Photovoltaic Power Generation Forecasting Based on Random Forest Algorithm

Since the 14th Five-Year Plan, the strategy of “carbon neutrality and carbon peak” has become one of the important strategies to realize China's economic and environmental development, and the clean energy structure will be the main way to achieve the double carbon goal. Solar energy has become one of the important energy sources because of its large resource amount and wide distribution range, and photovoltaic power generation technology is an important means to control solar power generation. The accuracy of photovoltaic power generation prediction plays an important role in the allocation and use of energy, so as to maximize the use of energy. However, unlike the continuous adjustable and controllable traditional power generation, solar power generation cannot be accurately controlled, and it has a great randomness, which brings great challenges to photovoltaic power generation. In order to make full use of solar energy resources and ensure the real-time balance of power generation, transmission and consumption of the power system when photovoltaic power generation is connected to the grid, it is crucial to accurately predict the power of photovoltaic power generation. In order to solve the problems of unsatisfactory prediction accuracy and poor robustness when training set missing data, this paper adopts K-means algorithm for similar day screening and combines it with random forest algorithm(RF) to construct a photovoltaic power generation prediction model. Through similar day screening, more accurate and applicable prediction models can be obtained for forecasting days;at the same time, the robustness of the model is improved by using random forest model. Through verification with data from a power plant in Wuhan, Hubei Province, the accuracy and stability of K-means-RF algorithm in photovoltaic power generation prediction are proved.