Hydroelectric generating set fault diagnosis method based on deep learning

Abstract

The invention relates to a hydroelectric generating set fault diagnosis method based on deep learning. The method comprises the following steps of firstly, acquiring a vibration signal when a hydroelectric generating set runs as a sample, and establishing a database; secondly, preprocessing data, namely reconstructing the original vibration signal; thirdly, segmenting the reconstructed data set into a training set, a verification set and a test set; fourthly, training a network combining a one-dimensional convolutional neural network (1-D CNN) and a gated recurrent unit (GRU), and optimizing network parameters to avoid network overfitting; and finally, establishing a hydroelectric generating set fault diagnosis model by utilizing the trained network parameters, and inputting test set samples into the model to realize the hydroelectric generating set fault diagnosis. According to the method, the accuracy of the hydroelectric generating set fault diagnosis can be improved.

Description

technical field [0001] The invention relates to the field of design of hydroelectric generating units, in particular to a method for fault diagnosis of hydroelectric generating units based on deep learning. Background technique [0002] Due to the increase in world population and the acceleration of industrialization, the rate of global energy consumption will become faster and faster. In addition, the consumption of traditional fossil energy will bring harm to the ecological environment, such as frequent occurrence of natural disasters and a series of disaster phenomena such as global warming. Therefore, the proposal of renewable energy is to protect the environment on which we live while developing. According to the global status report on renewable energy (REN212018), it is estimated that by the end of 2017, global renewable energy power production will account for 26.5% of the world's total power production, of which hydropower power production will account for 61.9% of...

AI Technical Summary

Problems solved by technology

However, traditional intelligent fault diagnosis methods can only provide limited fault diagnosis capabilities

It has some obvious shortcomings, which are summarized as follows: 1) Traditional intelligent fault diagnosis methods design and perform feature extraction and classification separately, which will affect the final diagnostic performance

This is a strategy that cannot be optimized at the same time

2) In the traditional intelligent fault diagnosis method, all the features extracted from the signal are manually made, which requires a lot of prior knowledge about signal processing technology and diagnostic expertise, and making these features is time-consuming and laborious

Images