Enhancing Electric Power Consumption Prediction Using Machine Learning and Genetic Algorithms

Electric power consumption prediction enables the development of smart strategies aimed at conserving energy, enhancing efficiency, and establishing sensible pricing mechanisms. In this paper, we used predictive models for electric power consumption using machine learning and genetic algorithm techniques. The dataset utilized consists of socio-demographic instances and power consumption-related measured values collected by Centrales Eléctricas de Nariño (CEDENAR), Colombia. Three predictive models—Random Forest (RF), Gradient Boosting Machines (GBM), and Support Vector Machines (SVM)—were employed to predict power consumption. Additionally, a genetic algorithm (GA) was applied to optimize the parameter values of each model. Performance metrics, including mean absolute error (MAE), mean absolute percentage error (MAPE), root mean squared error (RMSE), and coefficient of determination ($\mathrm{R}^{2}$), were used to evaluate the effectiveness of the models. Our findings indicate that all models, both in their standalone forms and after genetic algorithm optimization, demonstrate improvements in performance metrics. However, statistically, the RF + GA model exhibits superiority only in the MAE and MAPE metrics, achieving an $\mathrm{R}^{2}$ of 0.76, reflecting its enhanced predictive capability in electric power consumption estimation. These results offer valuable insights for energy management strategies and resource planning in the context of electricity consumption.