Abstract: It is an important means to consider the dynamic reconfiguration strategy of distribution network in the optimization of integrated energy system (IES) with electric-gas-thermal interconnection to improve the flexibility and economy of the system. In view of the complexity of the model and the low efficiency of solution after considering the dynamic reconfiguration of distribution network (DNR), a method based on double deep Q network (DDQN) and kernel ridge regression (KRR) is proposed to obtain the dynamic DNR strategy and incorporate it into the operation optimization model of IES. By transferring the time-consuming model learning process to the training process, only the trained model is applied, so as to quickly obtain the operation optimization results. Firstly, a database containing the IES states and static DNR results is established. According to the time series, the switching action cost is considered and the dynamic DNR strategy is formulated. Secondly, the KRR method is proposed to predict the optimal output of coupled units. Finally, based on DDQN, the mapping relationship between the IES states and the DNR results is exploited to achieve rapid dynamic DNR. The E33-G20-H16 system composed of IEEE 33 bus distribution network, 20 bus gas network and 16 bus thermal network and E78-G40-H32 is taken as an example, and the effectiveness of the proposed model and method is verified.