Abstract: To address the bottlenecks of low efficiency, excessive reliance on manual experience, and insufficient utilization of historical data in natural gas pipeline network fault handling, this paper proposes an auxiliary decision-making method for the optimal selection of control stations, integrating historical experience with deep reinforcement learning. First, a framework for collecting and preprocessing historical fault handling data is constructed to systematically transform unstructured dispatching records, such as operation logs, into structured datasets suitable for machine learning, achieving an initial quantification of experts' tacit knowledge. Second, the problem of selecting an optimal combination of control stations under fault scenarios is modeled as a Markov Decision Process (MDP). We meticulously define a state space incorporating fault information and network topological features, a discrete action space based on adjustments of key stations, and a phased reward function that quantifies the consistency between the agent's decisions and historical expert solutions. Third, a Double Deep Q-Network (Double-DQN) algorithm is employed to train the decision-making agent. Through an experience replay mechanism and the separation of the target network, the agent efficiently learns the strategy for selecting optimal station combinations under various fault conditions, guided by expert experience. Finally, the trained model is deployed in real-time fault handling scenarios to provide dispatchers with data-driven recommendations for optimal station selection. Experimental results demonstrate that the proposed method achieves a precision of 94.8% and a recall of 95.2% on the test set, indicating a high degree of correspondence between the model's recommendations and historical expert decisions. The proposed method effectively realizes the explicitation and modeling of dispatchers' tacit knowledge, showing great potential to significantly improve fault response speed and decision quality, and provides a new technical approach for the intelligent dispatching of natural gas pipeline networks.