A distribution network device group cooperative energy-saving optimization method and device based on federated learning

By employing federated learning to perform data consistency verification and topology alignment regularization on distribution network device groups, and combining this with model reference data for knowledge distillation modeling, feeder-level collaborative decision-making quantities are generated. This solves the problem of insignificant energy-saving effects of traditional distribution network device groups, and achieves the effect of reducing resource consumption and operating costs without data leakage.

CN121056328BActive Publication Date: 2026-07-21NANJING GUANGMAOLONG TECH CO LTD
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
NANJING GUANGMAOLONG TECH CO LTD
Filing Date
2025-08-25
Publication Date
2026-07-21

AI Technical Summary

Technical Problem

Traditional energy-saving methods for distribution network equipment groups are not very effective without leaking raw data, and cannot effectively reduce resource consumption and operating costs.

Method used

By adopting a federated learning-based approach, the federated upload data of the edge control agent cluster is obtained, and consistency verification and topology alignment regularization are performed. Combined with model reference data, knowledge distillation modeling is carried out to generate feeder-level collaborative decision quantities, forming a unified distribution packet to achieve collaborative energy-saving optimization.

Benefits of technology

Without leaking raw data, it significantly reduces resource consumption and operating costs, improves the efficiency, reliability and sustainability of network-wide collaborative control, achieves peak shaving and valley filling, reduces losses and saves energy, and reduces overload and overlimit events.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121056328B_ABST
    Figure CN121056328B_ABST
Patent Text Reader

Abstract

The application relates to a power distribution network device group cooperative energy-saving optimization method and device based on federal learning, which comprises the following steps: obtaining federal upload data of an edge control agent cluster, loading server running environment data and model reference data, performing consistency verification on the federal data to obtain qualified incremental packages and qualified topological fingerprints, robustly aggregating shared base parameters by using the qualified incremental packages to generate shared base temporary parameters, performing topological alignment regularization on the qualified topological fingerprints to obtain topological adaptive shared base parameters, distilling preset individualized strategy parameters according to the model reference data to obtain a distilled edge execution strategy parameter set, carrying out feeder level cooperative optimization according to the distilled parameters and a running environment to form a strategy issuing instruction set, and fusing the topological adaptive shared base parameters and the instruction set to generate a power distribution network device group cooperative energy-saving issuing package for the edge control agent cluster, so that the original data is not disclosed, and resource consumption and operation cost can be significantly reduced.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Multi-modal data federated learning-based farmer credit assessment method and system

    CN120258967A

  • Electric power analysis method and system based on artificial intelligence

    CN120494546A