Enhancing Short-Term PV Power Forecasting Using Deep Learning Models: A Comparative Study of DNN and CNN Approaches

Forecasting solar power is essential for increasing solar power plants' competitiveness in the energy market and reducing reliance on fossil fuels for social and economic advancement. The novel method of forecasting solar Photovoltaic (PV) power output through the use of deep learning and machine learning techniques specifically, Deep Neural Network (DNN) and Convolutional Neural Network (CNN) models is presented in this research. To anticipate solar energy generation, our technique uses pertinent weather characteristics and historical PV power data as inputs. Although several PV power forecasting models in the literature have varying degrees of accuracy, our CNN and DNN models provide a clear benefit by obtaining important features from raw local data. With this skill, precise short-term projections of solar PV power output from a few hours to several days-can be made. In addition, the integration of naive seasonal forecasting with CNN and DNN models improves the precision of short-term power generation forecasts by identifying complex dependencies and periodic patterns in the data. All things considered, our suggested method offers a viable way to raise the accuracy and efficiency of solar energy forecasting, which will help the energy industry embrace sustainable energy sources more widely.