Estimación de la curva de demanda eléctrica en Ecuador a mediano y largo plazo utilizando el método Mapreduce para la planificación óptima de la expansión de la generación

Abstract

Accurate electricity demand forecasting is essential to ensure reliable supply in developing power systems. Conventional methods present limitations when processing large data volumes and projecting under uncertainty. This article develops a methodology that integrates MapReduce with temporally validated econometric models to estimate demand for Ecuador's National Interconnected System. A total of 87,650 hourly records (2015-2019) were processed through statistical preprocessing, load pattern extraction via distributed processing, and calibration of six models with temporal backtesting. The exponential model proved optimal with MAPE of 1.9% and backtesting error <5.1%. Projections indicate growth from 4,500 MW (2020) to 6,900 MW (2030), requiring 2,100 MW expansion with USD 3,944 million investment distributed across hydroelectric (45%), renewable (35%), and thermal (20%) generation, ensuring system reliability.