Short-term Forecast and Long-term Simulation for Accurate Energy Consumption Prediction

Accurate energy consumption forecasting has become pivotal for many companies as a way to tailor the budget dedicated to energy purchase on their actual power demand, thus sustainably minimizing energy waste and expenses. For these companies, both short-term and long-term energy consumption forecasts are a matter of interest since they would like to both program last-minute buy and sell and also plan future investments for power optimization. For this purpose, in this paper, different Deep Neural Networks techniques will be tested to perform both a supervised short-term energy consumption forecasting and an unsupervised long-term simulation via generative learning since very long-term forecasting (i.e., more than 1 year) is usually too inaccurate. The first task will be performed by adopting both a Recurrent Neural Network and a Long Short-Term Memory Network, while the second one will be performed by adopting a Generative Adversarial Network. Result on public data from the Australian Energy Market Operator will support the proposal.