A VSG parameter self-adaptive control method based on reinforcement learning and related device

By constructing an intelligent agent based on reinforcement learning, the inertia and damping coefficients of a virtual synchronous generator are optimized in real time. This addresses the shortcomings of VSG parameter optimization methods, improves the stability and adaptability of the power grid, and enables the safe and reliable operation of the power grid.

CN122225530APending Publication Date: 2026-06-16ZHUHAI POWER SUPPLY BUREAU GUANGDONG POWER GIRD CO +1
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
ZHUHAI POWER SUPPLY BUREAU GUANGDONG POWER GIRD CO
Filing Date
2026-03-11
Publication Date
2026-06-16

AI Technical Summary

Technical Problem

In the existing technology, there is a lack of dedicated optimization methods for the inertia coefficient and damping coefficient of virtual synchronous generators (VSGs). Traditional optimization algorithms cannot meet the requirements of complex and ever-changing power grid environments in terms of real-time performance and computational complexity, resulting in insufficient grid frequency and voltage stability and an inability to adapt to the dynamic changes of high proportion of new energy grid connection.

Method used

A reinforcement learning-based agent is constructed by collecting power grid operation status data, constructing state and action spaces, defining a reward function, and training the reinforcement learning agent to output the adjustment amounts of inertia and damping coefficients, thereby achieving adaptive optimization of VSG parameters. A deep deterministic policy gradient algorithm is used to iteratively update network parameters, forming real-time feedback and closed-loop control.

Benefits of technology

It achieves real-time adaptive optimization of VSG parameters, improves the stability of power grid frequency and voltage, enhances the safety and reliability of the power grid, adapts to complex and ever-changing power grid operating conditions, and avoids the problem of parameter adjustment lag.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122225530A_ABST
    Figure CN122225530A_ABST
Patent Text Reader

Abstract

The application provides a VSG parameter self-adaptive control method based on reinforcement learning and a related device, and the steps of the method include collecting power grid operation state observation data; constructing a state space of a reinforcement learning agent based on the power grid operation state observation data, and simultaneously defining an action space and a reward function for VSG parameter adjustment; training the reinforcement learning agent, so that the reinforcement learning agent outputs an adjustment amount of an inertia coefficient and a damping coefficient in the action space based on the input of the state space and in combination with real-time feedback of the reward function; and inputting the adjustment amount into a VSG controller to dynamically update VSG operation parameters. The application constructs a state space, an action space and a reward function of a reinforcement learning agent based on power grid operation state observation data, and completes training, and dynamically outputs an adjustment amount of an inertia coefficient and a damping coefficient from the trained agent, so that adaptive optimization of VSG parameters is realized, and the safety and reliability of power grid operation can be significantly improved.
Need to check novelty before this filing date? Find Prior Art