Open Access

ARTICLE

Heterogeneous Graph Neural Network–Assisted Joint Clearing for Intra- and Inter-Provincial Electricity Spot Markets

Shijie Ji¹, Shuya Lei^2,*, Lan Ren², Renjie Wei², Ning Yang³, Yishen Wang²
1 Beijing Power Exchange Center, Beijing, 100031, China
2 China Electric Power Research Institute, Beijing, 100192, China
3 Beijing Kedong Electric Power Control System Co., Ltd., Beijing, 211000, China
* Corresponding Author: Shuya Lei. Email:
(This article belongs to the Special Issue: Grid Integration of Intermittent Renewable Energy Resources: Technologies, Policies, and Operational Strategies)

Energy Engineering https://doi.org/10.32604/ee.2026.075094

Received 24 October 2025; Accepted 30 December 2025; Published online 26 August 2026

Abstract

With the continuous growth of the scale of inter-provincial electricity transactions, the intra-provincial independent clearing model has gradually revealed limitations in terms of resource collaborative optimization and regional supply-demand balance, and the coordination mechanism between the two-tier markets of inter-provincial and intra-provincial levels urgently needs to be improved. To address this problem, this paper constructs a bi-level joint clearing model for the intra-provincial and inter-provincial electricity spot markets, which includes three stages: intra-provincial clearing, inter-provincial entity screening and inter-provincial joint clearing. The model aims to effectively connect the results of intra-provincial transactions with inter-provincial resource allocation through a hierarchical coordination mechanism, so as to realize the orderly linkage and global optimization of the two-tier markets. To improve the solution efficiency of large-scale mixed integer linear programming (MILP), this paper proposes an auxiliary solution method based on heterogeneous graph neural networks (HGNN). Through the heterogeneous graph modeling with multi-type nodes and multi-relation structures, the method realizes the expression of coupling information among the generation side, demand side, provinces and tie-lines, and performs pre-identification of integer variables to generate high-quality initial solutions, thereby improving the solution efficiency of MILP. Case study results show that, compared with traditional methods, the proposed method significantly reduces the solution time while maintaining the optimality of clearing results, effectively promotes the efficient allocation of inter-provincial electricity resources, and exhibits good real-time performance and scalability in a large-scale market environment.

---

Keywords

Electricity spot market; joint clearing; heterogeneous graph neural network; mixed-integer programming

---