In this paper, the upper optimization, based on the dynamic model of heating network, solves the problem of pipe network failure caused by too low temperature, the lower layer optimizes the capacities of the devices at the heat sources with the objective of economy according to the results of upper layer. Through the case analysis, the result shows that after optimization, the difference between the pipeline temperature and standard temperature is reduced by 33%, the minimum temperature of the pipe is increased from 52.5 °C to 76.7 °C, and the problem of pipeline failure is effectively avoided. Compared to the traditional capacity optimization model, the capacity planning cost of the two heat sources is reduced by 3.37% and 4.14% respectively. The bi- level programming model proposed in this paper can effectively avoid the operation failure of pipe network, and has better economic benefits after considering the variable efficiency characteristics of energy device.