Analysis of fusion of machine learning model and economic factors in electricity market price prediction

Abstract

The volatility of electricity market prices poses significant challenges for forecasting models, particularly amid dynamic economic conditions and fluctuating energy demands. This research introduces a novel model for electricity market price prediction that combines advanced machine learning (ML) techniques with economic energy factors to enhance forecasting accuracy and reliability. The proposed model integrates key economic energy indicators such as fuel prices (natural gas, coal, crude oil), inflation rates, currency exchange rates, industrial production indices, and electricity demand-supply ratios. These economic variables are combined with historical electricity price and load data to capture both short-term market fluctuations and broader economic influences. Data preprocessing using Z-score normalization was applied to standardize the input features, ensuring consistent scaling and improved model stability.  The forecasting architecture employs a multi-model ensemble strategy, utilizing a Least Squares Support Vector-fused Adaptive Random Forest (LSSV-ARF) model for electricity price prediction. The LSSV component captures nonlinear short-term fluctuations, while the ARF component adapts to evolving economic patterns. Empirical validation is conducted using real-world market data. The model is implemented in the Python platform, and the proposed method demonstrated substantial improvements over traditional models, such as MAE (0.79) and the MAPE (5.43%). These results confirm the effectiveness of integrating economic energy indicators with ML algorithms for electricity price forecasting, providing valuable insights for market participants, grid operators, and policymakers in managing energy pricing strategies.