In order to solve the problem of heavy workload in scoring power operation steps, this paper studies a scoring method for power operation steps based on a local general large language model, avoiding the disadvantage that traditional analysis methods cannot accurately segment power operation steps, and at the same time improving the accuracy of scoring. This paper uses local large language models such as ChatGLM3 to segment the operation steps, which has better accuracy than traditional segmentation methods. At the same time, a complete set of power operation step scoring methods is established to avoid the problem that traditional scoring methods cannot accurately score. By constructing step sequence vectors and updating step sequence vectors, it can accurately score the student steps compared to traditional scoring methods. Finally, experiments verified that the method proposed in this paper can improve the scoring accuracy of power operation steps.