Conventional approaches to heating network operation, such as single-objective optimization constrained solely by minimum energy consumption or macroscopic regulation strategies, often overlook critical issues. These include the rationality and optimality of flow distribution among different users, the balance between individual user supply and demand, and the fulfillment of indoor thermal comfort requirements. To address these limitations, this study proposes a bi-level multi-objective optimization framework that comprehensively evaluates three key factors: energy consumption, the deviation between heat supply and demand, and hydraulic balance. The Non-dominated Sorting Genetic Algorithm II (NSGA-II) is employed to solve this model and find the minimum value of the objective function – representing the optimal solution. This solution yields the optimal return water temperature and supply water flow rate parameters for the given operating moment, thereby guiding network operation to achieve the multi-objective optimum. Practical implementation in a residential district project in Tianjin demonstrated the effectiveness of the trained optimization model: it reduced pump energy consumption by 10%, achieved an average supply-demand deviation error of only 1.67%, and reduced the standard deviation of pressure drop between pipelines by 50%.