Demand and price forecasting in deregulated electricity market

Abstract

In the project initially, we started with secondary research for understanding the structure of the deregulated market. We also studied the AESO’s website for understanding the market in greater detail We read research papers on forecasting of demand and price in such a market. In this process, we found out different methods to forecast the above variables and they can be divided in two parts: Artificial Intelligence based methods and Statistical Methods. Statistical methods included methods that have been discussed in the courses we have studied and include autoregressive (AR) models, linear regression models, dynamic linear or nonlinear models, ARMAX models, and threshold AR models. AI based models include Artificial Neural Network (ANN) models. We also studied about the various tools and software to be used to forecast demand and price. We collected data using the AESO databases and cleaned up the excel files to get the data that we required for predicting the demand and price. Weather data was also taken from online weather databases. We initially started with data of one year and used a part of it to develop our model and the other to test our built model. For demand forecasting, we have used ARIMA model and for price forecasting we have used LSTM model. We studied and researched about these models. After prediction, we calculated the MAPE, MAE and RMSE to check the accuracy of our predictions, basis which we made further improvements to our model.