Comparing Univariate and Multivariate Methods for Short Term Load Forecasting

This paper introduces an improved multivariate method, which is vitally important to develop a short-term load forecasting module for planning and operation of distribution system. It has many applications including purchasing of energy, generation and infrastructure development etc. Thus, accuracy is very important in Load Forecasting. We have mentioned different time series forecasting approaches in this paper. Auto Regressive Integrated Moving Average model performs better in terms of accuracy than any other techniques as a univariate method. But ARIMAX, Auto Regressive Integrated Moving Average with exogenous variables, has proved itself as the most appropriate method in forecasting of the load profile for West Bengal using the historical data of the year of 2017. Several factors are there which influence the load demand. Weather or daily temperature has been taken into consideration here, which plays a major role in the daily load profile [2]. This paper computes Mean Absolute Error (MAE) for the mentioned forecasted model using ARIMA and ARIMAX in order to compare their applicability and discusses the merits and demerits for each to reach the optimum solution in week ahead and day ahead load profile prediction.