Enhancing Short-Term Load Forecasting: A Hybrid Approach with Prophet and Custom Feature Engineering

This paper presents an innovative hybrid strategy aimed at improving forecasting accuracy of short-term electric load forecasting by combining a custom feature engineering approach with the Prophet model. This work focuses on overcoming the limitations of conventional forecasting methods by capturing complex seasonality and nonlinear trends using Prophet’s additive regression capabilities, Custom feature engineering improves data insights and solves problems with forecasting. The suggested approach’s accuracy and forecast quality are demonstrated by experimental validation on an available real-world dataset. The results of multiple test cases and comparisons with existing models demonstrate the effectiveness of the proposed model. Benefits include resilience against outliers and missing numbers, domain-specific knowledge integration, and comprehensible results that give insightful knowledge about load predictions to utilities and energy providers. This hybrid approach represents a major advancement in the field of short-term load forecasting methods.