Building energy consumption control method and device based on GAT and multi-agent reinforcement learning

Abstract

The present invention provides a building energy consumption control method and device based on GAT and multi-agent reinforcement learning, which can be applied to the field of artificial intelligence and intelligent building technology. The method includes: using monitoring equipment to obtain initial environmental information and initial energy consumption information of the building; processing the initial environmental information to determine the correlation between the first functional area and the second functional area; updating the initial environmental information based on the correlation and adjacent feature information to obtain target environmental information; using the target environmental information and a preset analysis strategy to determine the environmental assessment result corresponding to the building; and regulating the initial energy consumption information based on the environmental assessment result and resource dynamic information to obtain target energy consumption information, wherein the target energy consumption information is less than the preset energy consumption threshold at the target time. This method can improve the accuracy of building energy consumption assessment and control results, has universal applicability for different scenarios, and improves the efficiency of building energy consumption assessment and control.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence and intelligent building technology, and in particular to a building energy consumption control method and device based on GAT and multi-agent reinforcement learning. Background Art

[0002] Building energy consumption control can not only significantly reduce energy costs and operating costs, but also improve building comfort, environmental friendliness and asset value. By effectively managing building energy consumption, multiple goals of economic, environmental and social benefits can be achieved. The relevant technologies mainly achieve building energy consumption control by establishing a building thermodynamic model, such as basic regulation control (Regulation-Based Control, RBC), which mainly uses feedback control strategies to adjust control variables so that the system output is close to the desired set value; Model Predictive Control (MPC), which mainly uses the system's mathematical model to predict the system behavior in the future, and achieves energy consumption control by optimizing control inputs.

[0003] Research has found that in the relevant technologies for building energy consumption analysis and control, due to the characteristics of building energy consumption prediction itself, such as many disturbance factors and complex and changeable environment, the generalization of traditional models is poor, and it is difficult to flexibly respond to different building environments; the reinforcement learning of a single agent cannot achieve distributed control, cannot meet the requirements of collaborative control and multi-region optimization, and the overall building energy consumption optimization effect is poor. Summary of the invention

[0004] In view of the above problems, the present invention provides a building energy consumption control method, device, equipment, storage medium and program product based on GAT and multi-agent reinforcement learning.

[0005] According to a first aspect of the present invention, a building energy consumption control method based on GAT (Graph Attention Networks) and multi-agent reinforcement learning is provided, comprising: obtaining initial environmental information and initial energy consumption information of the building by using a monitoring device, wherein the monitoring device corresponds to a functional area in the building, the functional area includes a first functional area and a second functional area, the first functional area is adjacent to the second functional area, and a plurality of the functional areas correspond one-to-one to a plurality of agents; processing the initial environmental information to determine the correlation between the first functional area and the second functional area; updating the initial environmental information based on the correlation and adjacent feature information to obtain target environmental information; determining an environmental assessment result corresponding to the building by using the target environmental information and a preset analysis strategy; and regulating the initial energy consumption information based on the environmental assessment result and resource dynamic information to obtain target energy consumption information, wherein the target energy consumption information is less than a preset energy consumption threshold at the target time.

[0006] According to an embodiment of the present invention, the above-mentioned initial environmental information is processed to determine the correlation between the above-mentioned first functional area and the above-mentioned second functional area, including: encoding the above-mentioned initial environmental information to obtain initial feature information; and using the above-mentioned initial feature information and the above-mentioned adjacent feature information to determine the correlation between the above-mentioned first functional area and the above-mentioned second functional area.

[0007] According to an embodiment of the present invention, the correlation between the first functional area and the second functional area is determined using the initial feature information and the adjacent feature information, including: determining the initial correlation between the first functional area and the second functional area based on the initial feature information and the adjacent feature information; and normalizing the initial correlation to obtain the correlation.

[0008] According to an embodiment of the present invention, the above-mentioned initial environment information is updated based on the above-mentioned correlation degree and adjacent feature information to obtain the target environment information, including: determining the target feature information based on the above-mentioned correlation degree and the above-mentioned adjacent feature information; and updating the above-mentioned initial environment information using the above-mentioned target feature information to obtain the above-mentioned target environment information.

[0009] According to an embodiment of the present invention, the above-mentioned target environment information and a preset analysis strategy are used to determine an environmental assessment result corresponding to the above-mentioned building, including: determining a plurality of operation evaluation information corresponding to the plurality of the above-mentioned functional areas based on the above-mentioned target environment information and the preset analysis strategy; and determining the above-mentioned environmental assessment result based on the plurality of the above-mentioned operation evaluation information.

[0010] According to an embodiment of the present invention, the above method also includes: forming information pairs of operation information and control results corresponding to multiple historical time periods, and storing them in an information library, so that users can retrieve target control information from the above information library according to target time period requirements.

[0011] According to an embodiment of the present invention, the target characteristic information is used to update the initial environmental information to obtain the target environmental information, including: acquiring target parameters and target adjacent characteristic information corresponding to the target characteristic information; and updating the initial environmental information using the target parameters and the target adjacent characteristic information to obtain the target environmental information, wherein the observation area corresponding to the target environmental information is larger than the observation area corresponding to the initial environmental information.

[0012] According to an embodiment of the present invention, the method further includes: obtaining evaluation energy consumption information corresponding to the environmental evaluation result, wherein the evaluation energy consumption information is used to characterize energy consumption information of the environmental evaluation result when the preset energy consumption threshold is met.

[0013] According to an embodiment of the present invention, the above-mentioned initial energy consumption information is regulated based on the above-mentioned environmental assessment results and resource dynamic information to obtain target energy consumption information, including: comparing the above-mentioned initial energy consumption information with the above-mentioned assessment energy consumption information to obtain a comparison result; and when the above-mentioned comparison result indicates that the above-mentioned initial energy consumption information is greater than the above-mentioned assessment energy consumption information, the above-mentioned initial energy consumption information is updated through a target operation based on the above-mentioned resource dynamic information to obtain the above-mentioned target energy consumption information.

[0014] The second aspect of the present invention provides a building energy consumption control device based on GAT and multi-agent reinforcement learning, including: an information acquisition module, used to use monitoring equipment to obtain initial environmental information and initial energy consumption information of the building, wherein the above-mentioned monitoring equipment corresponds to the functional area in the above-mentioned building, the above-mentioned functional area includes a first functional area and a second functional area, the above-mentioned first functional area is adjacent to the above-mentioned second functional area, and the multiple functional areas correspond one-to-one to multiple agents; a correlation determination module, used to process the above-mentioned initial environmental information and determine the correlation between the above-mentioned first functional area and the above-mentioned second functional area; an initial environmental information updating module, used to update the above-mentioned initial environmental information based on the above-mentioned correlation and adjacent feature information to obtain target environmental information; an environmental assessment result determination module, used to determine the environmental assessment result corresponding to the above-mentioned building using the above-mentioned target environmental information and a preset analysis strategy; and an information regulation module, used to regulate the above-mentioned initial energy consumption information based on the above-mentioned environmental assessment result and resource dynamic information to obtain target energy consumption information, wherein the above-mentioned target energy consumption information is less than a preset energy consumption threshold at the target time.

