Automatic Generation Algorithm of the Wiring Diagram of Power Transmission Network

The automatic generation of power grid flow diagrams is one of the important research topics in the field of power grid operations. Reinforcement learning is commonly used to describe and solve the problem of an intelligent agent interacting with the environment to maximize rewards or achieve specific objectives through learning strategies. This paper focuses on the application of reinforcement learning to solve the problem of reorganizing power grid flow diagrams, also known as “automatic diagram generation” in the power industry.The method proposed in this paper for the automatic generation of power grid diagrams can be divided into two major steps: the first step is substation layout, and the second step is line adjustment. The substation layout serves as the environment for reinforcement learning in this study, while the reorganization of substation line layout represents the specific objective to be achieved. The Q-learning algorithm is adopted to implement the reinforcement learning process.In the proposed approach, the reinforcement learning environment refers to a region defined by a specific criterion, such as administrative division or power supply area. All substations are recursively divided into groups, and the resulting groups are allocated to their corresponding regions on the canvas based on the number of substations and the centroid coordinates. In solving the substation layout within each group, an auxiliary reward function is utilized to generate the Q-table, which aims to minimize line crossings and overall line length.To ensure the effectiveness of the generated diagrams, additional preprocessing and post-processing techniques are applied, including fine-tuning of single-connected substations, adjustment of overlapping lines, and considering the sequence of overlapping lines.Overall, this research proposes a reinforcement learning-based approach for automatic generation of power grid flow diagrams, addressing the specific objectives of substation layout and line adjustment. The Q-learning algorithm is employed, and various preprocessing and post-processing techniques are utilized to achieve desirable diagram results.