Abstract— This research proposes a hybrid framework based on deep learning an‎d intelligent optimization to enhance the reliability of distribution networks in the presence of renewable energy sources. The primary objective is to minimize the economic cost of reliability (ECOST), considering the generation uncertainty of photovoltaic (PV) units. To this end, a deep learning neural network is utilized to forecast the hourly output power of five PV units installed at buses 21, 30, 33, 73, an‎d 81 of an 85-bus distribution network. The modelʹs training data were generated via simulation. The output of the deep learning model is fed into a Particle Swarm Optimization (PSO) algorithm to determine the optimal network topology while considering operational constraints (radiality, voltage, an‎d current). The reliability index is calculated by combining the effective load (load minus renewable generation), interruption cost, an‎d the dynamic failure rate of branches. Simulation results in the MATLAB 2023 environment demonstrate that the proposed approach significantly reduces the economic cost of power outages compared to the base case (traditional reconfiguration without forecasting). This improvement is particularly evident during peak load hours an‎d periods of PV generation fluctuations. The integration of accurate forecasting with dynamic optimization enables efficient management of uncertainty an‎d facilitates decision-making under real-world conditions.

احسان آزاد فارساني

پروين رستگاري , محمداسماعيل نظري