Renewable energy sources, particularly solar energy, are essential to meeting the world's growing energy needs. Real-time prediction of DC power generation is necessary for optimizing energy management and utilization of solar energy. We initially used regression models like Linear Regression, XGBoost, MLP Regressor, and Random Forest to predict DC power directly. While these models showed good accuracy, they were not ideal for time-series forecasting. To eliminate this limitation, we subsequently employed deep learning models LSTM, 1D CNN, and GRU to predict key features like AC frequency, AC voltage, DC link voltage, energy today, output current, total energy, output power, DC, pyranometer reading, temperature, and power factor based on date and time as input. These forecasted features were further used to predict the target variable, DC Power. The LSTM model most accurately predicted the intermediate features and the end DC power output among the deep learning models. Our LSTM network was trained on real data and achieved a Root Mean Squared Error (RMSE) of 0.039, a Mean Absolute Error (MAE) of 0.020, and an R² score of 0.909, indicating high prediction accuracy.
An AI-based system designed to simplify and improve the recruitment process through Artificial Intel...
