Wind power is becoming increasingly crucial as a renewable power source in the present day. However, due to the unpredictable nature of wind, accurate power forecasting is essential for a reliable energy system. Traditional approaches, such as Numerical Weather Prediction, have been utilized for wind power forecasting, but they often involve complex computations and meteorological models. Hence, machine learning algorithms have emerged as more compelling wind power forecasting tools. This paper focuses on predicting wind power using ten distinct models, including Gaussian Process Regression (GPR) with proper tunning of kernel function. Thereafter, the same wind power prediction is done with decision trees, linear regression, and Support Vector Machines (SVM), implemented through the MATLAB regression learner app and Python platform. The data from the Yalova wind farm is utilized for this analysis. Comparing these methods based on various forecasting errors such as Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and normalized RMSE (nRMSE) reveals that the modified exponential GPR model outperforms the others. It demonstrates superior performance compared to other models by significantly enhancing prediction accuracy and minimizing errors. In simpler terms, this indicates that the exponential Gaussian Process Regression model excels in making precise predictions while minimizing inaccuracies.
Forecasting of Wind Power Using Modified Exponential Gaussian Process Regression
2024-07-31
2911417 byte
Conference paper
Electronic Resource
English
Gaussian process regression for forecasting battery state of health
Online Contents | 2017
|A Variant Gaussian Process for Short-Term Wind Power Forecasting Based on TLBO
Springer Verlag | 2014
|A Variant Gaussian Process for Short-Term Wind Power Forecasting Based on TLBO
British Library Conference Proceedings | 2014
|