全球能源互联网2025，Vol.8Issue(1)：76-86,11.DOI:10.19705/j.cnki.issn2096-5125.2025.01.009
基于Stackelberg博弈与改进深度神经网络的多源调频协调策略研究
Research on Multi-source Frequency Regulation Strategies Based on the Stackelberg Game and Improved Deep Neural Network
摘要
Abstract
With the increase of new energy penetration in the power grid,the traditional frequency regulation of thermal power units can no longer meet the power quality demand.Aiming at the problem of large area control error in traditional automatic generation control systems in multi-source scenario,a multi-source frequency regulation strategy based on the Stackelberg game and improved deep neural network(S-DNN)is proposed.Firstly,an improved multilevel deep neural network(DNN)is proposed,which consists of a DNN layer,natural gradient boosting layer,and least squares support vector machine layer to sequentially and progressively complete the prediction,evaluation,and execution of actions,and output the total frequency regulation power command.This multilevel total frequency regulation power output model considers the dynamic impact of new energy penetration on the frequency regulation system,fully learns more features from historical information and real-time state,and improves the accuracy of frequency regulation instructions.Then,based on Stackelberg game theory,it considers the characteristics and synergy of multi-source frequency regulation,optimizes the power allocation among frequency regulation sources,and improves the economy of the system's secondary frequency regulation.Finally,the effectiveness of the proposed multi-source frequency regulation strategy is verified by case analysis.Compared with the traditional frequency regulation method,the proposed S-DNN multi-source frequency regulation strategy can effectively reduce the area control error and frequency deviation,and reduce the frequency regulation cost.关键词
多源系统/二次调频/Stackelberg博弈/深度神经网络/自然梯度提升/最小二乘支持向量机Key words
multi-source system/secondary frequency regulation/Stackelberg game/deep neural networks/natural gradient boosting/least squares support vector machine分类
动力与电气工程引用本文复制引用
王永文,赵雪锋,李夏叶,詹巍,单怡琳,闫启明,赵泽宇,杨锡运..基于Stackelberg博弈与改进深度神经网络的多源调频协调策略研究[J].全球能源互联网,2025,8(1):76-86,11.基金项目
国家电投集团四川电力有限公司科技项目(XNNY-WW-KJ-2021-16).Science and Technology Project of State Power Investment Group Sichuan Electric Power Co.,Ltd.(XNNY-WW-KJ-2021-16). （XNNY-WW-KJ-2021-16）
