Abstract: In order to make full use of demand response to improve energy utilisation and reduce energy consumption in multi-station convergence systems, an energy routing strategy for multi-station convergence systems based on neural networks and power big data for load forecasting is proposed. Customer comfort is firstly considered and measured using the PMV index, and models for incentive-based demand response and price-based demand response are developed. Load forecasting is then carried out based on neural network and power big data approaches, and a multi-station fusion system model is developed, including energy hubs such as gas turbines, gas boilers, absorption smart machines, waste heat boilers, electric chillers and P2G units. Finally, the solution is based on a differential evolutionary algorithm. The algorithm is extended with an energy demonstration zone in Beijing as the basic prototype. Before solving the model, a linear model solution without considering the demand response is obtained and then used as the initial solution for the differential evolution algorithm solution, which is conducive to improving the solution accuracy and solution speed. Comparing the model before and after performing demand response, it is found that performing demand response has a great effect on reducing costs and saving energy.