建筑与区域负荷预测模型研究进展OA

This paper reviews the research progress on load forecasting methods in the field of buildings and district heating systems,providing references for the se-lection of forecasting models.Existing load forecasting models include physical modeling methods,time series analysis methods,machine learning methods,deep learning models,and hybrid models.Physical modeling methods offer clear mechanisms but are complex to build and lack real-time performance.Time series mod-els are suitable for short-term forecasting when load patterns are stable.Machine learning methods such as artificial neural networks,support vector machines,and random forests achieve high accuracy in medium-to short-term forecasting.Among these,support vector machines are suitable for small-sample problems,while random forests exhibit good robustness under high-dimensional data conditions.Deep learning mod-els and hybrid models show significant advantages in handling load forecasting with strong nonlinearity and multi-variable coupling,making them suitable for dis-trict heating systems.Considering forecasting perfor-mance and engineering feasibility,it is recommended to prioritize deep learning models and hybrid models for short-term load forecasting.Explainable artificial intelligence,multi-modal data fusion,and digital twin technology are identified as important development di-rections for load forecasting in building and district heating systems.