Optimal Control of Distribution Switches Using Deep Neural Networks to Balance a Low-Cost Dynamic Photovoltaic Microgrid |
Ricardo Calloquispe-Huallpa1, Rachid Darbali-Zamora2, Erick E. Aponte-Bezares1, Anny Huaman-Rivera1 1 University of Puerto Rico at Mayagüez (UPRM), Mayaguez, PR, Puerto Rico /2Sandia National Laboratories, New Mexico , NM, United States |
Energy management is essential to maximize the efficiency, dependability, and stability of microgrids. Energy management in microgrids usually involves strategic switching operations to control loads and generation in a network. This can be achieved by employing intelligent agents, such as deep neural networks (DNNs), to make decisions regarding microgrid switching operations. In this paper, a novel approach was employed by enabling a DNN to train from a prior optimization using a genetic algorithm. An electrical model for a microgrid was built in MATLAB to study its dynamic behavior during switching operations. The proposed DNN achieved an accuracy of 96.2 %. Since the DNN trained from optimized results, it was able to perform instantaneous decision making. This approach confers a significant advantage, as the DNN bypasses the iterative process typically associated with metaheuristic algorithms, thus reducing computation time and effort. |