Triple Exponential Smoothing Models for Forecasting Residential Natural Gas Consumption in Colombia

Abstract

The growing demand for natural gas in Colombia highlights the need for accurate forecasting tools to support energy planning, especially in the residential sector, where consumption patterns are shaped by climatic, socioeconomic, and behavioral factors. This study proposes a methodological framework for forecasting national residential natural gas demand using the additive and multiplicative Holt–Winters exponential smoothing models. Monthly historical consumption data from April 2020 to July 2025, obtained from the UPME’s Commercialization Bulletin, are used to calibrate both models. A nonlinear optimization procedure based on the Simplex method is applied to identify the optimal smoothing parameters that minimize the Mean Absolute Error (MAE) and Mean Absolute Percentage Error (MAPE). The results show that both Holt–Winters models achieve high predictive accuracy, confirming their suitability for short-term forecasting in the residential gas sector. The optimized multiplicative model demonstrates consistent parameter stability across different seasonal lengths, whereas the additive version yields slightly lower forecasting errors, with MAPE and MAE values of approximately 2.15% and 3.70, respectively. The proposed framework offers a replicable, computationally efficient tool aligned with national energy planning objectives, providing valuable insights for distributors, regulators, and market agents involved in supply management and operational decision-making.