Data-driven Acceleration Method for Discrete Interprovincial Medium and Long-term Power Transactions Clearing Problem

The optimization of the interprovincial medium and long-term power transactions clearing problem aims to maximize social welfare through the selection of optimal purchasing and selling pairs. Considering the extensive range of time periods, buyers, sellers, transmission paths, and the discrete nature of selling in blocks, it is intractable to obtain the optimal solution within a practical computing time limit, even using the state-of-the-art discrete optimizer. To address this challenge, this paper proposes a data-driven approach to accelerate the solution process by fully exploring few-shot historical data. First, the stable structure of the solution pattern is identified and fixed in advance to reduce the model scale. Second, multiple warm-start solutions are selected using the bidding price feature. Finally, the solution process is accelerated to achieve an improved decision. Based on practical data in China, the case study demonstrates that the proposed method can improve the solution quality by 1 million RMB on average within the solution time limit of 1 hour, compared with directly solving the clearing problem using the state-of-the-art commercial solver.