Heat Index-Driven CNN-BiLSTM Load Forecasting under Hot and Humid Climates

As climate change intensifies, humid and hot regions are facing increasingly complex fluctuations in power load. Traditional load forecasting models often treat temperature as a key meteorological feature; however, in hot and humid environments, temperature alone fails to accurately capture the relationship between human thermal perception and electricity usage behavior This study proposes a CNN-BiLSTM model driven by the Heat Index for short-term power load forecasting in humid and hot climates. Based on hourly power load and meteorological data from Guangzhou in 2023, two model schemes were developed and compared—one using the Heat Index and the other using conventional temperature as input features. Experimental results demonstrate that the Heat Index more effectively captures the impact of hot-humid environment on power load. The model outperforms the temperature-driven scheme across multiple evaluation metrics, confirming the applicability and advantages of using the Heat Index as a forecasting feature in such climates.