Predictive Analysis of Electrical Load Forecasting and Classification of Feeder Efficiency with Random Forest Algorithm

Indonesian electricity companies are setting work objectives and allocating next year's investment using thorough corporate budget work plans. However, projecting kWh sales and setting priorities for electrical system improvements can be difficult. This study aims to address two problems: (1) the development of an electricity load estimation model to serve as a sales target and (2) the construction of a classification model to prioritize improvements based on feeder data characteristics for operational and investment budget purposes. To solve Problem (1), A Random Forest Regressor was implemented to enhance the accuracy of monthly electricity load forecasting from distribution system feeders. In addition, a Random Forest Classifier is applied to address the problem (2), enabling the prioritization of feeder improvements to support practical field implementation and assist corporate operational planning. The analysis yielded a regression model accuracy of 96,4% for predicting power load over the next 12 months. The efficiency classification model achieved an accuracy of 98,4%. Both models maintain accuracy within the company's accepted error tolerance limit of 10%.