Predicting Energy Generation in Ecuador Using Machine Learning

The Electric power generation in Ecuador faces structural challenges that limit the efficiency and sustainability of the system, particularly in the context of climate variability such as El Niño or prolonged droughts. The country currently relies mainly on hydroelectric sources managed by CELEC EP, making it vulnerable to extreme environmental events. This paper proposes a predictive model based on Long Short-Term Memory (LSTM) neural networks to forecast national energy generation. Using open historical data from the Government of Ecuador and applying Exploratory Data Analysis (EDA) techniques, patterns, anomalies, and seasonal trends in electricity production are identified. The research aims to overcome CELEC EP’s planning limitations by integrating artificial intelligence tools to improve system responsiveness and sustainability. A quantitative, non-experimental, and longitudinal methodological approach is adopted, evaluating different predictive models using standardized metrics to assess their applicability in the Ecuadorian context.