Prediction of Solar Power Irradiance using Machine learning and Deep Learning for Smart grid integration

To address the challenges posed by climate change and global warming, there is a growing emphasis on utilizing renewable energies. Among these, solar photovoltaic (PV) power generation, driven by solar irradiance (SR), has witnessed substantial growth in electricity production. However, the inherent variability in PV power generation poses challenges to the electric grid system, impacting stability, reliability, and operational planning, in addition to economic considerations. Therefore, accurate forecasting of solar energy becomes crucial for stabilizing grid operations, ensuring reliability, and facilitating the integration of large-scale PV power. This paper proposes an approach for solar energy prediction at TamilNadu(India) using machine and deep learning techniques. The effectiveness of the models is assessed for both real-time data and short-term solar energy forecasting, aiming to optimize management and meet security requirements. The study employs a unified solution based on a single tool and an appropriate predictive model. The dataset is sourced from the National Solar Radiation Database, encompassing features such as temperature, pressure, relative humidity, dew point, solar zenith angle, wind speed, and direction. The y-parameter, global horizontal irradiance (GHI), serves as the target variable. Various climatic models are trained using machine learning (ML) algorithms, including Random Forest (RF), Gradient Boosting, and a deep learning algorithm known as Long Short-Term Memory (LSTM). LSTM demonstrates superior accuracy and performance for both real-time and shortterm predictions compared to ML methods, which exhibit notable errors.