A Method for Constructing an Integrated Energy System Operation and Maintenance Decision-Making Agent Model Based on a Large Language Model

Abstract

This invention discloses a method for constructing an operation and maintenance decision-making agent model for integrated energy systems based on a large language model. The method first establishes a state dataset and a domain knowledge base for the integrated energy system; then, based on the domain knowledge base and state data at different times, it constructs a domain knowledge large language model for the integrated energy system; according to different states of the integrated energy system, the domain knowledge large language model generates feasible domains for control strategies, and evaluates each control strategy within these domains; based on the state dataset and the evaluation results of the feasible domains, it constructs an operation and maintenance decision-making agent model for the integrated energy system based on the large language model. By inputting real-time data of the integrated energy system into the operation and maintenance decision-making agent model, the optimal control strategy for that state can be obtained. The output of the model in this invention has a certain degree of interpretability, and this invention also overcomes the limitations of large language models requiring network connectivity or consuming large amounts of computing resources.

Description

Technical Field

[0001] This invention belongs to the field of operation, maintenance and scheduling control of integrated energy systems, and specifically relates to a method for constructing an operation and maintenance decision agent model for integrated energy systems based on a large language model. Background Technology

[0002] With the deepening of the carbon neutrality concept, integrated energy systems are undergoing a transition towards low-carbon development. Integrated energy systems involve a large number of devices, loads, and pipeline nodes, resulting in massive amounts of data, and their complexity is also significant. Therefore, it is difficult to reconstruct a model for control and regulation decisions under different conditions for the operation and maintenance of integrated energy systems. Furthermore, traditional methods often only consider objects that can be described by mathematical formulas, and in actual construction, compromises are often made with the operation and maintenance specifications and requirements of integrated energy systems to ensure model stability and complexity. Therefore, under these circumstances, traditional integrated energy operation and maintenance optimization decision-making algorithms may not fully reflect the real-world conditions of integrated energy systems, and the convergence of the model is also challenged as the complexity of integrated energy systems increases. Summary of the Invention

[0003] The purpose of this invention is to provide a method for constructing an integrated energy system operation and maintenance decision-making agent model based on a large language model. The integrated energy system operation and maintenance decision-making agent model constructed based on this invention can output control strategies that meet the operation and maintenance specifications and requirements of the integrated energy system given a system state. The integrated energy system operation and maintenance decision-making agent model constructed by this invention can consider mathematically difficult-to-express relationships when outputting the optimal control strategies for equipment, and the results obtained have a certain degree of interpretability.

[0004] To achieve the above objectives, the present invention provides the following technical solution:

[0005] A method for constructing an integrated energy system operation and maintenance decision-making agent model based on a large language model includes the following steps:

[0006] S1, construct the state dataset and domain knowledge base of the integrated energy system;

[0007] S2, Based on the domain knowledge base of the integrated energy system and the state data at different times, construct a large language model of the domain knowledge of the integrated energy system;

[0008] S3, based on the different states of the integrated energy system, the domain knowledge big language model of the integrated energy system generates a feasible domain of control strategies, and evaluates each control strategy in the feasible domain of control strategies;

[0009] S4. Based on the state dataset and the evaluation results of the feasible domain of the regulation strategy obtained in step S3, construct the integrated energy system operation and maintenance decision agent model based on the large language model.

[0010] In the above technical solution, step S1 further includes the following steps:

[0011] Step S11, the method for constructing the state dataset of the integrated energy system is as follows: by acquiring historical operating data of each load, equipment, and pipeline node in the integrated energy system, and representing it as a state dataset; the specific construction process is as follows: in time... The following will include historical operating data of the integrated energy system's equipment, load, and pipeline nodes, expressed as time. The state below Then, it is possible to construct a system containing... The state dataset at each time point is represented as follows: ;

[0012] Step S12, the method for constructing the domain knowledge base is as follows: collect the operation planning, safe operation, equipment control and operation characteristics of the integrated energy system and the various equipment pipelines, and establish a domain knowledge base based on this information for subsequent construction of a large language model of domain knowledge for the integrated energy system.

[0013] Furthermore, in step S2, the method for constructing the comprehensive energy system domain knowledge language model is as follows:

[0014] Construct large language model prompt words and use them as model input; the large language model prompt words specifically include the role setting of the integrated energy system, domain knowledge vector, state data at a specific time, and task objectives for generating and evaluating control strategies.

[0015] The control strategy and the evaluation result of the control strategy for the equipment under the corresponding state are used as the output of the model;

[0016] The method for constructing the domain knowledge vector is as follows: the domain knowledge base is divided into finer-grained word blocks, and these word blocks are vectorized and stored to obtain the domain knowledge vector.

[0017] Furthermore, step S3 includes the following steps:

[0018] The prompts corresponding to different states of the integrated energy system are input into the integrated energy system domain knowledge large language model. The integrated energy system domain knowledge large language model outputs the feasible domain of control strategies through a thought chain. At the same time, the integrated energy system domain knowledge large language model evaluates all control strategies in the feasible domain of control strategies based on the operation and maintenance specifications and requirements of the integrated energy system and outputs the evaluation results. The feasible domain of control strategies includes the control strategies of all adjustable devices in the integrated energy system.

