Failure prediction diagnosis algorithm of wind turbine generator gear box

Abstract

The invention discloses a failure prediction diagnosis algorithm of a wind turbine generator gear box based on kernel principle component analysis (KPCA) and a support vector machine. Based on full consideration of indexes of temperature change of all elements when a failure occurs on a gear box and output power change before and after the failure occurs, the KPCA algorithm is adopted to reduce an input dimension for characteristic extracting, irrelevant data is abandoned, the model training speed can be drastically increased, and the failure diagnosis time can be reduced. Meanwhile, the support vector machine is introduced to conduct classified training on the data so as to improve the generalization ability; moreover, with the help of an expert system, the result is analyzed and explained, accurate and detailed information can be provided for a human-computer interaction interface, and thus precise diagnosis of the failure is achieved.

Description

technical field [0001] The invention relates to the field of fault diagnosis algorithms, in particular to a fault prediction and diagnosis algorithm for a wind turbine gearbox based on KPCA and support vector machines. Background technique [0002] Over the past few years, the country has attached great importance to new energy power generation. Wind power has attracted widespread attention as the cleanest and most environmentally friendly new energy generation technology. The rapid development of wind power has also brought an unprecedented problem to wind power equipment manufacturing - fault diagnosis. The gear box of the transmission device in the mechanical system is one of the equipment with the highest failure rate in the wind turbine, and its operating status will directly affect the operating status and power output of the wind turbine. Therefore, it is necessary to carry out in-depth fault diagnosis of the gearbox of the wind turbine. Research. [0003] Currentl...

AI Technical Summary

Problems solved by technology

However, for multi-variable high-dimensional data, the neural network tends to fall into the local minimum problem, and there will be over-adaptation phenomenon.

Support Vector Machine (SVM) is slow to process multi-variable high-dimensional data, and takes a long time to calculate, resulting in slow diagnosis speed

Images