A Novel Hybrid Model Utilizing Data-Physical Integration for Urban Air-Conditioning Loads Forecasting

The increase in uncertainty in the new power system has made it difficult to balance power and electricity. As a flexible regulation resource, urban air-Conditioning loads (UACL) has a high proportion of load and can assist the power system in peak shaving and frequency modulation. However, in the day-ahead planning of the power grid, due to the lack of comprehensive and prediction perception of air-Conditioning load data, this article proposes a data-physical integrated load forecasting method for UACL. By using weather forecast data to drive physical models such as urban heat effect (UHE), urban building heat inertia, and user comfort, the mapping relationship of the prediction model is enhanced; in order to effectively represent the spatial-temporal relationship between weather and air-Conditioning load, a CNN prediction model is established. The proposed data-physical integration framework was tested in a city in China, showing high accuracy. When the power grid cannot collect accurate user information, the method proposed in this article can effectively perceive the day-ahead change trend of UACL.