Study on the load uncertainty of buildings and villages in western Sichuan Plateau based BP neural network integrating prior knowledge

Abstract

Due to its unique climatic conditions and occupant behaviors, the precise design of heating systems in the western Sichuan Plateau (WSCP) presents significant challenges. Existing research has primarily focused on climate change and solar energy utilization systems, with limited exploration of the combined impacts of weather variability and occupant behavior uncertainties on building and village heating demands. This study obtained three typical buildings and occupant behavior patterns through field surveys and on-site measurements. Further, it developed a novel algorithm combining a particle swarm optimization (PSO) algorithm with a backpropagation (BP) algorithm that incorporates prior knowledge to predict heating demands for buildings and villages under the dual uncertainties of weather and occupant behavior across different locations. The results show that considering both weather and occupant behavior uncertainties, the peak heating load of buildings may increase by up to 119.3 % compared to design conditions. Conversely, when considering the herding activities of the inhabitants, the overall peak heating demand of villages may decrease by 38.1 %. Additionally, in the WSCP, the mean uncertainty of building heating demand exhibits an upward trend with decreasing longitude and increasing latitude. Therefore, future designs of distributed and centralized heating systems in this region should fully incorporate the impacts of uncertainties from weather conditions and occupant behaviors. The findings of this study provide critical references for the precise design of heating systems in the WSCP.