Forecasting Solar And Wind Energy Production Using Artificial Intelligence

Abstract

Background: Saudi Arabia’s Vision 2030 aims to generate 50% of electricity from renewables by 2030, leveraging abundant solar and wind resources. However, variable weather conditions challenge accurate energy production forecasting, requiring advanced AI models to ensure grid stability and efficient energy management.Aim: To develop an AI-based model integrating machine learning (ML) and deep learning (DL) to enhance solar and wind energy forecasting accuracy using meteorological data, supporting Vision 2030’s sustainability goals.Patients and Methods: Using Kaggle datasets with weather variables (temperature, solar radiation, wind speed, humidity), the study preprocessed data to address missing values and outliers. ML models (Random Forest, XGBoost, K-Nearest Neighbors, Extra Trees) and DL models (Deep Neural Networks) were trained and evaluated via RMSE, MAE, and R². A Streamlit dashboard was built for real-time forecasting.Results: XGBoost excelled, with the lowest RMSE (402.94 for solar, 187.61 for wind) and highest R² (0.9737 for solar, 0.9794 for wind). Random Forest performed well, while DNN showed lower accuracy (R² = 0.5269 for solar). The model predicted 102,568.73 MWh daily solar output, supporting 3.4 million homes and reducing CO2 emissions by 96,414.61 tons daily.

Conclusions: The AI model, particularly XGBoost, enhances renewable energy forecasting, aiding grid stability and aligning with Vision 2030. The interactive dashboard improves usability. Future work should explore advanced DL and real-time data integration.