[0015] A third aspect of the present invention provides an electronic device, comprising: one or more processors; a memory for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors execute the above method.

[0016] The fourth aspect of the present invention also provides a computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, causes the processor to execute the above method.

[0017] The fifth aspect of the present invention also provides a computer program product, including a computer program, which implements the above method when executed by a processor.

[0018] According to the building energy consumption control method, device, equipment, storage medium and program product based on GAT and multi-agent reinforcement learning provided by the present invention, the initial environmental information of different functional areas in the building is processed to obtain the correlation between the first functional area and the adjacent functional area, so as to obtain the environmental assessment result of the building according to the correlation, adjacent feature information and preset analysis strategy, and on this basis, the initial energy consumption information is adjusted and controlled to obtain the target energy consumption information. Since the environmental assessment result is determined according to the correlation between multiple functional areas, and the target energy consumption information is obtained by regulation based on resource dynamic information, the defect of only considering the energy consumption information, environmental information and resource information of a single area in the related technology is avoided, the multi-area collaborative control is met, the accuracy of the overall energy consumption assessment and control results of the building is improved, and it has universal applicability to different scenarios, further improving the flexibility and efficiency of building energy consumption assessment and control. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The above and other objects, features and advantages of the present invention will become more apparent through the following description of the embodiments of the present invention with reference to the accompanying drawings, in which:

[0020] Figure 1 An application scenario diagram of a building energy consumption control method, device, equipment, storage medium, and program product based on GAT and multi-agent reinforcement learning according to an embodiment of the present invention is shown;

[0021] Figure 2 A flow chart of a building energy consumption control method based on GAT and multi-agent reinforcement learning according to an embodiment of the present invention is shown;

[0022] Figure 3 A flow chart showing a method for determining an environmental assessment result according to an embodiment of the present invention;

[0023] Figure 4 A system architecture diagram of an initial model (G-Qmix) training method according to an embodiment of the present invention is shown;

[0024] Figure 5 It shows a structural block diagram of a building energy consumption information evaluation and control device based on GAT and multi-agent reinforcement learning according to an embodiment of the present invention;

[0025] Figure 6 A block diagram of an electronic device suitable for implementing a building energy consumption control method based on GAT and multi-agent reinforcement learning according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0026] Below, embodiments of the present invention will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present invention. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of embodiments of the present invention. However, it is apparent that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of known structures and technologies are omitted to avoid unnecessary confusion of concepts of the present invention.

[0027] The terms used herein are only for describing specific embodiments and are not intended to limit the present invention. The terms "comprise", "include", etc. used herein indicate the existence of the features, steps, operations and / or components, but do not exclude the existence or addition of one or more other features, steps, operations or components.

[0028] All terms (including technical and scientific terms) used herein have the meanings commonly understood by those skilled in the art unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0029] When using expressions such as "at least one of A, B, and C, etc.", they should generally be interpreted according to the meaning of the expression commonly understood by those skilled in the art (for example, "a system having at least one of A, B, and C" should include but is not limited to a system having A alone, B alone, C alone, A and B, A and C, B and C, and / or A, B, C, etc.).

[0030] In the technical solution of the present invention, the user information (including but not limited to user personal information, user image information, user device information, such as location information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved are all information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of the relevant data comply with relevant laws, regulations and standards, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0031] In the process of conceiving the present invention, it was found through research that in the relevant technologies, the building energy consumption control lacks consideration of the characteristics of building energy consumption prediction itself, such as many disturbance factors and complex and changeable environment. The generalization of traditional models is poor, and it is difficult to flexibly respond to different building environments; the reinforcement learning of a single intelligent agent cannot achieve distributed control, cannot meet the requirements of collaborative control and multi-region optimization, and the overall building energy consumption optimization effect is poor.

[0032] In view of this, the present invention processes the initial environmental information of different functional areas in the building to obtain the correlation between the first functional area and the adjacent functional areas, and thus obtains the environmental assessment result of the building according to the correlation, adjacent feature information and the preset analysis strategy. On this basis, the initial energy consumption information is adjusted and controlled to obtain the target energy consumption information. Since the environmental assessment result is determined according to the correlation between multiple functional areas, and the target energy consumption information is obtained by regulation based on resource dynamic information, it avoids the defects of the related technology of only considering the energy consumption information, environmental information and resource information of a single area, meets the multi-area collaborative control, improves the accuracy of the overall energy consumption assessment and control results of the building, and has universal applicability to different scenarios, further improving the flexibility and efficiency of the building energy consumption assessment and control.

[0033] An embodiment of the present invention provides a building energy consumption control method, device, equipment, storage medium and program product based on GAT and multi-agent reinforcement learning, the method comprising: using a monitoring device to obtain initial environmental information and initial energy consumption information of a building, wherein the monitoring device corresponds to a functional area in the building, the functional area comprises a first functional area and a second functional area, and the first functional area is adjacent to the second functional area; processing the initial environmental information to determine the correlation between the first functional area and the second functional area; updating the initial environmental information based on the correlation and adjacent feature information to obtain target environmental information; using the target environmental information and a preset analysis strategy to determine an environmental assessment result corresponding to the building; and regulating the initial energy consumption information based on the environmental assessment result and resource dynamic information to obtain target energy consumption information, wherein the target energy consumption information is less than a preset energy consumption threshold at the target time.

[0034] Figure 1 An application scenario diagram of a building energy consumption control method, device, equipment, storage medium, and program product based on GAT and multi-agent reinforcement learning according to an embodiment of the present invention is shown.

[0035] like Figure 1 As shown, the application scenario according to this embodiment may include a first terminal device 101, a second terminal device 102, a third terminal device 103, a network 104, a server 105, and a monitoring device 106. The network 104 is used to provide a medium for a communication link between the first terminal device 101, the second terminal device 102, the third terminal device 103, the server 105, and the monitoring device 106. The network 104 may include various connection types, such as wired, wireless communication links, or optical fiber cables, etc.

