Enhancing DER Scheduling in Active Distribution Networks: A Symbolic Deep Reinforcement Learning Approach for Improved Interpretability and Performance

Distributed Energy Resources (DERs) play a crucial role in enhancing the flexibility, security, and economic operation of Active Distribution Networks (ADNs). Deep Reinforcement Learning (DRL) algorithms have made significant progress in optimizing DER scheduling, particularly in dynamic distribution networks, due to their ability to handle continuous action spaces. However, the “black-box” nature of DRL models poses challenges related to safety and reliability in the practical deployment of power systems, limiting their widespread use. To address this issue, this paper proposes a novel symbolic Kolmogorov-Arnold Policy Gradient (S-KAPG) method, a symbolic policy network that optimizes scheduling performance while improving interpretability. S-KAPG leverages the universal approximation properties of the Kolmogorov-Arnold theorem and generates understandable symbolic expressions through a symbolic decision process, significantly enhancing the model’s transparency. Experimental validation on the IEEE 33-bus distribution system demonstrates that S-KAPG outperforms the traditional DDPG method in key performance indicators such as reward, voltage deviation, and system robustness, showcasing its superior advantages. Furthermore, S-KAPG provides more intuitive explanations in feature extraction, model visualization, and symbolic modeling. Finally, the accuracy of the generated explanations is verified through practical system applications, ensuring that the theoretical results can be effectively translated into real-world applications.