Comparative Analysis of Six Short-Term Load Forecasting Models for a Distribution Transformer

Abstract

The paper presents the short-term (day-ahead) time-series load forecasting models based on the ARIMA, SVR, MLR, ML/ANN i.e. NARX, ANFIS, and Exponentially Weighted Elasticnet with Fourier Series (EWENFS) at 990 kVA Distribution Transformer (DT) level. The proposed load forecasting models significantly improve the forecasting errors for DT level day-ahead forecasting for each season. The models show that ARIMA and EWENFS models yield better predictions as compared to other models. However, the reasons behind the high error of ANN needs to be examined to improve our learning. For summer, monsoon & winter seasons minimum MAE obtained are 10.27 kVA, 6.97 kVA, and 7.19 kVA respectively.