Multi-path-constrained inter-provincial medium- and long-term transaction mechanism and parallel acceleration algorithm for its clearing model

• Designs a novel multi-path-constrained inter-provincial transaction mechanism for inter-provincial path-specific clearing. • Formulates a clearing model for the proposed mechanism, integrating transmission security constrains with temporal coupling. • Develops a variable-step DQAM algorithm enabling temporal scalable parallel computation for efficient clearing. The inter-provincial medium- and long-term transaction (IPMLT) mechanism plays a key role in China’s power market by facilitating electricity delivery from generation-rich provinces to high-demand consumption provinces. However, two key challenges hinder its efficient implementation: (1) under the “one line, one price” policy, buyers incur different transmission tariffs and losses depending on their transaction paths, necessitating explicit path matching and bid adjustments in both quantity and price; and (2) the shift toward time-segmented clearing, which better aligns with spot market operations, imposes stricter physical realizability requirements that go beyond traditional available transfer capability (ATC) constraints. To address these challenges, we propose a Multi-Path-Constrained IPMLT (MPC-IPMLT) mechanism. It introduces path-specific centralized clearing by explicitly modeling buyer–seller-path relationships, accounting for transmission tariffs and losses, and embedding time-coupling constraints to reflect ramping limits across periods. A spatiotemporal clearing model is formulated to support this mechanism, capturing both geospatial transmission characteristics and inter-temporal operational limits. However, the resulting large-scale spatiotemporal coupled optimization problem poses significant computational challenges. To address this, we develop a variable-step diagonal quadratic approximation method (v-DQAM) that relaxes inter-temporal ramping constraints to enable scalable parallel computation. Numerical results based on real-world IPMLT data demonstrate that the proposed v-DQAM achieves a speedup of up to 7.24 times compared to direct solving, while maintaining a relative error below 0.01%, thereby ensuring both efficiency and accuracy in large-scale market clearing.