[0036] The monitoring device 106 can be referred to as an agent or as a part of an agent. An agent is an independent decision-making unit in multi-agent reinforcement learning. Each agent can observe the environmental status of an independent functional area, take actions, and obtain feedback from the environment. Agents include but are not limited to air quality sensors, temperature sensors, humidity sensors, air pressure sensors, air flow meters, light sensors, noise meters, or multi-functional environmental monitors, or monitoring systems formed by integrating multiple monitors, such as a monitoring system for heating, ventilation, and air conditioning (HVAC). An agent corresponds to each independent functional area in a building (such as a machine room or other functional room). The entire building can be controlled or adjusted by multiple agents to control or adjust the building energy consumption. It is understandable that an agent can be an intelligent decision-making system integrating a monitoring device 106, a server 105, and other functional service devices. An agent can include a first agent and multiple second agents adjacent to the first agent. The monitoring device 106 can be connected to the server 105 or a server cluster through a network 104.

[0037] The user may use at least one of the first terminal device 101, the second terminal device 102, and the third terminal device 103 to interact with the server 105 through the network 104 to receive or send messages, etc. Various communication client applications may be installed on the first terminal device 101, the second terminal device 102, and the third terminal device 103, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social platform software, etc. (only for example).

[0038] The first terminal device 101, the second terminal device 102, and the third terminal device 103 may be various electronic devices having display screens and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, desktop computers, and the like.

[0039] The server 105 may be a server that provides various services, such as a background management server (only as an example) that provides support for websites browsed by users using the first terminal device 101, the second terminal device 102, and the third terminal device 103. The background management server may analyze and process the received data such as user requests, and feed back the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal device.

[0040] It should be noted that the building energy consumption information evaluation control method based on GAT (Graph Attention Networks) and multi-agent reinforcement learning provided in the embodiment of the present invention can generally be executed by the server 105. Accordingly, the building energy consumption control device based on GAT and multi-agent reinforcement learning provided in the embodiment of the present invention can generally be set in the server 105. The building energy consumption control method based on GAT and multi-agent reinforcement learning provided in the embodiment of the present invention can also be executed by a server or server cluster that is different from the server 105 and can communicate with the first terminal device 101, the second terminal device 102, the third terminal device 103 and / or the server 105. Accordingly, the building energy consumption control device based on GAT and multi-agent reinforcement learning provided in the embodiment of the present invention can also be set in a server or server cluster that is different from the server 105 and can communicate with the first terminal device 101, the second terminal device 102, the third terminal device 103 and / or the server 105.

[0041] It should be understood that Figure 1 The number of terminal devices, networks, monitoring devices and servers in the embodiment is only for illustration. Any number of terminal devices, monitoring device networks and servers may be provided according to the implementation requirements.

[0042] Figure 2 A flow chart of a building energy consumption control method based on GAT and multi-agent reinforcement learning according to an embodiment of the present invention is shown.

[0043] like Figure 2 As shown, the building energy consumption control method based on GAT and multi-agent reinforcement learning of this embodiment may include operations S210 to S250.

[0044] In operation S210, initial environmental information and initial energy consumption information of the building are obtained using a monitoring device, wherein the monitoring device corresponds to a functional area in the building, the functional area includes a first functional area and a second functional area, the first functional area is adjacent to the second functional area, and the multiple functional areas correspond one-to-one to the multiple intelligent entities.

[0045] According to an embodiment of the present invention, the initial environmental information may include initial temperature information, initial humidity information, initial air pollution information, initial noise information, and initial oxygen content information. The initial energy consumption information may characterize the building energy consumption information in the initial state. The functional area may characterize the independent functional rooms in the building. It is understandable that the entire building generally includes multiple independent functional areas, each functional area may have other adjacent functional areas, and in multi-agent reinforcement learning, multiple agents may correspond one-to-one to multiple functional areas.

[0046] In operation S220, the initial environment information is processed to determine the degree of association between the first functional area and the second functional area.

[0047] According to an embodiment of the present invention, the degree of association can represent the degree of association between adjacent different functional areas. Based on the multi-agent algorithm, when each independent functional area is regarded as an agent, the degree of association can represent the attention weight value between different agents. The larger the attention weight value, the greater the degree of association between the two agents, and the greater the role in the evaluation and control of building energy consumption.

[0048] In operation S230, the initial environment information is updated based on the association degree and the adjacent feature information to obtain the target environment information.

[0049] According to an embodiment of the present invention, the adjacent feature information may represent the feature information of a functional area (second functional area) adjacent to the first functional area. It is understandable that there may be multiple second functional areas, and the first functional area and the second functional area in the present invention are only for illustration and distinction, and are not limited here. The adjacent feature information may also represent the feature information of a functional area (first functional area) adjacent to the second functional area. The target environment information may represent the environment information obtained after the adjacent feature information is integrated.

[0050] In operation S240, an environmental assessment result corresponding to the building is determined using the target environment information and a preset analysis strategy.

[0051] According to an embodiment of the present invention, the preset analysis strategy may be an intensity learning algorithm for evaluating the building environment. The environmental evaluation result may be an overall indoor environment evaluation result after integrating the entire building or multiple buildings in the entire area.

[0052] In operation S250, the initial energy consumption information is regulated based on the environmental assessment result and the resource dynamic information to obtain target energy consumption information, wherein the target energy consumption information is less than a preset energy consumption threshold at the target time.

[0053] According to an embodiment of the present invention, resource dynamic information can represent resource usage information and resource price information in different time periods / at different times, and resources include but are not limited to electric energy resources, natural gas resources and water resources information. It can be understood that the energy consumption value of the entire building / functional area in the target energy consumption information is less than the energy consumption value of the entire building / functional area in the initial energy consumption information.

[0054] According to an embodiment of the present invention, by processing the initial environmental information of different functional areas in a building, the correlation between the first functional area and the adjacent functional area is obtained, so as to obtain the environmental assessment result of the building according to the correlation, adjacent characteristic information and a preset analysis strategy, and on this basis, the initial energy consumption information is adjusted and controlled to obtain the target energy consumption information. Since the environmental assessment result is determined according to the correlation between multiple functional areas, and the target energy consumption information is obtained by regulation based on resource dynamic information, the defect of only considering the energy consumption information, environmental information and resource information of a single area in the related technology is avoided, the multi-area collaborative control is met, the accuracy of the overall energy consumption assessment and control results of the building is improved, and it has universal applicability to different scenarios, further improving the flexibility and efficiency of the building energy consumption assessment control.

[0055] According to an embodiment of the present invention, the initial environmental information is processed to determine the association between the first functional area and the second functional area, including: encoding the initial environmental information to obtain initial feature information; and using the initial feature information and adjacent feature information to determine the association between the first functional area and the second functional area.

[0056] According to an embodiment of the present invention, the initial environmental information can represent the environmental information obtained by monitoring the environment (temperature, humidity, air quality, noise and oxygen content) of the first functional area through the first intelligent agent in a historical time period / time point, and multiple intelligent agents correspond one-to-one to multiple functional areas.

