Load Forecast using ANN & VAR techniques for North Eastern Regional (NER) Grid of India

Abstract

The prediction of electric power or energy demand is required for efficient, economical, and reliable operation of the power system. Considering the importance of power demand forecasting, different models, as well as many new techniques have been proposed in recent times. In this paper, two forecasting methods have been used and compared. The artificial intelligence method, Artificial Neutral Network (ANN) is compared with Vector Auto-regressive (VAR) which is a statistical method. The methods are used to predict the Hourly day ahead short-term load for the NER states of India with two cases i.e. Weekday and Weekend (Saturday & Sunday). The result for the Assam State of India for a period of three days i.e., 23rd–25th January' 2021 has been presented in this paper. The comparison utilizes the Mean Absolute Percentage Error (MAPE). The simulation results have shown lower average values of MAPE for the three days in the ANN model (7.01 % w.r.t. SEM and 7.42% w.r.t. SCADA) than in the VAR model (7.49 % w.r.t. SEM and 7.94% w.r.t. SCADA) indicating better accuracy by the use of the ANN method in predicting the electric power demand forecast of the North-Eastern Region of India. Further, the performance of the ANN model in load prediction is also compared with other machine learning methods such as Random Forest (RF) and Support Vector Machines (SVM).