District heating systems exhibit characteristics of high thermal inertia and strong nonlinearity, making it challenging to establish precise and universally applicable dynamic control models to enhance heating quality. This paper proposes a predictive control method for district heating based on regional adaptive group control hydraulic balance. By dynamically adjusting the outlet temperature of the heat exchange station and the valve openings of each user, it addresses the imbalance between supply and demand and the uneven distribution of hot and cold in district heating systems. This method first employs an adaptive group control approach to rapidly balance the hydraulic regime of all pipelines within the heating area. Secondly, it utilizes an Informer neural network combined with historical network information and future weather conditions to predict the average indoor temperature within the region. Finally, based on this model, a model predictive control method was designed to achieve energy-saving and comfort goals for the heating system. The study designed a hydraulic balance experimental platform based on a real community in Qingdao, China, and tested the proposed methods. The results demonstrate that compared to LSTM-based MPC and empirical heating curve control methods, the Informer-based MPC method considering hydraulic balance can more accurately control indoor temperature, while also saving 4.24% and 11.34% of energy consumption. The research findings mentioned above demonstrate that this method can significantly reduce the energy consumption and operational costs of heating systems, while simultaneously enhancing the stability of the heating systems and improving user comfort.