ELECTRICAL LOAD FORECASTING USING MACHINE LEARNING TECHNIQUES AND THEIR COMPARISON

Abstract

The work presented gives hourly electrical load forecasting as a time series forecasting model using multilayer deep learning Long Short-Term Memory neural network Technique and its detailed comparative study with various Machine Learning Techniques based on their Mean Squared Error, Mean Absolute Percentage Error and Training time. Load Forecasting has immense potential to help in modulating the generation and distribution potentials of our smart grids in accordance to the requirement so that optimum power is generated and supplied through various channels which would be effective in grid management and operations. The MAPE of the model presented below is 0.41.