سامانه همایش های دانشگاه آزاد اسلامی واحد نجف آباد

صفحه اصلی / ششمین کنفرانس بین المللی بهینه سازی مصرف انرژی الکتریکی

Distribution Network Reconfiguration in the Presence of Distribution Generation Using Deep Reinforcement Learning

نویسندگان :

Amirhossein Ghaemipour¹ Habib Rajabi Mashhadi² Seyed Hossein Mostafavi³

1- دانشگاه فردوسی مشهد 2- دانشگاه فردوسی مشهد 3- دانشگاه فردوسی مشهد

کلمات کلیدی :

Distribution network،reconfiguration،Distribution Generation،Deep reinforcement learning،Deep Q Network (DQN) algorithm

چکیده :

Due to the increasing complexity of distribution networks for various reasons, such as the widespread use of distributed generation, demand response, various load management programs, and smart devices in network control, efficient distribution network operation has become an important issue. Utilizing distribution network reconfiguration (DNR) in these situations can provide appropriate flexibility to achieve the goals of the network operator. In this research, reinforcement learning theory is used for distribution network reconfiguration, which, unlike other methods, does not require any previously prepared data or parameters. Based on trial and error and receiving feedback on its performance from the environment, it reaches the optimal solution. The deep Q network (DQN) algorithm minimizes power losses and improves the network voltage profile. This design has been tested and verified on an IEEE 16-bus distribution network.