Short-Term Power Prediction of Wind Turbine Applying Machine Learning and Digital Filter

As wind energy development increases, accurate forecasting helps to develop sensible power generation plans and ensure a balance between supply demand. Machine-learning-based models possess exceptional predictive capabilities, data manipulation prior model training is also key focus of this research. This study trained deep Long Short-Term Memory (LSTM) neural network learn the processing results Savitzky-Golay filter, which can avoid overfitting due fluctuations noise in measurements, improving generalization performance. The optimum frame length match second-order filter was determined by comparison. In single-step prediction, method reduced root-mean-square error 3.8% compared directly with measurements. produced smallest errors all steps multi-step advance prediction. proposed ensures accuracy and, on that basis, improves timeliness effective forecasts.