Joint dispatch and economic collaboration of multiple regional energy systems via Transformer-based load prediction and two-stage stochastic optimization

Enhancing resource complementarity and operational economy across multiple regional energy systems requires effective collaboration and optimized dispatch. Current approaches suffer from inaccurate demand prediction and inadequate handling of stochasticity and fairness. This study addresses the challenges by proposing a synergistic framework that integrates Transformer-based load prediction with two-stage stochastic optimization. The variational mode decomposition is used to extract multi-scale time-series features embedded in the load, and each mode is trained and predicted independently using the Transformer model to mitigate the errors and improve accuracy. Uncertainty scenarios are generated and fed into a stochastic optimization model that jointly determines electricity trading prices and regional dispatch strategies. The superiority of the proposed method has been verified. The case study results show that the proposed joint dispatch can reduce total operating costs and carbon emissions by 10.69 % and 10.11 %, respectively, while internal interaction contributes to 19.14 % of the total electricity demand. The fairness constraint can effectively balance the regional benefit distribution, and the cost reduction rates of different systems converge to approximately 10 % while maintaining cooperation incentives. The analysis further considers the impact of time-of-use tariffs, electricity purchase constraints, and energy storage degradation. This research provides a practical and promising approach to advancing regional energy sharing for low-carbon and profitable operation.