A hybrid time series forecasting approach integrating fuzzy clustering and machine learning for enhanced power consumption prediction

Power demand estimation in Tetouan, Morocco, uses fuzzy clustering with machine learning-based time series forecasting models as the main subject of research. This paper tackles an important requirement for forecasting methods that accurately predict electricity use in areas with changing demand to enhance energy management capabilities. An evaluation of 52,417 records containing six characteristics derived from three power networks formed the basis of this analysis. A comparison of Random Forest, Support Vector Machine, K-Nearest Neighbors, Extreme Gradient Boosting, and Multilayer Perceptron models took place through Root Mean Square Error, Mean Absolute Error, and R² metric evaluation. Model performance improved after fuzzy clustering integration, resulting in the multilayer perceptron achieving its best results with RMSE at 355.42, MAE at 246.43, and R² of 0.9889. The hybrid approach is an original practical solution that improves the forecasting accuracy of power consumption.