基于改进深度强化学习理论的台区微电网电压控制策略OA

The inherent characteristics of distributed energy resources(DERs),including their small scale,volatility,and intermittency,pose significant challenges to the design of operational strategies for distribution area microgrids.Despite the successful integration of diverse DERs and external power grids within microgrid systems,voltage management has become increasingly complex.In light of this,a real-time voltage control strategy for distribution area microgrids was proposed based on deep reinforcement learning theory.Firstly,a recurrent neural network(RNN)model was employed to accurately identify and handle corrupted or missing data in the source-load power data within the system,ensuring data quality.Subsequently,a voltage management model for the distribution area microgrid was constructed,comprehensively considering line losses during power transmission and the potential risk of voltage violations.Given the complex nonlinear constraints inherent in this model,an improved deep reinforcement learning algorithm was adopted for efficient solution,and an ε-greedy decreasing strategy was adopted to guide the agent's action selection,overcoming the limitations of traditional methods.Finally,to validate the effectiveness and feasibility of the proposed strategy,comparative tests were conducted against traditional control strategies.The results show that the voltage control strategy presented exhibits significant advantages in multiple key indicators,including reducing voltage fluctuations and minimizing network losses.