[0019] Furthermore, step S4 specifically includes the following steps:

[0020] The states at different times are used as inputs to the integrated energy system operation and maintenance decision-making agent model based on the large language model. The optimal control strategy under the corresponding state is used as the output of the integrated energy system operation and maintenance decision-making agent model based on the large language model. The error between the evaluation result of the feasible region of the control strategy and the output result of the integrated energy system operation and maintenance decision-making agent model based on the large language model under the corresponding state is used as the loss function of the integrated energy system operation and maintenance decision-making agent model based on the large language model under the corresponding state. Finally, contrastive learning is used to train the integrated energy system operation and maintenance decision-making agent model based on the large language model on the generated feasible region of the control strategy.

[0021] The present invention also provides a comprehensive energy system operation and maintenance decision agent model based on a large language model constructed using the above method. When the comprehensive energy system operation and maintenance decision agent model based on a large language model is used in a real comprehensive energy system, the current state of the system is input into the comprehensive energy system operation and maintenance decision agent model based on a large language model, and the comprehensive energy system operation and maintenance decision agent model based on a large language model can output the optimal control strategy under the corresponding state.

[0022] The beneficial effects of this invention are:

[0023] The integrated energy system operation and maintenance decision-making agent model constructed by the method of this invention can output control strategies that meet the operation and maintenance specifications and requirements of integrated energy systems given a system state. The proposed method for constructing the integrated energy system operation and maintenance decision-making agent model employs a large language model of integrated energy system domain knowledge to supervise the learning of the output results of the operation and maintenance decision-making agent model under different system states at different times. This allows for the construction of the operation and maintenance decision-making agent model using natural language that is more in line with human expression habits. It can also consider relations that are difficult to express mathematically, and the results obtained when the constructed agent model outputs the optimal control strategy for the devices have a certain degree of interpretability. Furthermore, this invention overcomes the limitations of large language models requiring network connectivity or consuming large amounts of computing resources. Attached Figure Description

[0024] Figure 1 This is a flowchart of the method for constructing an integrated energy system operation and maintenance decision agent model based on a large language model as described in this invention;

[0025] Figure 2 This is a schematic diagram of the construction method for the integrated energy system operation and maintenance decision agent model based on a large language model as described in this invention. Detailed Implementation

[0026] The present invention will now be further described in conjunction with the accompanying drawings and specific embodiments.

[0027] This invention provides a method for constructing an integrated energy system operation and maintenance decision-making agent model based on a large language model. The construction process and principle are as follows: Figure 1 and Figure 2 As shown, the method specifically includes the following steps:

[0028] S1, construct the state dataset and domain knowledge base of the integrated energy system;

[0029] S2, Based on the domain knowledge base of the integrated energy system and the state data at different times, construct a large language model of the domain knowledge of the integrated energy system;

[0030] S3, based on the different states of the integrated energy system, the domain knowledge big language model of the integrated energy system generates a feasible domain of control strategies, and evaluates each control strategy in the feasible domain of control strategies;

[0031] S4. Based on the state dataset and the evaluation results of the feasible domain of the regulation strategy obtained in step S3, construct the integrated energy system operation and maintenance decision agent model based on the large language model.

[0032] Step S1 includes the following steps:

[0033] Step S11, the method for constructing the state dataset of the integrated energy system is as follows: by acquiring historical operating data of each load, equipment, and pipeline node in the integrated energy system, and representing it as a state dataset; the specific construction process is as follows: in time... The following will include historical operating data of the integrated energy system's equipment, load, and pipeline nodes, expressed as time. The state below Then, it is possible to construct a system containing... The state dataset at each time point is represented as follows: ;

[0034] Step S12, the method for constructing the domain knowledge base is as follows: collect the operation planning, safe operation, equipment control and operation characteristics of the integrated energy system and the various equipment pipelines, and establish a domain knowledge base based on this information for subsequent construction of a large language model of domain knowledge for the integrated energy system.

[0035] In step S2, the method for constructing the comprehensive energy system domain knowledge language model is as follows:

[0036] Construct large language model prompt words and use them as model input; the large language model prompt words specifically include the role setting of the integrated energy system, domain knowledge vector, state data at a specific time, and task objectives for generating and evaluating control strategies.

[0037] The control strategy and the evaluation result of the control strategy for the equipment under the corresponding state are used as the output of the model;

[0038] The method for constructing the domain knowledge vector is as follows: the domain knowledge base is divided into finer-grained word blocks, and these word blocks are vectorized and stored to obtain the domain knowledge vector.

[0039] Step S3 includes the following steps:

[0040] The prompts corresponding to different states of the integrated energy system are input into the integrated energy system domain knowledge large language model. The integrated energy system domain knowledge large language model outputs the feasible domain of control strategies through a thought chain. At the same time, the integrated energy system domain knowledge large language model evaluates all control strategies in the feasible domain of control strategies based on the operation and maintenance specifications and requirements of the integrated energy system and outputs the evaluation results. The feasible domain of control strategies includes the control strategies of all adjustable devices in the integrated energy system.

