CN-121981440-A - Comprehensive energy system optimization regulation and control method based on working condition clustering and multi-agent model

CN121981440ACN 121981440 ACN121981440 ACN 121981440ACN-121981440-A

Abstract

The invention relates to a comprehensive energy system optimization regulation and control method based on working condition clustering and a multi-agent model, which is applicable to the field of comprehensive energy system operation regulation and control. The method comprises the steps of determining a system external working condition parameter set related to a preset optimization objective function based on a structure of a target comprehensive energy system, constructing an external working condition data set of the system, classifying all working condition samples in the data set to form an external working condition data subset corresponding to all working condition classifications, determining an optimal regulation strategy result corresponding to all the samples, constructing a training data set corresponding to all the working condition classifications, taking the working condition samples in the training data set as input, taking the optimal regulation strategy result corresponding to all the working condition samples as output, training an optimal regulation proxy model corresponding to the comprehensive energy system, acquiring the current system external working condition, determining the working condition classification corresponding to the current working condition, and inputting the current working condition into the proxy model corresponding to the working condition classification to obtain the optimal regulation strategy result corresponding to the current working condition.

Inventors

WANG ZIHAO
HUANG YANZHONG
FANG HAO
XU JUNYANG
LI DEDI
WU YUECHAO
LUO YUANLIN
ZHENG LI

Assignees

中国电建集团华东勘测设计研究院有限公司
杭州华辰电力控制工程有限公司

Dates

Publication Date: 20260505
Application Date: 20251226

Claims (10)

1. The comprehensive energy system optimization regulation and control method based on the working condition clustering and the multi-agent model is characterized by comprising the following steps of: Determining a system external working condition parameter set associated with a preset optimization objective function based on the structure of the target comprehensive energy system; constructing an external working condition data set of the system based on the external working condition parameter set of the system and the historical operation working condition data of the target comprehensive energy system; classifying the working condition samples in the data set based on the parameter data of the working condition samples in the external working condition data set to form an external working condition data subset corresponding to the working condition classification; based on working condition samples in the data subset, combining a preset optimization objective function, and determining an optimal regulation strategy result corresponding to each sample; Constructing a training data set corresponding to each working condition classification based on the working condition samples in the data subsets and the optimal regulation strategy results corresponding to the working condition samples; Taking a working condition sample in the training data set as input, taking an optimal regulation strategy result corresponding to the working condition sample as output, and training an optimal regulation proxy model corresponding to each working condition classification; the method comprises the steps of obtaining the current external working condition of the system, determining the working condition classification corresponding to the current working condition, inputting the current working condition into the agent model corresponding to the working condition classification, and obtaining the optimal regulation strategy result corresponding to the current working condition.
2. The method for optimizing and controlling the comprehensive energy system based on the working condition clusters and the multi-agent model according to claim 1, wherein the set of the external working condition parameters of the system is collected to an external working condition parameter library, and the external working condition parameter library comprises end load parameters, energy-related meteorological parameters and energy-related price parameters.
3. The method for optimizing and controlling a comprehensive energy system based on condition clustering and a multi-agent model according to claim 1, wherein the constructing an external condition data set of the system based on the system external condition parameter set and the historical operation condition data of the target comprehensive energy system comprises: Based on each parameter in the parameter set, each parameter data at the same moment is obtained from the historical operation condition data, and a condition sample in the external condition data set is formed.
4. The method for optimizing and controlling a comprehensive energy system based on a working condition cluster and a multi-agent model according to claim 1, wherein the classifying the working condition samples in the data set based on each parameter data of the working condition samples in the external working condition data set to form an external working condition data subset corresponding to each working condition classification comprises: Based on all parameter data of the working condition samples in the external working condition data set, clustering the working condition samples in the external working condition data set by adopting a K-means algorithm to obtain K different external working condition data subsets.
5. The method for optimizing and controlling a comprehensive energy system based on condition clustering and multiple agent models according to claim 1, wherein the training the comprehensive energy system optimizing and controlling agent model corresponding to each condition classification by taking the condition sample in the training data set as input and the optimal control strategy result corresponding to the condition sample as output comprises the following steps: And learning the training data set corresponding to each working condition classification by adopting a machine learning algorithm to obtain a proxy model corresponding to each working condition classification.
6. The method for optimizing and controlling the comprehensive energy system based on the working condition clustering and the multi-agent model according to claim 1, wherein the steps of obtaining the current external working condition of the system and determining the working condition classification corresponding to the current working condition comprise the following steps: Based on the external working condition parameter data of the current working condition, calculating the distance between the working condition classification clustering centers and the working condition classification with the minimum distance is used as the working condition classification of the current working condition.
7. The method for optimizing and regulating the comprehensive energy system based on the working condition clustering and the multi-agent model according to claim 1, wherein the preset optimizing objective function is to minimize the system running cost, maximize the end load supply rate or minimize the carbon dioxide emission of the system.
8. An integrated energy system optimizing and regulating device based on working condition clustering and a multi-agent model is characterized by comprising: the parameter determining module is used for determining a system external working condition parameter set associated with a preset optimization objective function based on the structure of the objective comprehensive energy system; The data set construction module is used for constructing an external working condition data set of the system based on the system external working condition parameter set and the historical operation working condition data of the target comprehensive energy system; The data classification module is used for classifying the working condition samples in the data set based on the parameter data of the working condition samples in the external working condition data set to form an external working condition data subset corresponding to the working condition classification; The sample optimizing module is used for determining an optimal regulation strategy result corresponding to each sample based on working condition samples in the data subset and combining with a preset optimization objective function; the training set construction module is used for constructing training data sets corresponding to all working condition classifications based on the working condition samples in the data subsets and the optimal regulation strategy results corresponding to the working condition samples; the model training module is used for taking the working condition samples in the training data set as input, taking the optimal regulation strategy results corresponding to the working condition samples as output, and training the comprehensive energy system optimization regulation proxy model corresponding to each working condition classification; The model optimizing module is used for acquiring the current external working condition of the system, determining the working condition classification corresponding to the current working condition, inputting the current working condition into the agent model corresponding to the working condition classification, and obtaining the optimal regulation strategy result corresponding to the current working condition.
9. A storage medium having stored thereon a computer program executable by a processor, wherein the computer program when executed implements the steps of the integrated energy system optimization regulation method based on a condition cluster and a multi-agent model according to any one of claims 1 to 7.
10. An integrated energy system optimizing and regulating device, comprising a memory and a processor, wherein the memory stores a computer program which can be executed by the processor, and the method is characterized in that the computer program is executed to realize the steps of the integrated energy system optimizing and regulating method based on the working condition clustering and the multi-agent model according to any one of claims 1-7.