[0057] In a feasible embodiment, each agent can obtain the initial environmental information such as temperature, humidity, PM2.5, noise, CO2, etc. of each functional area through monitoring, forming a local observation at time t. The initial environmental information can be recorded as , .

[0058] According to an embodiment of the present invention, the initial environment information can be encoded by a deep learning model to learn the complex features of the initial environment information, and obtain the initial feature information that can better represent the environment information, which is recorded as It can be understood that the adjacent feature information of the second functional area adjacent to the first functional area can be obtained by the same method, that is, The degree of association can represent the degree of association (attention weight) between the first functional area and all adjacent second functional areas. The larger the attention weight value, the higher the degree of association between the first functional area and the second functional area.

[0059] In a feasible embodiment, the initial feature information can be obtained through a multi-layer perception model (MLP), including: the initial environment information of the first agent As input features, they are input to the MLP model; through multiple hidden layers (each layer contains multiple neurons), the input features are transformed nonlinearly. These transformations are usually performed through activation functions (such as ReLU), which enables the model to capture complex patterns in the data; at the same time, each neuron in the hidden layer has corresponding weights and biases, which are trained through the back-propagation algorithm to optimize these parameters to minimize the prediction error; the hidden layer can summarize the extracted features and pass them to the output layer to generate initial feature information; finally, the initial feature information is input into the Graph Attention Networks (GAT) model to realize the observation of the status of important adjacent second agents (important neighbor nodes). It can be understood that the agents can correspond to the nodes in the graph attention model.

[0060] According to an embodiment of the present invention, the correlation between the first functional area and the second functional area is determined using initial feature information and adjacent feature information, including: determining the initial correlation between the first functional area and the second functional area based on the initial feature information and the adjacent feature information; and normalizing the initial correlation to obtain the correlation.

[0061] According to an embodiment of the present invention, the initial correlation degree can represent the attention value between the first agent and the second agent, and the correlation degree can represent the attention weight between the first agent and multiple adjacent second agents obtained by processing the attention value using an activation function and a normalization function.

[0062] In a feasible embodiment, the method of using the graph attention model to determine the degree of association includes: constructing an attention value e between a first agent and a plurality of adjacent second agents i，k , e i，k The calculation method is as shown in the following formula (1):

[0063] (1);

[0064] Among them, e i，k represents the initial association (attention weight) between the first agent i and multiple adjacent second agents k, and W can represent the weight parameter, is the initial feature information, It is the adjacent feature information.

[0065] Furthermore, the attention weights can be processed using activation functions (such as LeakyReLU) and normalization functions (such as softmax) to obtain the relevance (Attention coefficient), The calculation method is shown in the following formula (2):

[0066] (2);

[0067] Among them, N i Represent all neighboring agents of the first agent i, It can characterize the characteristic information of all neighboring agents of the first agent i.

[0068] According to an embodiment of the present invention, the correlation between the first intelligent agent and the second intelligent agent is dynamically determined by comprehensively considering adjacent feature information, thereby avoiding the defect of considering only the energy consumption information of a single area, meeting the needs of multi-area coordinated control, enhancing the intelligent agent's understanding and reference of the adjacent environment, and enhancing the flexibility and robustness of the model.

[0069] According to an embodiment of the present invention, the initial environment information is updated based on the correlation degree and the adjacent feature information to obtain the target environment information, including: determining the target feature information based on the correlation degree and the adjacent feature information; and updating the initial environment information using the target feature information to obtain the target environment information.

[0070] According to an embodiment of the present invention, the target feature information may represent the feature information of the first functional area obtained by combining the initial feature information with the neighbor feature information. The target environment feature information may represent the feature information obtained by adaptively allocating weights of different neighbor nodes by performing an aggregation operation on neighbor nodes (second agent) using an attention model (e.g., a graph attention model).

[0071] In a feasible embodiment, the obtained correlation between the first agent i and the second agent k is , adjacent feature information By using the activation function, the target feature information of the first agent can be obtained; by expanding the observation range of the first agent, based on the target feature information, the target environment information after integrating the larger observation range can be obtained.

[0072] According to an embodiment of the present invention, target environmental information and preset analysis strategies are used to determine an environmental assessment result corresponding to a building, including: determining multiple operation evaluation information corresponding to multiple functional areas based on the target environmental information and the preset analysis strategy; and determining the environmental assessment result based on the multiple operation evaluation information.

[0073] According to an embodiment of the present invention, the operation evaluation information can characterize the environmental and energy consumption effects achieved by each agent on the environmental information and energy consumption information when performing the corresponding action operation. The preset analysis strategy is used as a deep learning algorithm for evaluating the building environment. The environmental evaluation result can characterize the environmental evaluation information obtained by comprehensively considering the operation evaluation information of all agents. The operation evaluation information of multiple functional areas is determined based on the target environmental information, the deep learning algorithm and the neural network, so that the multiple operation evaluation information is mixed and processed by the neural network to obtain the environmental evaluation result.

[0074] For example, the preset analysis strategy can be a deep reinforcement learning algorithm (Deep Q-Network, DQN) combined with a neural network (Long Short-Term Memory, LSTM), such as Deep Recurrent Q-Learning for Partially Observable MDPs (DRQN) can be used to characterize and solve partially observable Markov decision processes. It should be noted that the specific strength learning algorithm can be determined according to actual needs and is not limited here.

[0075] Figure 3 A flow chart of a method for determining an environmental assessment result according to an embodiment of the present invention is shown.

[0076] like Figure 3 As shown, the method for determining the environmental assessment result may include the following operations S310 to S330.

[0077] In operation S310, the indoor environment information is observed. The action / operation (such as lowering the air-conditioning temperature, expanding the air outlet, or closing doors and windows, etc.) performed by the first agent i at the previous moment (time t-1) can be recorded as , through which the environmental state s in the corresponding functional area at the current time t can be affected t , so that the first agent i observes the current indoor environment information.

[0078] In operation S320, the operation evaluation information is determined. DRQN ​​(Deep Recurrent Q-Learning for Partially Observable MDPs) can be used to collect the value (Q value function) of each agent's own operation / action, and then obtain the operation evaluation information, which is recorded as , , characterizing the joint action-observation history of the agent.

[0079] In operation S330, the environmental assessment result is determined. Using a neural network (such as MIX Network), the output results of all agent i networks can be As input, a mixture of all agents , get the global Q value Q total The MIX Network contains two hidden parameters W1 and W2, which are used to construct the loss function to achieve Q total It should be noted that the loss function can be expressed as the mean square error between the predicted joint Q value and the target joint Q value; the action of each agent at the current moment Through the policy function to be sure.

[0080] According to the embodiments of the present invention, the operation information of the intelligent agent is processed in real time through deep learning and neural network models, which satisfies the dynamic update of environmental assessment results and provides the latest environmental information, thus achieving real-time monitoring and rapid response to different functional areas of the building. At the same time, the deep learning system can self-adjust and optimize according to new information, thereby adapting to changes in environmental conditions and improving the accuracy of prediction and control results.

