A power system scheduling method based on scene mapping and a storage medium

CN121836310BActive Publication Date: 2026-06-09HUNAN UNIV
View PDF 4 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
HUNAN UNIV
Filing Date
2026-03-13
Publication Date
2026-06-09

AI Technical Summary

Technical Problem

Reinforcement learning agents suffer from poor generalization of scheduling decisions and insufficient security in extreme scenarios due to the offset between training and testing scenarios.

Method used

A scene mapping model based on generative adversarial networks is constructed. Through adversarial training, extreme scenes are mapped to the distribution space of normal scenes. The decision-making actions are corrected for safety by combining a physical model. The parameters are updated by alternating between the generator and the discriminator to generate a mapped scene that conforms to the adjustable capacity range of the power system.

Benefits of technology

It significantly improves the security, economy, and reliability of scheduling schemes in extreme scenarios, ensuring the effective operation and decision-making efficiency of reinforcement learning agents in extreme scenarios.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121836310B_ABST
    Figure CN121836310B_ABST
Patent Text Reader

Abstract

The application belongs to the technical field of power system automatic dispatching, in particular to a power system dispatching method based on scene mapping and a storage medium, comprising: firstly, a scene mapping model based on a generative adversarial network is constructed to map extreme scenes which are difficult to handle to regular scenes which are the same as training data; then, a reinforcement learning intelligent agent based on the regular scenes is used to make a quick decision for the mapped scenes; finally, a motion correction model considering safety constraints is designed to physically correct the initial motion of the intelligent agent decision to ensure that it meets the actual dispatching requirements under extreme scenes. Through the three-stage framework of "scene mapping-intelligent decision-motion correction", the application effectively solves the problem of decline in generalization performance caused by data distribution deviation when reinforcement learning deals with extreme scenes, and improves the dispatching decision-making ability of the power system under high-risk and low-probability extreme events and the safe and stable operation level of the power grid.
Need to check novelty before this filing date? Find Prior Art