A Data‐Driven Acceleration Method for Production Simulation of Power Systems with High‐Penetration Renewables

ABSTRACT The 8760‐h production simulation is a crucial component in power system planning and operation. With the increasing integration of high‐penetration renewable energy, even simplified approaches encounter computational efficiency challenges, primarily due to the surge in integer variables, continuous variables, and operational constraints. To address this issue, this paper proposes an accelerated production simulation method that leverages a data‐driven model to infer the solution of the relaxed clustered unit commitment (RCUC), instead of explicitly solving the mixed‐integer linear programming (MILP) formulation. Specifically, a multi‐channel convolutional neural networks‐transformer (MCCNNs‐Transformer) architecture is developed to obtain the hourly thermal cluster outputs and daily operation costs. It captures multi‐scale temporal features while utilizing a self‐attention mechanism to effectively represent the physical coupling characteristics in power systems. Furthermore, a stratified normalization strategy is introduced to differentially process multi‐type inputs, preserving the inherent features of sequences and mitigating distortion caused by uniform normalization. Case studies conducted on a provincial power grid in China demonstrate that the proposed method significantly enhances computational efficiency while maintaining high accuracy, validating the feasibility of data‐driven method in accelerating RCUC‐based production simulation.