a certain amount of energy savings (target reduction amount) for a certain period (target period) from the energy usage converted into money. It's a step.

The target period and target reduction amount are used as important data in the prediction stage.

The notification step is a step in which the main server 100 informs the user terminal 300 of the benefits that can be provided when energy is saved based on the target value.

The prediction step is a step in which the main server 100 predicts the energy demand that can be used in each controller 200 and predicts and applies the energy demand based on statistics.

The controller 200 controls each unit household device including a heat exchanger, air purifier, and kitchen hood, a hot water distributor, differential pressure flow control valve, water leak detection sensor, and PTS sensor, and the usage of hot water, hot water, gas, and heating, respectively. Connected to the weighing meter,

A temperature sensor, humidity sensor, fine dust concentration sensor, and carbon dioxide concentration measurement sensor are formed inside, a PID control device is formed, a control device capable of deep learning data with an artificial intelligence neural network is configured, and a network connection is possible. is formed

The unit household devices are not limited to total heat exchangers, air purifiers, and kitchen hoods, and may include all devices related to ventilation and exhaust systems.

The hot water distributor, differential pressure flow control valve, water leak detection sensor, and PTS sensor are sensors related to the hot water and water supply system, and the valve is configured to be controlled by the regulator 200.

Each meter measures its usage, and when the usage data is measured by the controller 200, it is converted into energy usage, converted into money, and used as data.

The temperature sensor, humidity sensor, fine dust concentration sensor, and carbon dioxide concentration measurement sensor continuously store data by time and date, and are automatically controlled by an artificial intelligence neural network based on these data.

A human proximity sensor (proximity sensor) is installed in the controller 200 to determine whether the user is living there or not and provides the information to the artificial intelligence neural network so that it can be used to save energy.

The artificial intelligence neural network uses a control algorithm, which includes a state value transmission step in which the state value (s) input from a specific unit generation is transmitted to a deep reinforcement learning neural network, a reward prediction neural network, and a PID control neural network;

The compensation prediction neural network is composed of an optimal temperature prediction neural network and an energy prediction neural network, and the optimal temperature prediction neural network and the energy prediction neural network learn about the state value (s) and then determine and transmit the compensation value (γ). and;

The deep reinforcement learning neural network consists of an actor neural network and a critical neural network. The critical neural network combines the transmitted state value (s) and the expected action value (A) with the reward value (γ) and the value value ( An action value calculation step in which Q) is calculated and transmitted to the Actor neural network, and the Actor neural network uses the value value (Q) and the state value (s) to calculate the optimal action value (A);

The PID control neural network consists of a PID calculation step in which the action value (A) is stored as learning data, the optimal PID value (K) is calculated using a machine learning technique, and the optimal PID value (K) is calculated and transmitted to the control device.

The control algorithm uses a supervised learning algorithm and operates with the basic algorithm optimized using a training set (learning data).

The learning data is data classified by time period and date for each household (temperature, humidity, Co2 concentration, solar radiation, etc.) collected from the unit household device 11 and the controller (controlled unit).

The longer the period of accumulation of the above-mentioned household data, the better, but for buildings without any accumulated data, it can be calculated by using a dataset that calculates the optimal value by collecting household data in similar environments.

It is appropriate that both the reward prediction neural network and the deep reinforcement learning neural network of the above algorithm are performed with the basic algorithm optimized.

The compensation prediction neural network consists of an optimal temperature prediction neural network and an energy prediction neural network. The optimal temperature prediction neural network applies a compensation function to calculate compensation values for changes in temperature and humidity, and the energy prediction neural network calculates electrical energy usage, A compensation function is applied to calculate compensation for changes in heat energy usage.

The reward prediction neural network may include a scheduler, which may be configured to recognize the user's usage pattern and transmit a compensation value for scheduling it.

Examples of reward functions used in the reward prediction neural network include:

can be used, and when the state value (s) and action value (A) are at a specific moment (t), the reward value (γ) is learned and calculated as the expected value.

The compensation prediction neural network determines the final compensation value (γ) as the average of the compensation values evaluated in the optimal temperature prediction neural network and the energy prediction neural network and transmits it to the deep reinforcement learning neural network.

The deep reinforcement learning neural network consists of an actor neural network and a critical neural network. The critical neural network learns and calculates the value value (Q) using the state value (s) and reward value (γ).

An example of the equation used for the above value (Q) is

You can use the Bellman expectation equation.

