A physics-informed artificial intelligence framework for virtual operational data generation to enhance planning and scheduling abilities in expanding district heating systems

To address the key challenges in urban district heating network expansion—the absence of historical data for new heat exchange stations, complex hydraulic coupling effects, and inherent system imbalance—this paper proposes a physics-informed generative AI framework for virtual operational data generation. The framework develops a graph theory-driven topology-lag fusion modeling method to accurately quantify the spatiotemporal lag correlation between stations and fully characterize the global hydraulic coupling mechanism; puts forward a Building-Thermal-Informed Liquid Neural Network (BTI-LNN) model with both strong physical interpretability and superior dynamic time-series fitting capability; and establishes a data generation system constrained by both national engineering standards and fundamental physical equations, a design that breaks the strong dependence of traditional methods on historical data. Case studies verify that, under high and low load conditions, the coefficient of determination (R2) between the data generated by the proposed method and the measured values exceeds 0.960 and 0.953, respectively, with the generation errors of water supply flow rate and temperature being less than 1.4 t/h and 0.8°C, respectively. The method thus holds significant engineering application and promotion value.