Automation and Safety Control of New Energy Power Dispatching Based on Multi-Agent Network System

Abstract

During the power dispatching process, the increased access of new energy sources may cause frequency and voltage fluctuations and affect system stability. Therefore, this article adopts the new energy power dispatching automation and safety control method based on a multi-agent network system, models the power generation, energy storage, and load units as autonomous agents, and combines the long-short term memory (LSTM) network to predict the new energy power and improve the dispatching precision. At the same time, the distributed optimal power flow (DOPF) algorithm is used to realize distributed power allocation and optimization among multiple agents. Collaboration capabilities are enhanced through policy self-learning. A wide area measurement system (WAMS) monitors frequency and voltage in real time. Deep reinforcement learning (DRL) is used to design feedback control strategies to dynamically respond to new energy fluctuations and improve system stability. A simulation comparison experiment is conducted with the traditional centralized dispatching system. The multi-agent network system can reduce frequency and voltage deviations. The average dispatching accuracy is about 99.24%. At the same time, the power grid’s frequency fluctuation and voltage deviation are reduced, and the system’s stability is enhanced. In addition, the system shortens fault detection and recovery time, optimizes the utilization of new energy, and improves the efficiency of resource allocation. The results indicate that the system in this article performs well in improving the automation, stability, and efficient utilization of new energy in the power system, verifying its effectiveness as an innovative solution for power dispatching.