| Summary: | An increase in renewable energy demand and its energy mix caused the use of solar power to become crucial. However, the uncertainty of solar power generation due to weather conditions challenges solar power producers in planning large-scale solar projects. This aim is to apply a random forest (RF) algorithm for solar power production forecasting. The dataset used in this project is a combination of weather data from Solcast company and solar power production centers in selected states in Malaysia. The grid search method is applied to find the best hyperparameter configuration for random forests. Two parameters are tested: the number of trees and tree depth. The study reveals that more trees in an RF leads to a better model but only significantly improves with more trees. The relationship between tree depth and R-squared value becomes more linear. Too shallow or too deep tree depth can cause underfitting or overfitting, making it crucial to find the optimal depth for the model. It is found that the best number of trees is 11, and the best depth is set to 4. Besides that, the result shows that the achieved R-squared value is 0.9591. Testing the algorithm with different datasets is recommended to ensure it can be applied to any solar power production center location. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024.
|
|---|