[0081] According to an embodiment of the present invention, the method further includes: forming information pairs consisting of operation information and control results corresponding to multiple historical time periods, and storing them in an information library, so that the user can retrieve target control information from the information library according to the requirements of the target time period.

[0082] According to an embodiment of the present invention, taking into account the diversity of building types (such as industrial buildings, residential buildings and public buildings), the information collected historically can be used to generate an information database, which can provide information retrieval according to subsequent actual application needs.

[0083] In a feasible embodiment, the construction method of the information database may include: collecting operation information (such as operation instructions, control parameters) and control results (such as system response, effect data) for different types of buildings in historical time periods; pairing the operation information with the corresponding control results to form information pairs, for example, operation instruction A corresponds to control result B; thereby, a database table structure can be designed to store information pairs, and the table structure can include operation information, control results and other related metadata fields; the paired information pairs are inserted into the designed information database, and a suitable database system (such as SQL or NoSQL) can be selected to store data according to needs.

[0084] According to an embodiment of the present invention, by constructing an information database, the efficiency of information storage and retrieval (information pairs of operation information and regulation results) can be ensured, thereby protecting the integrity and confidentiality of the data. At the same time, an interface or query mechanism is provided according to different actual needs, thereby improving the work efficiency of retrieving and analyzing stored operation information and regulation result pairs from the information database.

[0085] According to an embodiment of the present invention, the target environment information is updated using the target feature information to obtain the target environment information, including: acquiring target parameters and target adjacent feature information corresponding to the target feature information; and updating the initial environment information using the target parameters and the target adjacent feature information to obtain the target environment information, wherein the observation area corresponding to the target environment information is larger than the observation area corresponding to the initial environment information.

[0086] According to an embodiment of the present invention, a method of obtaining target environment information through a graph attention model may include: through a first network layer, based on the relevance and adjacent feature information , determine the target feature information, recorded as , The calculation method is as shown in the following formula (3):

[0087] (3);

[0088] in, is the activation function. Through this operation, the feature information of the first node i and multiple adjacent second nodes k are aggregated to obtain a new feature vector , The updated features of node i in the graph (target feature information) can be represented. In the same way, the target adjacent feature information can be obtained .

[0089] Furthermore, the obtained target adjacent feature information can be used as input information ( ) is input into the second network layer of the model to expand the observation range of the first agent and obtain the target environment information, which is recorded as , The calculation method of can be shown as the following formula (4):

[0090] (4);

[0091] in, These are the preset parameters of the second network layer. It is the label adjacent feature information obtained by the first network layer.

[0092] According to an embodiment of the present invention, the method further includes: acquiring assessed energy consumption information corresponding to the environmental assessment result, wherein the assessed energy consumption information is used to characterize energy consumption information when the environmental assessment result meets a preset energy consumption threshold.

[0093] According to an embodiment of the present invention, a preset evaluation strategy may be used to comprehensively evaluate environmental information, operation information, and environmental evaluation results within a target time period to obtain evaluated energy consumption information corresponding to the building.

[0094] In a feasible embodiment, an energy consumption prediction model (such as a random forest) can be established, and air conditioning control information and indoor environmental information can be used as input to predict energy consumption; the accuracy of the energy consumption prediction model can be verified by comparing the recorded actual energy consumption data (such as meter readings) with the predicted values; indicators such as energy efficiency ratio (EER) or coefficient of performance (COP) can be used to evaluate the energy efficiency performance of the air conditioning system; thereby analyzing the relationship between indoor environmental data and air conditioning operation information, identifying energy consumption patterns and key factors, and on this basis, data mining technology can be used for in-depth analysis; based on the evaluation results, energy consumption optimization strategies can be generated, such as adjusting air conditioning settings, improving indoor environmental conditions, etc., so that the indoor evaluation results meet the relevant indoor environmental standards while the energy consumption meets the requirements of the preset energy consumption threshold.

[0095] According to an embodiment of the present invention, initial energy consumption information is regulated based on environmental assessment results and resource dynamic information to obtain target energy consumption information, including: comparing the initial energy consumption information with the assessment energy consumption information to obtain a comparison result; and when the comparison result indicates that the initial energy consumption information is greater than the assessment energy consumption information, updating the initial energy consumption information through a target operation based on the resource dynamic information to obtain the target energy consumption information.

[0096] According to an embodiment of the present invention, the energy consumption information for evaluation may be an energy consumption threshold determined based on relevant building energy consumption standards. When the initial energy consumption information is greater than the energy consumption threshold, it indicates that the building energy consumption within the target time period is difficult to meet the building energy consumption standards. It is necessary to regulate some energy-consuming equipment in the building (such as air-conditioning equipment and heating equipment) through an intelligent agent so that the target energy consumption information meets the energy consumption standards.

[0097] In a feasible embodiment, taking power resource information as an example, the coordinated configuration between different functional areas of the building can be used to achieve the regulation of building energy consumption, which may include: obtaining and integrating dynamic electricity price data, including peak and valley electricity prices, time period electricity prices and other information, to understand the fluctuations in electricity prices in different time periods; using historical time period electricity demand information, usage information and forecasting models to estimate the electricity demand in each area, and considering the impact of dynamic electricity prices on electricity consumption behavior; generating resource intelligent scheduling strategies based on the above information and optimization algorithms (such as linear programming, genetic algorithms), and adjusting the allocation of power resources according to dynamic electricity prices and electricity demand, for example, giving priority to operating high-energy-consuming equipment during low electricity price periods; implementing regional-level power control strategies, adjusting power supply according to occupancy and electricity prices in different areas, and formulating flexible control strategies taking into account random occupancy; at the same time, real-time monitoring of electricity usage in the building, and dynamically adjusting control strategies according to actual electricity price changes and energy demand.

[0098] In a feasible embodiment, a HVAC control method can be developed based on a graph attention model (GAT) and a multi-agent control algorithm (such as the Q-mix algorithm) to maximize the reduction of the overall energy consumption of a building while fully considering dynamic electricity prices and multi-area random occupancy strategies, targeting the indoor environment and building energy consumption in different functional areas of the building without reducing the living experience and maintaining the five constants (constant temperature, constant humidity, constant cleanliness, constant quietness, and constant oxygen) indoors. It can be understood that the target model (G-Qmix) obtained based on the graph attention model (GAT) and the multi-agent control algorithm (such as the Qmix algorithm) is a model obtained after training.

