Akademik Veri Yönetim Sistemi
Electric Power Systems Research, cilt.258, 2026 (SCI-Expanded, Scopus)
This research introduces a hybrid strategy to address the Optimal Power Flow (OPF) challenge by integrating the Adaptive Artificial Hummingbird Algorithm (AAHA) with Convolutional Neural Networks (CNNs). The methodology consists of two stages. First, the AAHA algorithm generates various operating scenarios to optimize control variables under system constraints. Second, a CNN trained on these scenarios provides fast and reliable predictions, expanding the search space and mitigating the risk of falling into local optima. Case studies conducted on IEEE systems with 30, 57, and 118 buses demonstrate a notable reduction in generation costs and transmission losses. Furthermore, computation time was reduced by more than 95% compared to the standalone AAHA. The robustness and accuracy of the predictive model are confirmed through comprehensive performance metrics, sensitivity analyses, and statistical indicators. These results, together with comparisons against recent metaheuristic methods reported in the literature, validate the model's effectiveness. These findings demonstrate the flexibility and capability of the proposed methodology to operate effectively across power systems of varying scales. The approach presents a reliable data-driven solution for addressing static OPF constraints while ensuring the secure and efficient operation of electrical networks.