The Actor neural network learns and calculates the optimal action value (A) using the state value (s) and value value (Q).

The action value (A) generated by the Actor neural network is transmitted to the Critic neural network and stored as the new previous state value.

The PID value is a value that controls the manipulated amount according to the action value (A), and determines the control amount (manipulated amount) including the proportional, integral, and differential terms.

The PID control neural network stores PID values according to the state value (a) and action value (A) as learning data and learns them to predict the optimal value of the PID value, such as Monte Carlo learning technique, Q-learning technique, etc. You may also use .

Through the PID control neural network of the artificial intelligence neural network, the PID value of each unit generation is transmitted to the controller 200 and can be automatically controlled.

Three modes can be formed in the controller 200 of the present invention, which can be configured as an automatic control mode, manual control mode, and eco control mode.

The automatic control mode is a mode that completely uses an artificial intelligence neural network to control the environment, and the manual control mode is a mode that allows the user to directly control the environment.

The eco control mode allows the user to set a goal for energy saving, and the artificial intelligence neural network controls unit generation devices and valves so that the user can save energy while being comfortable.

In the present invention described above, when a user sets an energy saving target through the controller 200, this is notified to the main server 100, and the main server 100 can provide benefits to the user.

In addition, the main server 100 collects this energy saving target data and can stably predict future energy demand.

The compensation prediction neural network of the artificial intelligence neural network is composed of an optimal temperature prediction neural network and an energy prediction neural network, and allows calculation of compensation values by considering not only temperature but also energy usage data, thereby creating a comfortable environment as well as highly efficient energy savings. The effect occurs.

According to the user-based AI energy saving and demand forecasting system formed as a preferred embodiment of the present invention, the target value selected for saving energy to be used in the future can be used as predictive data without using the energy pattern used by the user. If the user chooses to save energy, not only can he receive separate benefits, but he can also save money by reducing the amount used, and even if the user selects the eco mode to save energy, the controller's artificial intelligence The control device provides the user with the most comfortable environment possible.

Although the present invention has been described with a focus on preferred embodiments with reference to the accompanying drawings, it is clear to those skilled in the art that various modifications can be made without departing from the scope of the present invention encompassed by the claims described later from this description.

100: Main server (100) 200: Controller (200)
300: User terminal (300)

Claims (4)

delete delete delete In an environment where each controller is installed in each household and each controller, user terminal, and main server are connected to the Internet, a registration step in which the main server registers each controller; A tracking step in which the main server determines whether each controller is using one of the automatic control mode, manual control mode, or eco control mode and tracks each controller in which the eco control mode is set; a classification step of classifying energy-using devices attached to each controller tracked by the main server by energy usage; a target value setting step in which the main server receives a target energy saving amount (target value) from each controller after the classification step; a notification step of notifying the user terminal of benefits that can be provided when the main server saves energy based on the target value; It consists of a prediction step in which the main server predicts the corresponding energy demand that can be used in each controller,
The controller includes each unit household device including a heat exchanger, air purifier, and kitchen hood, a hot water distributor, differential pressure flow control valve, water leak detection sensor, and PTS sensor, and a meter that measures the usage of hot water, domestic hot water, gas, and heating, respectively. It is connected to the controller, and a temperature sensor, humidity sensor, fine dust concentration sensor, and carbon dioxide concentration measurement sensor are formed inside the controller, a PID control device is formed, and a control device capable of deep learning data using an artificial intelligence neural network is formed. It works,
The artificial intelligence neural network of the controller uses a control algorithm, and the control algorithm includes a state value transmission step in which the state value (s) input from a specific unit generation is transmitted to a deep reinforcement learning neural network, a compensation prediction neural network, and a PID control neural network;
The compensation prediction neural network is composed of an optimal temperature prediction neural network and an energy prediction neural network, and the optimal temperature prediction neural network and the energy prediction neural network learn about the state value (s) and then determine and transmit the compensation value (γ). and;
The deep reinforcement learning neural network consists of an Actor neural network and a Critic neural network. The Critic neural network combines the transmitted state value (s) and the expected action value (A) with the reward value (γ) and the value value ( An action value calculation step in which Q) is calculated and transmitted to the Actor neural network, and the Actor neural network uses the value value (Q) and the state value (s) to calculate the action value (A);
In the PID control neural network, the user-based PID calculation step is comprised of storing the action value (A) as learning data, calculating the PID value (K) using a machine learning technique, and transmitting it to the control device. AI energy saving and demand forecasting system.