Floating Solar Power Generation Forecasting Model Based On Long Short-Term Memory And Extreme Gradient Boosting Methods

Abstract

The purposes of this research were to (1) build a forecasting model for floating solar energy generation based on Long Short-Term Memory and Extreme Gradient Boosting methods and (2) to evaluate the performance of a model based on Long Short-Term Memory and Extreme Gradient Boosting methods. The research methodology follows the deep learning pipeline, which consists of three main steps: In the first step, the data were collected from the electricity generation records and sensor readings obtained from a floating solar power facility with a capacity of 45 megawatts during February–October 2023, which consisted of 6,511 examples and 11 features. In the second step, feature extraction and classification were applied using the hybrid model by using two deep learning algorithms: 1) Long Short-Term Memory (LSTM) for memorizing long-term dependencies of time-series data, and 2) Extreme Gradient Boosting (XGBoost) for learning from uncertain data and predicting high performance; and In the third step, model evaluation was assessed using metrics including the mean absolute error (MAE), mean square error (MSE) and root mean square error (RMSE) for the indicated values from the forecasting model. The experiment result shows an average MAE of 0.0577, an average MSE of 0.0143, and an average RMSE of 0.1196 that represent suitable values in a real situation.