[0099] In a feasible embodiment, a training method for a target model (G-Qmix) may include: obtaining sample initial environmental information and sample initial energy consumption information of a sample building using a monitoring device, wherein the monitoring device corresponds to a sample functional area in the sample building, and the functional area includes a sample first functional area and a sample second functional area, and the sample first functional area is adjacent to the sample second functional area; processing the sample initial environmental information to determine a sample correlation between the sample first functional area and the sample second functional area; updating the sample initial environmental information based on the sample correlation and sample adjacent feature information to obtain sample target environmental information; and training an initial model using the sample initial environmental information, the sample initial energy consumption information, the sample environmental assessment results, the sample energy consumption information, and their respective corresponding labels to obtain a target model.

[0100] Figure 4 A system architecture diagram of a target model (G-Qmix) training method according to an embodiment of the present invention is shown.

[0101] like Figure 4 As shown, the initial model may include an initial graph attention sub-model 401 and an initial multi-agent sub-model 402. The sample initial environment information 410 of each sample agent in the historical time period is obtained, and the sample initial environment information 410 is encoded through a multi-layer perception network layer to obtain sample initial information 420, and the sample initial information 420 includes sample initial feature information 421 and sample adjacent feature information 422; thereby, the sample initial information 420 is sent to the first network layer of the initial graph attention sub-model 401 to construct an attention value 431 between sample agents, and the sample agent may include a sample first agent and an adjacent sample second agent. The attention value 431 is processed using an activation function and a normalization function to obtain an attention weight 432 between the sample first agent and the sample second agent; based on the attention weight 432, the model is built in combination with the observation results of the adjacent sample second agent to generate target feature information 440.

[0102] Further, the target feature information 440 is used as input to the second network layer of the initial graph attention sub-model 401, and the observation range of each sample agent is expanded to obtain the sample target environment information 450; then, the initial multi-agent sub-model 402 is used to use DRQN ​​to aggregate the Q value function of each sample agent to determine the value information (sample operation evaluation information 460) brought by the operation / action taken in the current state; further, through MIX Network model (initial multi-agent sub-model), takes the sample operation evaluation information 460 as input, mixes the Q values ​​of all sample agents, and determines the sample environment evaluation result 470; the current action of each sample agent can be determined by the strategy function, and by executing the current action, the HVAC can be controlled to affect the temperature, humidity, cleanliness, quietness, and oxygen content, and the sample indoor environment feedback result 480 is obtained; and based on the reward mechanism, the action is rewarded, and the environmental reward result based on the five constant states and the energy consumption reward result based on the energy consumption state are obtained respectively, and then the sample comprehensive reward result 490 is obtained; and the above steps are iterated until the prediction result of the intermediate model meets the preset accuracy requirement, and the intermediate model is determined as the target model 400 (including the target graph attention sub-model and the target multi-agent sub-model).

[0103] In a feasible embodiment, based on the indoor environment feedback result and the preset reward mechanism, the reward result of the action is formed, wherein the reward result of the five constant states can be , the reward result based on the energy consumption state can be , and determine the comprehensive reward result R, which can be calculated by the following formula (5):

[0104] (5);

[0105] in, Represents the weight parameter of the environmental reward result, The weight parameter representing the energy consumption reward result, the specific value can be assigned based on experience or determined through training.

[0106] According to an embodiment of the present invention, by using a graph attention model in coordination with a multi-agent model, the agent can observe the status of neighboring agents, and use the neighbor status as an important component of indoor environment observation, thereby achieving optimal global building environment results and energy consumption results.

[0107] Based on the above-mentioned building energy consumption control method based on GAT and multi-agent reinforcement learning, the present invention also provides a building energy consumption control device based on GAT and multi-agent reinforcement learning. Figure 5 The device is described in detail.

[0108] Figure 5 A structural block diagram of a building energy consumption control device based on GAT and multi-agent reinforcement learning according to an embodiment of the present invention is shown.

[0109] like Figure 5 As shown, the building energy consumption control device based on GAT and multi-agent reinforcement learning of this embodiment includes an information acquisition module 510, a correlation determination module 520, an initial environmental information update module 530, an environmental assessment result determination module 540 and an information regulation module 550.

[0110] The information acquisition module 510 is used to acquire the initial environment information and initial energy consumption information of the building by using the monitoring equipment, wherein the monitoring equipment corresponds to the functional area in the building, the functional area includes a first functional area and a second functional area, the first functional area is adjacent to the second functional area, and the multiple functional areas correspond to the multiple intelligent agents one by one. In one embodiment, the information acquisition module 510 can be used to perform the operation S210 described above, which will not be repeated here.

[0111] The correlation determination module 520 is used to process the initial environment information and determine the correlation between the first functional area and the second functional area. In one embodiment, the correlation determination module 520 can be used to perform the operation S220 described above, which will not be described in detail here.

[0112] The initial environment information updating module 530 is used to update the initial environment information based on the association degree and the adjacent feature information to obtain the target environment information. In one embodiment, the initial environment information updating module 530 can be used to perform the operation S230 described above, which will not be described in detail here.

[0113] The environmental assessment result determination module 540 is used to determine the environmental assessment result corresponding to the building using the target environmental information and the preset analysis strategy. In one embodiment, the environmental assessment result determination module 540 can be used to perform the operation S240 described above, which will not be described in detail here.

[0114] The information regulation module 550 is used to regulate the initial energy consumption information based on the environmental assessment result and the resource dynamic information to obtain the target energy consumption information, wherein the target energy consumption information is less than the preset energy consumption threshold at the target time. In one embodiment, the information regulation module 550 can be used to perform the operation S250 described above, which will not be repeated here.

[0115] According to an embodiment of the present invention, through the information acquisition module 510, the correlation determination module 520, the initial environment information update module 530, the environment assessment result determination module 540 and the information control module 550 in the building energy consumption control device based on GAT and multi-agent reinforcement learning, the initial environment information of different functional areas in the building is processed to obtain the correlation between the first functional area and the adjacent functional area, so as to obtain the environmental assessment result of the building according to the correlation, the adjacent feature information and the preset analysis strategy, and on this basis, the initial energy consumption information is adjusted and controlled to obtain the target energy consumption information. Since the environmental assessment result is determined according to the correlation between multiple functional areas, and the target energy consumption information is obtained by regulation based on resource dynamic information, the defect of only considering the energy consumption information, environmental information and resource information of a single area in the related technology is avoided, the multi-area collaborative control is met, the accuracy of the overall energy consumption assessment and control results of the building is improved, and it has universality for different scenarios, further improving the flexibility and efficiency of building energy consumption assessment control.

[0116] According to an embodiment of the present invention, the association determination module 520 includes: an encoding submodule and an association determination submodule. The encoding submodule is used to encode the initial environment information to obtain initial feature information. The association determination submodule is used to determine the association between the first functional area and the second functional area using the initial feature information and the adjacent feature information.

