Data Processing Methods in Solar Power Generation Forecasting Tasks

his paper addresses the problem of determining the weighting coefficients of parameters affecting the generation volumes using the random forest method. This approach enabled the exclusion of meteorological data that had minimal impact. Using these results, the study investigates the impact of various data preprocessing methods on solar power generation forecasting using the k-nearest neighbors (kNN) algorithm. During the study, a dataset was formed and processed using various preprocessing methods, including logarithmic normalization, double logarithmic normalization, Min-Max Scaling, and the exclusion of zero generation values. The results showed that logarithmic normalization is the most effective approach to improving forecast accuracy. The advantages and disadvantages of other approaches are also discussed, and their feasibility is justified.