Short-Term Seasonal Energy Forecasting

Load Forecasting is an important tool used to predict future electrical utility data in a Smart Grid. Forecasting uses past data to predict trends for future energy usage. This paper investigates the application of exponential smoothing, ARIMA and recursive neural networks in TensorFlow on utility data taken from 20 different US Utility zones over four years, from 2004-2007. These methods will be used to forecast and predict the load data for each yearly quarter in 2008.