Comparative Analysis of Linear and Nonlinear Regression Models for Peak Load Forecasting in Distribution Feeders

Abstract

Peak load forecasting provides essential information for planning and operating electrical distribution systems. This study investigates the forecasting performance of linear and quadratic polynomial regression models using monthly historical load observations from seven distribution feeders over the 2021–2022 period. To account for differences in loading patterns, each feeder was analysed independently. The linear model uses the time index as the predictor, whereas the quadratic polynomial approach is represented by a second-order polynomial regression. Model accuracy was assessed using Mean Absolute Percentage Error (MAPE), Root Mean Square Error (RMSE), and the coefficient of determination (R²). The results reveal variations in forecasting performance among the feeders. In general, quadratic polynomial regression yields slightly lower RMSE values, although the relative performance based on MAPE and R² differs across feeders. The differences between the two approaches are relatively small for most feeders, suggesting that greater model complexity does not necessarily provide a substantial improvement in forecasting accuracy. These findings highlight the importance of considering feeder-specific load characteristics and multiple evaluation indicators when selecting a forecasting model. Overall, the proposed approach offers a simple and computationally efficient alternative for feeder-level peak load forecasting based solely on historical load information.