Learning-Assisted Model for Optimal Placement of Protective Devices and Switches in Power Distribution Systems

The optimal placement of protective devices and sectionalizing switches plays a crucial role in enhancing the reliability of power distribution systems. However, the optimal switch placement requires a significant computational burden, which makes its implementation on large-scale power distributions challenging. To tackle such an issue, this article uses machine learning methods to propose a scalable and computationally-efficient model for simultaneously deploying breakers, fuses, and sectionalizing switches in large-scale power distribution systems. The model can find the optimal location and number of switching devices to optimize equipment costs and reliability indices. In particular, a learning-based assisting model is used to assist the mathematical switch optimization problem by, reducing the problem's search space. The input features of the assisting model are determined by applying explainable artificial intelligence tools. Then, the assisting model is trained by using transfer learning techniques. Finally, the proposed learning-assisted mathematical switch optimization problem is solved for large-scale systems by using commercial solvers. The efficiency and scalability of the proposed model are demonstrated through implementation on real power distribution systems. The numerical results show that the proposed learning-assisted model significantly outperforms state-of-the-art methods regarding scalability, computational cost, and optimality.