基于纵向联邦强化学习的居民社区综合能源系统协同训练与优化管理方法

The Collaborative Training and Management-optimized Method for Residential Integrated Energy System Based on Vertical Federated Reinforcement Learning

电、热、气等能源系统的实时管理能力关系到居民社区用能需求能否得到充分满足。该文首先基于深度Q网络(deep Q-learning network,DQN)强化学习算法建立了一种综合考虑居民社区实时负荷大小、能源销售与采购价格的能源系统优化管理模型。其次,该文面向不同能源系统之间的数据壁垒现象,提出一种基于纵向联邦学习技术的居民社区综合能源系统协同训练方法。该方法能够在3种能源系统DQN优化管理模型的训练过程中,通过交互DQN优化管理模型的网络梯度等参数信息来提升模型的训练速度。最后,算例分析验证该文所构建DQN优化管理模型的有效性,同时验证所提出居民社区综合能源系统协同训练方法能够优化各能源系统的经济效益及模型训练效率。

The ability of real-time management for electricity,heat,and gas energy system is closely related to whether the energy demand of residential community could not be met.This paper first created an energy management-optimized model which comprehensively considering the real-time energy demand,energy purchasing and selling price of energy system based on the Deep Q-learning network algorithm.Besides,focused on the data barriers phenomenon between the different energy systems,it provided a cooperative training method for residential community’s integrated energy system based on the vertical federated learning technology.During the training process,this method could improve the training efficiency of DQN management-optimized model by interacting its network parameters,likes gradient.Finally,numerical simulation verified the effectiveness of designed DQN management-optimized model,and it verified that the proposed cooperative training method for residential community’s integrated energy systems could achieve a better economic benefit and training efficiency for each energy system.