Integrated Energy System Planning Method Based on Deep Graph Embedding and Distributed Resilience Consensus
Year: 2026
Type: conference-paper
Abstract: To address the challenges in integrated energy system planning, including the heterogeneous coexistence of physical constraints of devices and game-theoretic constraints among multiple participants, the difficulty of distributed deployment of resilience regulation schemes under extreme uncertainties, and the high demand for data privacy protection of participants, this paper proposes a planning method integrating deep graph embedding and distributed resilience consensus. Firstly, a dual-layer heterogeneous graph model is established: the bottom physical layer models the multi-energy flow networks of electricity, heat, and hydrogen along with device operating constraints, while the upper game layer models the payment relationships and carbon quota interaction rules among energy participants. The Structural Deep Network Embedding (SDNE) method is employed to nonlinearly compress the high-dimensional constraint space into low-dimensional resilience feature vectors, retaining two key indices: controllability and coupling degree. Secondly, a dynamic leader rotation consensus protocol driven by resilience contribution is proposed, in which each energy participant exchanges only low-dimensional resilience representation vectors and marginal benefit signs. By replacing hash commitment with decision variable similarity verification, distributed resilience negotiation with linear communication complexity is achieved. Finally, an adaptive planning model based on hierarchical Information Gap Decision Theory (IGDT) and dynamic resilience radius is proposed. The resilience contributions obtained from distributed consensus are utilized to allocate uncertainty tolerance radii to different energy participants in a differentiated manner, and multi-scenario heatmaps are generated through conditional fragility sensitivity analysis, realizing closed-loop coupling between resilience planning and distributed optimization results. A northern China park-level integrated energy system is used as a case study for validation. The results show that, compared with centralized planning, the proposed method increases the total system cost by only 2.8% while ensuring privacy protection of each energy participant, but improves the load guarantee rate from 86.3% to 91.2% under extreme load peaks and renewable energy disconnection scenarios. Compared with fully decentralized independent planning, the total cost is reduced by 16.2%, and the load guarantee rate is increased by 42.3%. The method has low computational complexity, can explicitly locate key fragile links of the system, and effectively breaks the “privacy-efficiency-resilience” trilemma in integrated energy system planning, providing a feasible path and technical support for practical engineering deployment.
Authors: Qing Xu, Lingqiu Chen, Tingfang Tan, Yingqi Li
Language: English
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7204124641