A prediction method for fan blade icing based on feature selection and xgboost

Abstract

The invention relates to a fan blade icing prediction method based on feature selection and XGBoost. The fan blade icing prediction method comprises the following steps that (1) SCADA monitoring dataof a fan are obtained, denoising, dimensionality reduction and normalization are conducted on the SCADA monitoring data, and the timestamp dimensionality a time format is converted to a time sequenceformat; (2) feature selection is conducted through a Relief feature selection method to extract the key dimensionality and to reduce the dimensionality of the monitoring data, and the monitoring datawith the dimensionality reduced are classified into training samples and test samples; (3) the training samples are substituted into an XGBoost model to be trained, and the trained model is measured and evaluated according to the precision ratio, the recall ratio, the accuracy and the comprehensive index F1; and (4) to-be-verified monitoring data are substituted into the trained model to obtain aprediction result indicating whether or not the fan is iced, and a decision is made according to the prediction result to make the fan operate normally. Compared with the prior art, the fan blade prediction method has the advantages that the prediction accuracy is high, and the prediction speed is high.

Description

technical field [0001] The invention relates to the field of fault diagnosis of wind turbines, in particular to a method for predicting icing of fan blades based on feature selection and XGBoost. Background technique [0002] In recent years, the development of wind power generation is getting faster and faster, and there are more and more problems. Generally, the places with abundant wind energy resources are mostly located in cold plateau areas. The high altitude and low temperature in these places directly lead to the freezing of blades, the loss of materials and Structural performance changes and load changes pose a greater threat to the power generation performance and safe operation of wind turbines. At present, the real-time data of fan operation is mainly monitored and stored by the SCADA system. The method of judging the blade icing fault status is mainly to compare the deviation between the actual power of the fan and the theoretical power. When the deviation reach...

AI Technical Summary

Problems solved by technology

However, when the alarm is triggered, a large area of ​​ice on the blades has often occurred, and operating under such conditions will increase the risk of blade breakage and damage

Images