Description

Comprehensive energy system optimization regulation and control method based on working condition clustering and multi-agent model Technical Field The invention relates to a comprehensive energy system optimization regulation method based on working condition clustering and a multi-agent model. The method is suitable for the field of comprehensive energy system operation regulation and control. Background Renewable energy systems such as wind power and photovoltaic can effectively relieve energy crisis and reduce carbon emission, but high-proportion penetration of the renewable energy systems in a traditional power grid can cause stability and reliability of the power system to be reduced, and system operation is affected. The comprehensive energy system can integrate various renewable energy forms in the area efficiently through coordination optimization and efficient complementation among the multipotency flows, and develop optimization regulation research of the comprehensive energy system, and has important significance and value for promoting the consumption of renewable energy and ensuring the reliable supply of terminal multielement load. At present, most of the optimal regulation methods in the comprehensive energy field search the optimal operation scheme by adopting an algorithm optimizing mode, and as the comprehensive energy system has the characteristics of various equipment types, complex topological structure, strong thermal coupling and the like, the calculation cost and the time cost of the method are very high, and the optimal regulation requirement of the actual comprehensive energy system is difficult to meet. The agent model has the potential to solve the bottleneck problem of low calculation efficiency of the existing method, can directly learn the mapping relation between different operation conditions and the corresponding optimal operation scheme through massive training data, and has extremely high calculation efficiency. In the prior art, a plurality of cases for optimizing and controlling a comprehensive energy system by using a proxy model exist, for example, an active learning proxy optimization method (China patent application, publication No. CN 117744894A) of the comprehensive energy system adopts a cluster extreme learning machine to construct a proxy model for learning the relations among different operation schemes, energy saving rates, operation costs and carbon dioxide emission, and a comprehensive energy system operation and maintenance decision proxy model construction method (China patent application, publication No. CN 117973537A) based on a large language model selects the large language model as the proxy model for predicting the optimal operation scheme under different working conditions. In the prior art, the agent model is used for optimizing and controlling various operation conditions, but the actual comprehensive energy system has a wide operation condition range, the prediction accuracy of the actual comprehensive energy system under various operation conditions is difficult to ensure by the existing method, and the economy and the high efficiency of the system operation are obviously affected. Disclosure of Invention Aiming at the problems, the invention provides a comprehensive energy system optimization regulation method based on working condition clustering and a multi-agent model. The technical scheme adopted by the invention is that the comprehensive energy system optimizing and regulating method based on the working condition clustering and the multi-agent model comprises the following steps: Determining a system external working condition parameter set associated with a preset optimization objective function based on the structure of the target comprehensive energy system; constructing an external working condition data set of the system based on the external working condition parameter set of the system and the historical operation working condition data of the target comprehensive energy system; classifying the working condition samples in the data set based on the parameter data of the working condition samples in the external working condition data set to form an external working condition data subset corresponding to the working condition classification; based on working condition samples in the data subset, combining a preset optimization objective function, and determining an optimal regulation strategy result corresponding to each sample; Constructing a training data set corresponding to each working condition classification based on the working condition samples in the data subsets and the optimal regulation strategy results corresponding to the working condition samples; Taking a working condition sample in the training data set as input, taking an optimal regulation strategy result corresponding to the working condition sample as output, and training an optimal regulation proxy model corresponding to each working condition classification;