[0117] According to an embodiment of the present invention, the correlation determination submodule includes: an initial correlation determination unit and a normalization processing unit. The initial correlation determination unit is used to determine the initial correlation between the first functional area and the second functional area based on the initial feature information and the adjacent feature information. The normalization processing unit is used to normalize the initial correlation to obtain the correlation.

[0118] According to an embodiment of the present invention, the initial environment information update module 530 includes: a target feature information determination submodule and an initial environment information update submodule. The target feature information determination submodule is used to determine the target feature information based on the correlation degree and the adjacent feature information. The initial environment information update submodule is used to update the initial environment information using the target feature information to obtain the target environment information.

[0119] According to an embodiment of the present invention, the environment assessment result determination module 540 includes: an operation evaluation information determination submodule and an environment assessment result determination submodule. The operation evaluation information determination submodule is used to determine multiple operation evaluation information corresponding to multiple functional areas based on target environment information and a preset analysis strategy. The environment assessment result determination submodule is used to determine an environment assessment result based on multiple operation evaluation information.

[0120] According to an embodiment of the present invention, the device also includes: an information pair composition module, which is used to compose information pairs of operation information and control results corresponding to multiple historical time periods, and store them in an information library, so that users can retrieve target control information from the information library according to target time period requirements.

[0121] According to an embodiment of the present invention, the initial environment information updating submodule includes: a neighbor feature information acquiring unit and an initial environment information updating unit. The neighbor feature information acquiring unit is used to acquire target parameters and target neighbor feature information corresponding to the target feature information. The initial environment information updating unit is used to update the initial environment information using the target parameters and the target neighbor feature information to obtain the target environment information, wherein the observation area corresponding to the target environment information is larger than the observation area corresponding to the initial environment information.

[0122] According to an embodiment of the present invention, the device also includes: an evaluation energy consumption information acquisition module, which is used to obtain evaluation energy consumption information corresponding to the environmental evaluation result, wherein the evaluation energy consumption information is used to characterize the energy consumption information of the environmental evaluation result when a preset energy consumption threshold is met.

[0123] According to an embodiment of the present invention, the information control module 550 includes: an information comparison submodule and an initial energy consumption information update submodule. The information comparison submodule is used to compare the initial energy consumption information with the evaluation energy consumption information to obtain a comparison result. The initial energy consumption information update submodule is used to update the initial energy consumption information through a target operation based on resource dynamic information to obtain target energy consumption information when the comparison result indicates that the initial energy consumption information is greater than the evaluation energy consumption information.

[0124] According to an embodiment of the present invention, any multiple modules among the information acquisition module 510, the association determination module 520, the initial environment information update module 530, the environment assessment result determination module 540 and the information regulation module 550 can be combined into one module for implementation, or any one of the modules can be split into multiple modules. Alternatively, at least part of the functions of one or more of these modules can be combined with at least part of the functions of other modules and implemented in one module. According to an embodiment of the present invention, at least one of the information acquisition module 510, the association determination module 520, the initial environment information update module 530, the environment assessment result determination module 540 and the information regulation module 550 can be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on a chip, a system on a substrate, a system on a package, an application specific integrated circuit (ASIC), or can be implemented by hardware or firmware such as any other reasonable way of integrating or packaging the circuit, or implemented in any one of the three implementation methods of software, hardware and firmware or in a suitable combination of any of them. Alternatively, at least one of the information acquisition module 510, the association determination module 520, the initial environmental information update module 530, the environmental assessment result determination module 540 and the information regulation module 550 can be at least partially implemented as a computer program module, which can perform corresponding functions when the computer program module is executed.

[0125] Figure 6 A block diagram of an electronic device suitable for implementing a building energy consumption control method based on GAT and multi-agent reinforcement learning according to an embodiment of the present invention is shown.

[0126] like Figure 6 As shown, the electronic device according to an embodiment of the present invention includes a processor 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage part 608 to a random access memory (RAM) 603. The processor 601 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or a related chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 601 may also include an onboard memory for caching purposes. The processor 601 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present invention.

[0127] In RAM 603, various programs and data required for the operation of electronic device 600 are stored. Processor 601, ROM 602 and RAM 603 are connected to each other via bus 604. Processor 601 performs various operations of the method flow according to the embodiment of the present invention by executing the program in ROM 602 and / or RAM 603. It should be noted that the program can also be stored in one or more memories other than ROM 602 and RAM 603. Processor 601 can also perform various operations of the method flow according to the embodiment of the present invention by executing the program stored in the one or more memories.

[0128] According to an embodiment of the present invention, the electronic device 600 may further include an input / output (I / O) interface 605, which is also connected to the bus 604. The electronic device 600 may further include one or more of the following components connected to the input / output (I / O) interface 605: an input portion 606 including a keyboard, a mouse, etc.; an output portion 607 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage portion 608 including a hard disk, etc.; and a communication portion 609 including a network interface card such as a LAN card, a modem, etc. The communication portion 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to the input / output (I / O) interface 605 as needed. A removable medium 611, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 610 as needed, so that a computer program read therefrom is installed into the storage portion 608 as needed.

[0129] The present invention also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiment; or may exist independently without being assembled into the device / apparatus / system. The above computer-readable storage medium carries one or more programs, and when the above one or more programs are executed, the method according to the embodiment of the present invention is implemented.

[0130] According to an embodiment of the present invention, the computer-readable storage medium may be a non-volatile computer-readable storage medium, for example, it may include but is not limited to: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present invention, the computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, an apparatus or a device. For example, according to an embodiment of the present invention, the computer-readable storage medium may include the ROM 602 and / or RAM 603 described above and / or one or more memories other than ROM 602 and RAM 603.

[0131] The embodiment of the present invention also includes a computer program product, which includes a computer program, and the computer program contains program code for executing the method shown in the flowchart. When the computer program product is run in a computer system, the program code is used to enable the computer system to implement the building energy consumption control method based on GAT and multi-agent reinforcement learning provided by the embodiment of the present invention.

[0132] The computer program executes the above functions defined in the system / device of the embodiment of the present invention when it is executed by the processor 601. According to the embodiment of the present invention, the system, device, module, unit, etc. described above can be implemented by a computer program module.

[0133] In one embodiment, the computer program may rely on tangible storage media such as optical storage devices, magnetic storage devices, etc. In another embodiment, the computer program may also be transmitted and distributed in the form of signals on a network medium, and downloaded and installed through the communication part 609, and / or installed from a removable medium 611. The program code contained in the computer program may be transmitted using any appropriate network medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.

[0134] In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 609, and / or installed from the removable medium 611. When the computer program is executed by the processor 601, the above functions defined in the system of the embodiment of the present invention are performed. According to the embodiment of the present invention, the system, device, means, module, unit, etc. described above can be implemented by a computer program module.

