Forecasting of Photovoltaic Power with ARO based AI approach

Renewable power from sunlight can power the future smart grid with massive amounts of electricity. Systems struggle with solar energy's unpredictability and intermittent nature. Unpredictability of solar electricity hinders smart grid optimization and planning. Photovoltaic (PV) power generation must be accurately estimated to reduce power interruptions. PV power must be accurately predicted to avoid grid disturbances from PV facilities. Thus, we describe a transfer learning and AlexNet-based CNN architecture for short-term power forecasting. Past power, solar radiation, wind speed, and temperature readings determine the input. AlexNet's hyper-parameters are optimized using the artificial rabbit method. By adding selective opposition to ARO, local solution tracking efficiency is improved. CNN input features are created from all input parameters as 2D feature maps. After analyzing real PV data from Limberg, Belgium, the math shows that PV systems work.