Fan key part fault diagnosis method

Abstract

The invention relates to a fan key part fault diagnosis method, which comprises the steps of obtaining an operation monitoring record and a fault record of an SCADA system, and determining an experiment data set; preprocessing the experimental data set, and encoding the preprocessed fault data and normal data; carrying out stratified sampling on the preprocessed experimental data set, calculatingthe weight of each observation characteristic parameter in the SCADA system in fault multi-classification by utilizing a ReliefF algorithm, screening out the observation characteristic parameters withrelatively high relevancy, and updating the experimental data set; dividing the updated experimental data set into a training set and a test set, inputting the training set into an XGBoost model fortraining, determining an optimal target function of XGBoost, and optimizing the XGBoost model; and inputting the test set data into the optimized XGBoost model, and performing fault multi-classification test on the test set data. The method can be adjusted according to actual faults in recorded data of the SCADA system, and is more feasible, simple and high in fault recognition accuracy.

Description

technical field [0001] The invention belongs to the technical field of fan fault diagnosis, and in particular relates to a fault diagnosis method for key parts of a fan. Background technique [0002] The environmental conditions of wind farm construction sites are harsh, and they are generally located in mountains, deserts or seas, which also leads to frequent failures of wind turbines and difficult maintenance. Foreign scholars have studied the fault data of wind farms in Germany, Denmark, Sweden, and Finland and found that electrical system, sensor, and hydraulic system faults are very common, but more than half of the faults that cause wind turbine shutdowns are related to generators and gearboxes. . The gearbox is the component with the longest downtime per failure, mainly due to difficult repairs inside the pod. Generator failure rates are low, but downtime is high. Conversely, the control system had the highest cumulative failure rate but a lower cumulative downtime...

AI Technical Summary

Problems solved by technology

First, there are difficulties in feature extraction and selection

Due to the large number of monitoring characteristic parameters of the wind turbine SCADA system, how to select the special monitoring parameters related to the fault has a great impact on the accuracy of fault diagnosis. However, the commonly used methods still rely on experience to determine the parameters, which lacks a certain scientific basis.

Second, there is a lack of high-precision fault diagnosis methods

Therefore, it is difficult to achieve high diagnostic accuracy in practical applications

Third, the fault diagnosis is single, and most fault methods only diagnose a specific fault or two or three faults, but require various complex algorithms, so it is of little significance in the actual application of wind farm fault diagnosis

Images