Wind turbine generator main bearing temperature prediction method based on deep learning

Abstract

The invention provides a wind turbine generator main bearing temperature prediction method based on deep learning. The method comprises the following steps: firstly, collecting sensor data collected by an SCADA (Supervisory Control and Data Acquisition) system of a wind turbine generator in a wind field, then cleaning and re-sampling the data, selecting SCADA parameter variables related to temperature change characteristics of a main bearing, and constructing a training sample and test sample data set; secondly, a neural network model based on deep learning is constructed through stacking self-coding, and internal characteristics of normal operation data of the main bearing are fully mined through repeated training; and finally, adding a regression prediction layer at the top of the model to further finely adjust the whole deep learning model until intelligent prediction of the temperature of the main bearing is met. Through the method, indexes such as prediction precision and errors of the method are superior to those of a traditional shallow learning model, and technical auxiliary support can be well provided for operation state monitoring and fault early warning of the main bearing.

Description

technical field [0001] The present invention mainly relates to the technical field related to the monitoring and identification of the operating status of large-scale direct-drive wind turbines, in particular to a method for predicting the main bearing temperature of wind turbines based on deep learning. Background technique [0002] With the rapid development of wind power technology and wind power industry, more and more wind farms and larger-capacity wind turbines have been put into use. However, most wind farms are located in remote or inaccessible mountains, wilderness and ocean, and the working environment of wind turbines is harsh and complex. These factors lead to frequent failures of wind turbines, and the difficulty and inconvenience of replacing components (especially large components). This ultimately led to a dramatic increase in operation and maintenance (O&M) costs, especially for offshore wind farms. Therefore, the development of condition monitoring and fau...

AI Technical Summary

Problems solved by technology

These models are shallow machine learning models with limited data feature learning capabilities

In addition, as the depth of the model increases, the network is prone to overfitting and the parameters are difficult to adjust

Images