Long-term Hourly Cooling Load Forecasting through Combination of Global-domain and Sub-domain

Carbon neutrality has become a popular topic worldwide, and governments in various countries have started implementing strict measures to manage energy consumption. In developed countries, buildings account for 20-40% of total social energy consumption, with heating, ventilation, and air conditioning (HVAC) comprising almost half of the building's total energy usage. Therefore, accurate forecasting of hourly cooling loads can provide a strong basis for optimal HVAC control strategies to balance energy consumption and thermal comfort levels. This paper proposes an approach to accurately forecast HVAC hourly cooling load for the next three months. The approach combines both global-domain and sub-domain features and models. The evaluation results indicate that the proposed method achieved improved forecasting performance compared to the autoregressive model and long short-term memory model.