Convolutional Artificial Recurrent Neural Network (CAR-NN) deep learning model for accurate load forecasting in distribution grid /

Abstract

With the introduction of renewable energy sources (RES), the world has entered into a discussion where complete shift from conventional energy sources (CES) to renewable energy sources is being considered a more viable option. However, conventional energy sources provide a capacity factor (CP) of over 72% (Coal, LNG, Furnace oil power plants) whereas renewable energy sources have not been able to provide a CP over 25%. Moreover, CES are mature technologies with extensive research and industrial applications available for use which are over 200 years old. RES, however, is only 40-50 years old with not much research and industrial applications to completely replace CES. So, a complete cutover from CES to RES is not possible at the given time. But the world is going through an energy crisis at present and a quick solution is required. A possible solution is to make a hybrid grid with inclusion of both CES and RES into the grid and smartly manage the available energy as per demand. But to enable the grid to make such important decisions, we need to forecast the demand which will help our grid make better decision in energy distribution. Machine Learning (ML) and Neural Network (NN) techniques will enable our grid to forecast the demand and empower the grid to make such important decisions.