FORECASTING UTILITY DEMAND BASED ON HISTORIC DATA APPLYING DEEP LEARNING TECHNIQUES

Abstract

Load Forecasting has been one of the most emerging area of research for the last several years. Power system scheduling, reducing the expense of spot purchase of power, day to-day operation and efficiency are some of the very interesting outcomes that can be explored by load forecasting. The development of Smart Grid and Energy Management System, aggregates large-size of data adding to the complexity of the system. Big Data Analytics is a modern day technique that can extract information from these complex and large datasets. Typical load profiles exhibit periodicity, allowing to extract patterns from demand time series and available historical recordings. However there are many factors that cause strong variations of the demand patterns from the predicted values.Deep learning models can learn from a considerable volume of big data, insufficient data that contains missing values, heterogenous data. Artificial intelligence (AI) can be combined with big data technology to solve complex problems in demand forecasting. This project is aimed at comparing the load prediction based on Artificial Neural Network(ANN), Long Short Term Memory(LSTM) and Bidirectional Long Short Term Memory(BLSTM).