[Objective] In combined heat and power (CHP) heating systems, long-distance pipelines exhibit thermal inertia, transport delay and heat loss. These characteristics can provide flexibility for CHP dispatch, but they also increase the difficulty of describing temperature dynamics in optimization models. Steady-state and simplified delay models are usually unable to accurately represent temperature propagation, while full dynamic difference models introduce many internal spatial -node variables into online optimization. To improve temperature -propagation accuracy while controlling dispatch-model complexity, this paper proposes a dynamic two-port heat transfer model based on the Crank-Nicolson difference scheme. 

[Methods] The fluid energy conservation equation of the heating pipeline is discretized by the Crank-Nicolson scheme, which has second-order accuracy in both time and space. A four-point space-time grid is used to describe heat transport, and the dynamic mapping relationship between inlet and outlet temperature sequences is derived. By using the linear structure of the discretized equations, an offline matrix-recursion method is developed to algebraically eliminate intermediate spatial nodes. Thus, the outlet temperature of each pipeline is expressed as a linear combination of the inlet temper ature sequence, the initial 2 temperature state and the environmental heat-loss term. The full Crank-Nicolson dynamic model is then transformed into a low-dimensional two-port transmission relationship suitable for electric -thermal coordinated dispatch. The proposed model is embedded into a CHP dispatch framework considering heat network constraints, node mixing, heat source and load balance, temperature limits, CHP operating region constraints, conventional unit constraints and wind power accommodation const raints. 

[Results] Case studies show that the proposed model produces temperature responses close to those of the full Crank-Nicolson direct difference model. In the double-pipe heating system, the average absolute errors of the outlet temperatures of the s upply and return pipelines are 0.12 °C and 0.11 °C, respectively, and the online computation time decreases from 0.22 s to 0.012 s. In the multi-branch heating network, when the spatial step decreases from 20m to 5m, the online computation time of the full Crank-Nicolson model increases from 1.41 s to 5.31 s, whereas that of the proposed model changes only from 0.055 s to 0.060s. 

[Conclusion] The proposed dynamic two-port model retains the temperature propagation characteristics of the full Crank-Nicolson model while reducing the online optimization scale. After being embedded into CHP dispatch, it enables heating-network thermal storage and transport delay to release CHP downward regulation capability during high wind power periods, providing a low-dimensional modeling approach for flexible peak shaving and low-carbon operation of CHP systems in practical engineering applications.