OPTIMIZATION OF DISTRIBUTED GENERATION PLACEMENT AND SIZING IN DISTRIBUTION SYSTEMS USING MULTI - OBJECTIVE DEEP REINFORCEMENT LEARNING

Abstract

Distributed Generations (DGs) play a significant role in modern distribution systems by reducing power losses, improving voltage stability, and enhancing system reliability. However, determining the optimal placement and sizing of DGs is a complex problem with diverse objectives and vast search spaces. This paper introduces the Multi - Objective Deep Reinforcement Learning (MODRL) algorithm to address this challenge. The objective function is designed to optimize power losses, voltage deviation, and investment costs. The method is validated on 33- bus and 69-bus distribution systems, with results compared to traditional algorithms (GA, PSO) and modern approaches (COA, WOA, FA). The results demonstrate that MODRL outperforms other methods, achieving significant power loss reduction while providing the best voltage stability and the lowest investment cost.