Hourly energy consumption forecasting by LSTM and ARIMA methods

Energy consumption represents the overall quantity of energy necessary for a specific activity or process, typically quantified in kilowatt-hours (kWh). In 2019, global electricity final consumption amounted to 22,848 terawatt-hours (TWh), indicating a 1.7% growth compared to 2018. During the same period, the Organization for Economic Co-operation and Development (OECD) recorded a total electricity final consumption of 9,672 TWh, reflecting a 1.1% decrease from the previous year. Conversely, non-OECD countries experienced a rise in final electricity consumption, reaching 13,176 TWh, marking a 3.8% increase over the 2018 figures. Time series analysis is a statistical method that examines data collected over successive time intervals. By identifying patterns and trends in historical data, this approach facilitates predictions and forecasts about future values. In our problem, forecasting and estimating the energy consumption in megawatts for the west regions of USA have achieved a good performance, with 97% for the R-squared metrics for LSTM and 98% for ARIMA models.