Design of a Model for Building Excellent Management Capabilities for Multi-Agent Ecological Collaboration in Power Grids

View at publisher Summarise with VIRAT
Authors: Zhiguang Hu, Hao Zuo, Jie Yang, Weide Gu, Zhaoyang Zhang
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Addressing the issues of low collaborative efficiency, data privacy leaks, and weak dynamic adaptability caused by the surge in multiple stakeholders in power grids under the "dual carbon" objective, this paper constructs a three-layer multi-agent ecological collaborative management capability model consisting of "core control - collaborative decision-making - execution feedback," and designs a Federated Attention Deep Reinforcement Learning (FA-DRL) algorithm. This algorithm integrates the gradient encryption transmission of federated learning, the high-value stakeholder focus of the attention mechanism, and the dynamic decision-making capability of reinforcement learning. It achieves a balance between privacy protection and collaborative optimization through a "local training - encrypted aggregation - weight optimization" process. Experiments are conducted based on actual data (1.2TB, 5-minute granularity) from a provincial power grid in 2023, comparing the results with traditional DRL and FL+DRL algorithms. The results show that the collaborative decision-making time of FA-DRL is 1.23 seconds, which is 65.4% lower than that of traditional DRL; the comprehensive energy consumption of the power grid is 4.82×105kWh/day, which is 8.7% lower than that of traditional DRL; the renewable energy absorption rate reaches 92.3%, which is 4.1% higher than that of FL+DRL; the privacy risk value is 0.03, and the decision-making accuracy is still 92.7% under ±25% output fluctuation. This study demonstrates that the model and algorithm can effectively improve the collaborative management capability of the power grid and provide technical support for the operation of new power systems.

Fields of Study & Subjects

Smart Grid and Power Systems Smart Grid Energy Management Integrated Energy Systems Optimization

Publication details

Journal / source
IEEE International Conference on Power, Electronics and Green Energy (ICPEGE)
Pages
1461–1466
Publication date
28 January 2026
Language
EN