Investigation for Applicability of Deep Learning Based Prediction Model in Energy Consumption Analysis

The electricity demand is rising in our day-to-day life and business; therefore, it is important to forecast electricity consumption to balance the demand and supply chain. The prediction mechanism will enable us to meet the demand quickly and use the available resources optimally. It will help us to reduce the waste of energy. Several statistical methods have been proposed to solve this issue, but they do not achieve the goal due to tedious topology and monitoring mechanisms. This research aims to forecast the load hourly or annually without rigorous monitoring and with fewer resources using Deep Learning models. In this work, Recurrent Neural Network (RNN), Long-Short Term Memory (LSTM), and Gated Recurrent Unit (GRU) based models are implemented. Their performances are measured using different evaluation metrics such as Mean Squared Error (MSE), Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) to identify the best-performing model to forecast the load. The dataset is formed by collecting data from several open-source platforms containing the required parameters affecting the load day or night throughout all seasons. The proposed model can be utilized in the optimal production and supply of energy by accurately forecasting the energy load at a time.