Climate change and electricity demand in Brazil: a stochastic approach

Abstract

We present a framework for incorporating weather uncertainty into electricity demand forecasting when weather patterns cannot be assumed to be stable, such as in climate change scenarios. This is done by first calibrating an econometric model for electricity demand on historical data, and subsequently applying the model to a large number of simulated weather paths, together with projections for the remaining determinants. Simulated weather paths are generated based on output from a global circulation model, using a method that preserves the trend and annual seasonality of the first and second moments, as well as the spatial and serial correlations. The application of the framework is demonstrated by creating long-term, probabilistic electricity demand forecasts for Brazil for the period 2016e2100 that incorporates weather uncertainty for three climate change scenarios. All three scenarios indicate steady growth in annual average electricity demand until reaching a peak of approximately 1071e1200 TWh in 2060, then subsequently a decline, largely reflecting the trajectory of the population projections. The weather uncertainty in all scenarios is significant, with up to 400 TWh separating the 10th and the 90th percentiles, or approximately ±17% relative to the mean.