Prediction of Energy Demand in Smart Grid using Hybrid Approach

Predicting the energy consumption to provide valuable decision information for service providers and customers is a challenging task on the electricity market. The combination of mathematical and ANN methods results in a precise method of energy prediction than a single mathematical or ANN approach for predicting time-series data starting with different applications. The combination of (linear) seasonal auto-regressive integrated moving average (SARIMA) and (nonlinear) long-short term memory (LSTM) models are adopted to construct an advanced hybrid SARIMA-LSTM model for forecasting the time series data. Compared to existing models such as SARIMA and LSTM, a combination of the SARIMA-LSTM technique produces more accurate predictions. The results of the simulation reveal that better prediction is achieved by the proposed hybrid model. Further, the prediction model explores detailed energy consumption patterns with seasonal properties which makes both the service provider and customer to reduce their energy generation and consumption cost, respectively.