Advanced techniques for wind energy production forecasting: Leveraging multi-layer Perceptron + Bayesian optimization, ensemble learning, and CNN-LSTM models

Abstract

Due to the shortage of fossil fuels in many countries, power plants that rely on fossil fuels will be phased out in favor of wind turbines as the primary source of energy generation. These fossil fuel plants wreak havoc on the natural world, making humans and other life forms susceptible to illness. The production potential of wind turbines was investigated. Consequently, methods such as XGBOOST, Multi-Layer Perceptron with Bayesian Optimization (MLP + BO), Gradient Boosting Regression Tree (GBDT), Ensemble (gradient boosting and xgboost), and CNN Long Short-Term Memory (CNN-LSTM) have been utilized. A mean square error (MSE) of 7.2 in 45 seconds was achieved using the Ensemble technique, and an MSE of 6.8 in 450 seconds was obtained with the CNN-LSTM method. Wind power is readily available and straightforward to acquire globally, indicating its potential as a reliable and sustainable energy source.

Keywords

• Wind turbines
• Ensemble (gradient boosting and xgboost)
• Multi-layer Perceptron with bayesian optimization
• CNN-LSTM method