[0041] Step S4 specifically includes the following steps:

[0042] The states at different times are used as inputs to the integrated energy system operation and maintenance decision-making agent model based on the large language model. The optimal control strategy under the corresponding state is used as the output of the integrated energy system operation and maintenance decision-making agent model based on the large language model. The error between the evaluation result of the feasible region of the control strategy and the output result of the integrated energy system operation and maintenance decision-making agent model based on the large language model under the corresponding state is used as the loss function of the integrated energy system operation and maintenance decision-making agent model based on the large language model under the corresponding state. Finally, contrastive learning is used to train the integrated energy system operation and maintenance decision-making agent model based on the large language model on the generated feasible region of the control strategy.

[0043] Accordingly, the present invention also provides a comprehensive energy system operation and maintenance decision-making agent model based on a large language model constructed using the above method. When the comprehensive energy system operation and maintenance decision-making agent model based on a large language model is used in a real comprehensive energy system, the current state of the system is input into the comprehensive energy system operation and maintenance decision-making agent model based on a large language model, and the comprehensive energy system operation and maintenance decision-making agent model based on a large language model can output the optimal control strategy under the corresponding state.

Claims

1. A method for constructing an integrated energy system operation and maintenance decision-making agent model based on a large language model, characterized in that, Includes the following steps: S1, construct the state dataset and domain knowledge base of the integrated energy system; S2, Based on the domain knowledge base of the integrated energy system and the state data at different times, construct a large language model of the domain knowledge of the integrated energy system; S3, based on the different states of the integrated energy system, the domain knowledge big language model of the integrated energy system generates a feasible domain of control strategies, and evaluates each control strategy in the feasible domain of control strategies; The feasible domain of the control strategy includes the control strategies of all adjustable devices in the integrated energy system. S4. Based on the state dataset and the evaluation results of the feasible domain of the regulation strategy obtained in step S3, construct the integrated energy system operation and maintenance decision agent model based on the large language model. Step S1 includes the following steps: Step S11, the method for constructing the state dataset of the integrated energy system is as follows: by acquiring historical operating data of each load, equipment, and pipeline node in the integrated energy system, and representing it as a state dataset; the specific construction process is as follows: in time... The following will include historical operating data of the integrated energy system's equipment, load, and pipeline nodes, expressed as time. The state below Then, construct the containing The state dataset at each time point is represented as follows: ; Step S12, the method for constructing the domain knowledge base is as follows: collect the operation planning, safe operation, equipment control and operation characteristics of the integrated energy system and the various equipment pipelines, and establish a domain knowledge base based on this information.

2. The method for constructing an integrated energy system operation and maintenance decision-making agent model based on a large language model according to claim 1, characterized in that, In step S2, the method for constructing the comprehensive energy system domain knowledge language model is as follows: Construct large language model prompt words and use them as model input; the large language model prompt words specifically include the role setting of the integrated energy system, domain knowledge vector, state data at a specific time, and task objectives for generating and evaluating control strategies. The control strategy and the evaluation result of the control strategy for the equipment under the corresponding state are used as the output of the model; The method for constructing the domain knowledge vector is as follows: the domain knowledge base is divided into finer-grained word blocks, and these word blocks are vectorized and stored to obtain the domain knowledge vector.

3. The method for constructing an integrated energy system operation and maintenance decision-making agent model based on a large language model according to claim 2, characterized in that, Step S3 specifically involves: The prompt words corresponding to different states of the integrated energy system are input into the integrated energy system domain knowledge large language model. The integrated energy system domain knowledge large language model outputs the feasible domain of control strategies through the thinking chain method. At the same time, the integrated energy system domain knowledge large language model evaluates all control strategies in the feasible domain of control strategies based on the operation and maintenance specifications and requirements of the integrated energy system and outputs the evaluation results.

4. The method for constructing an integrated energy system operation and maintenance decision-making agent model based on a large language model according to claim 1, characterized in that, Step S4 specifically involves: The states at different times are used as inputs to the integrated energy system operation and maintenance decision-making agent model based on the large language model. The optimal control strategy under the corresponding state is used as the output of the integrated energy system operation and maintenance decision-making agent model based on the large language model. The error between the evaluation result of the feasible region of the control strategy and the output result of the integrated energy system operation and maintenance decision-making agent model based on the large language model under the corresponding state is used as the loss function of the integrated energy system operation and maintenance decision-making agent model based on the large language model under the corresponding state. Finally, contrastive learning is used to train the integrated energy system operation and maintenance decision-making agent model based on the large language model on the generated feasible region of the control strategy.

5. A comprehensive energy system operation and maintenance decision-making agent model based on a large language model constructed according to any one of claims 1-4, characterized in that, When the integrated energy system operation and maintenance decision-making agent model based on the large language model is used in a real integrated energy system, the current state of the system is input into the integrated energy system operation and maintenance decision-making agent model based on the large language model, and the integrated energy system operation and maintenance decision-making agent model based on the large language model can output the optimal control strategy under the corresponding state.

Images