A Novel Acceleration Technique for Power System Production Simulation Based on an Improved K-Means Algorithm

Traditional production simulation models face significant challenges in solving optimization problems in real power grids due to high computational complexity, low efficiency, and insufficient accuracy, which pose difficulties for power system planning and operation. This paper focuses on reduction techniques for power production simulation models. In terms of scenario reduction, a temporal dimension reduction approach is employed to identify similar hours, days, and weeks within the 8,760 hours of a year. Multiple factors, including load, photovoltaic output, and wind power output, are comprehensively considered. An improved Euclidean distance calculation is incorporated to effectively reduce the original scenarios, significantly eliminating redundant scenarios and thereby accelerating production simulation. A production cost simulation based on real system data from a province demonstrates that the runtime after scenario reduction is only 0.002 times that of the original simulation. The results verify a substantial reduction in simulation time with minimal error, fully confirming the effectiveness of the proposed method.