Wind power generation prediction based on WRT and neural network

Abstract The objective of this study is to meticulously forecast wind power generation utilizing Wavelet Transform and neural network techniques. Leveraging extensive wind power generation data, we delve into the intricate characteristics of the dataset to formulate a prediction model that seamlessly integrates WRT and neural network principles. The aim is to attain a precise prediction of wind power generation, thereby furnishing invaluable decision-making insights for optimal dispatching and energy management in the realm of wind power generation. This endeavour holds significant implications for enhancing the efficiency and sustainability of wind energy utilization. Firstly, WRT is used to preprocess the wind power generation data. Utilizing wavelet decomposition and reconstruction techniques, we meticulously extract the pertinent information embedded within the data, thereby mitigating the detrimental effects of noise and outliers on prediction outcomes. Subsequently, a neural network is employed as the core prediction model, facilitating the realization of wind power generation prediction through the meticulous training of network parameters. To rigorously assess the predictive prowess of our model, we leverage actual wind power generation data for rigorous testing and benchmark it against other prediction methodologies. This comprehensive evaluation ensures the reliability and accuracy of our approach in forecasting wind power generation. The results show that the wind power generation prediction model based on WRT and neural network has high prediction accuracy and stability. Specifically, the prediction accuracy of the model on the test set is over 90 %, and the prediction error is reduced by over 20 % compared with the traditional prediction methods. This study also further analysed the source of forecasting error, found that the main reason is due to the uncertainty and volatility of wind power data. Therefore, the research proposed targeted improvement scheme, including optimizing WRT parameters, improving neural network structure and so on. The wind power generation forecasting model based on WRT and neural network proposed in this paper has high forecasting performance and practical value, and can provide strong support for optimal scheduling and energy management of wind power generation.