Long short-term memory as a forecasting tool for electricity demand in the Colombian market

High global energy demand poses environmental, social, and economic challenges, especially in developing countries such as Colombia. Effectively forecasting energy demand can facilitate decision-making and contribute to the optimization of resources within an electrical energy market. The goal of this work is to evaluate the feasibility of forecasting Colombia’s electricity demand with a 24-hour horizon considering different exogenous variables. The cross-industry standard process for data mining methodology, excluding the deployment phase, was used. This study analyzed time series hourly demand data from January 1, 2020, to August 31, 2024, comprising a total of 40,920 data points. The results showed that the best model captured more than 96% of the variability of the data on the basis of the coefficient of determination, with a mean absolute percentage error of 2.29%. This result suggests that the proposed model can support system operators, generators, marketers, and large consumers by providing an accurate forecast of energy demand in Colombia.