[0135] According to an embodiment of the present invention, the program code for executing the computer program provided by the embodiment of the present invention can be written in any combination of one or more programming languages. Specifically, these computing programs can be implemented using high-level process and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, Java, C++, python, "C" language or similar programming languages. The program code can be executed entirely on the user computing device, partially on the user device, partially on the remote computing device, or entirely on the remote computing device or server. In the case of a remote computing device, the remote computing device can be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computing device (e.g., using an Internet service provider to connect through the Internet).

[0136] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present invention. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0137] It will be appreciated by those skilled in the art that the features described in the various embodiments and / or claims of the present invention may be combined and / or combined in various ways, even if such combinations and / or combinations are not explicitly described in the present invention. In particular, the features described in the various embodiments and / or claims of the present invention may be combined and / or combined in various ways without departing from the spirit and teachings of the present invention. All of these combinations and / or combinations fall within the scope of the present invention.

[0138] The embodiments of the present invention are described above. However, these embodiments are only for the purpose of illustration, and are not intended to limit the scope of the present invention. Although each embodiment is described above, it does not mean that the measures in each embodiment cannot be used in combination. The scope of the present invention is defined by the attached claims and their equivalents. Without departing from the scope of the present invention, those skilled in the art may make various substitutions and modifications, which should all fall within the scope of the present invention.

Claims

1. A building energy consumption control method based on GAT and multi-agent reinforcement learning, characterized in that: include: Acquire initial environmental information and initial energy consumption information of the building by using a monitoring device, wherein the monitoring device corresponds to a functional area in the building, the functional area includes a first functional area and a second functional area, the first functional area is adjacent to the second functional area, and the plurality of functional areas correspond one-to-one to the plurality of intelligent agents; Processing the initial environment information to determine a correlation between the first functional area and the second functional area; Based on the correlation degree and adjacent feature information, the initial environment information is updated to obtain target environment information; Determine an environmental assessment result corresponding to the building using the target environmental information and a preset analysis strategy; and The initial energy consumption information is regulated based on the environmental assessment result and the resource dynamic information to obtain target energy consumption information, wherein the target energy consumption information is less than a preset energy consumption threshold at a target time; Wherein, updating the initial environment information based on the association degree and the adjacent feature information to obtain the target environment information includes: Determining target feature information based on the association degree and the adjacent feature information, wherein the adjacent feature information represents feature information of a functional area adjacent to the first functional area, and the association degree is determined based on a graph attention model; Acquire target parameters and target adjacent feature information corresponding to the target feature information; Using the target parameter and the target adjacent feature information, the initial environment information is updated to obtain the target environment information, wherein an observation area corresponding to the target environment information is larger than an observation area corresponding to the initial environment information; Wherein, using the target environment information and a preset analysis strategy to determine an environmental assessment result corresponding to the building includes: Determining a plurality of operation evaluation information corresponding to the plurality of functional areas based on the target environment information and a preset analysis strategy, wherein the preset analysis strategy represents a deep reinforcement learning algorithm combined with a neural network for evaluating a building environment; and The environmental assessment result is determined based on the plurality of operation evaluation information, wherein the environmental assessment result is an overall indoor environmental assessment result after integrating the building or a plurality of buildings in the entire area.

2. The method according to claim 1, characterized in that Processing the initial environment information to determine the correlation between the first functional area and the second functional area includes: Encoding the initial environment information to obtain initial feature information; and The degree of association between the first functional area and the second functional area is determined by using the initial feature information and the adjacent feature information.

3. The method according to claim 2, characterized in that Determining the degree of association between the first functional area and the second functional area using the initial feature information and the adjacent feature information includes: determining an initial degree of association between the first functional area and the second functional area based on the initial feature information and the adjacent feature information; and The initial correlation degree is normalized to obtain the correlation degree.

4. The method according to claim 1, characterized in that: The control information includes operation information and control results corresponding to a plurality of historical time periods, and the method further includes: The operation information and the control result are combined into an information pair and stored in an information library, so that the user can retrieve the target control information from the information library according to the requirements of the target time period.

5. The method according to claim 1, characterized in that: The method further comprises: Acquire assessment energy consumption information corresponding to the environmental assessment result, wherein the assessment energy consumption information is used to characterize energy consumption information when the environmental assessment result meets the preset energy consumption threshold.

6. The method according to claim 5, characterized in that The initial energy consumption information is regulated based on the environmental assessment result and the resource dynamic information to obtain target energy consumption information, including: Comparing the initial energy consumption information with the evaluated energy consumption information to obtain a comparison result; and When the comparison result indicates that the initial energy consumption information is greater than the evaluated energy consumption information, the initial energy consumption information is updated through a target operation based on the resource dynamic information to obtain the target energy consumption information.

7. A building energy consumption control device based on GAT and multi-agent reinforcement learning, characterized in that: include: An information acquisition module, used to acquire initial environmental information and initial energy consumption information of the building by using a monitoring device, wherein the monitoring device corresponds to a functional area in the building, the functional area includes a first functional area and a second functional area, the first functional area is adjacent to the second functional area, and the plurality of functional areas correspond one-to-one to the plurality of intelligent agents; a correlation determination module, configured to process the initial environment information and determine a correlation between the first functional area and the second functional area; An initial environment information updating module, used to update the initial environment information based on the association degree and adjacent feature information to obtain target environment information; an environmental assessment result determination module, configured to determine an environmental assessment result corresponding to the building by using the target environmental information and a preset analysis strategy; and An information control module, configured to control the initial energy consumption information based on the environmental assessment result and resource dynamic information to obtain target energy consumption information, wherein the target energy consumption information is less than a preset energy consumption threshold at a target time; Among them, the initial environment information update module includes: a target feature information determination submodule, configured to determine target feature information based on the correlation degree and the adjacent feature information, wherein the adjacent feature information represents feature information of a functional area adjacent to the first functional area, and the correlation degree is determined based on a graph attention model; A neighbor feature information acquisition unit, used to acquire target parameters and target neighbor feature information corresponding to the target feature information; an initial environment information updating unit, configured to update the initial environment information using the target parameter and the target adjacent feature information to obtain the target environment information, wherein an observation area corresponding to the target environment information is larger than an observation area corresponding to the initial environment information; Among them, the environmental assessment result determination module includes: an operation evaluation information determination submodule, configured to determine a plurality of operation evaluation information corresponding to the plurality of functional areas based on the target environment information and a preset analysis strategy, wherein the preset analysis strategy represents a deep reinforcement learning algorithm combined with a neural network for evaluating a building environment; and The environmental assessment result determination submodule is used to determine the environmental assessment result based on the multiple operation evaluation information, wherein the environmental assessment result is an overall indoor environment assessment result after integrating the building or multiple buildings in the entire area.

Images