Abstract：With the continuous development of renewable energy sources and the increasing share of

renewable energy in the grid，optimal coordination of maintenance work becomes increasingly important in order to ensure the safety of power supply in power systems considering renewable energy access. Current tools for

maintenance planning are constrained by operational safety standards and the complexity of the grid，and have

problems such as low operability and high computational effort to simulate accidents. To reduce the burden of

manual computation，the use of machine learning models was proposed to predict the outcome of emergency

situations in a fast and reliable manner. The method was tested in a regional facility in Lanzhou，covering voltage

levels of 10 kV and 220 kV. By testing and comparing a plain Bayesian classifier，a support vector machine（SVM）

and a decision tree-based model，it was shown that the decision tree-based random forest algorithm is consistently

better than other algorithms in identifying safe serviceable time periods with an accuracy rate higher than 90%. In

addition，it was shown experimentally that the expected growth in renewable energy generation will affect the

future serviceability of the power system，with a 20% increase in non-safe serviceable time periods in some areas.

      Key words：renewable energy；machine learning；maintenance plan；reliability；tree-based model