Method for evaluating mechanical state of high-voltage shunt reactor based on vibration characteristics

Patent metadata

Field	Value
Publication number	US-12175352-B2
Application number	US-202117244947-A
Country	US
Kind code	B2
Filing date	Apr 29, 2021
Priority date	May 20, 2020
Publication date	Dec 24, 2024
Grant date	Dec 24, 2024

Abstract

A method for evaluating the mechanical state of a high-voltage shunt reactor based on vibration characteristics is disclosed, relating to the technical field of electrical equipment fault diagnosis. The method includes: based on historical state data and real-time vibration and noise signal data of the high-voltage shunt reactor and through an LSTM neural network time series prediction method, comparing deviation between predicted characteristic value and actual characteristic value, and determining whether the high-voltage shunt reactor has mechanical defects or failures. By using the historical state data and the real-time vibration and noise signal data of the high-voltage shunt reactor, an LSTM neural network time series prediction method, as well as comparison of the deviation between the predicted characteristic value and the actual characteristic value, etc., the evaluation of the mechanical state of the high-voltage shunt reactor is realized.

First claim

The invention claimed is:

1. A method for evaluating mechanical state of a high-voltage shunt reactor based on vibration characteristics, comprising: based on historical state data and real-time vibration and noise signal data of the high-voltage shunt reactor and through an LSTM neural network time series prediction model, comparing deviation between predicted characteristic value and actual characteristic value, and determining whether the high-voltage shunt reactor has mechanical defects or failures;
wherein the method specifically comprises: collecting vibration signals and noise signals on a surface of an oil tank of the high-voltage shunt reactor, extracting characteristic values in the vibration signals and noise signals and form a time series, combining with the time series prediction model to calculate a comprehensive deviation factor between the predicted characteristic value and the actual characteristic value, and determining whether mechanical defects or failures with unnatural trends occur inside the oil tank of the high-voltage shunt reactor by the comprehensive deviation factor;
wherein the method comprises: step S 1 : collecting vibration signals and noise signals of the high-voltage shunt reactor and performing data preprocessing; S 2 : establishing an LSTM neural network model and setting parameters; S 3 : predicting signal fundamental frequency amplitudes by using an LSTM neural network; S 4 : evaluating operating state of the high-voltage shunt reactor by using the comprehensive deviation factor;
wherein specifically steps of step S 2 comprise:
step S 201 : determining the number of neurons in an input layer, a hidden layer and an output layer; and step S 202 : constructing a prediction model;
in step S 201 : determining that the number of neurons of the input layer is 480, the output layer is 48, and the hidden layer is 24 in the neural network;
in step S 202 : constructing a neural network model in a form of multi-step training on the data, the steps have time-series relationship; in each iteration, using an output with current information as a portion of an input for a next time step;
wherein specifically steps of step S 1 comprise: step S 101 : arranging measuring points; step S 102 : collecting equipment; step S 103 : collection period; step S 104 : extracting characteristics from the signals; and step S 105 : filtering a characteristic sequence;
wherein steps of step S 4 comprise: step S 401 : calculating the comprehensive deviation factor; step S 402 : determining whether the comprehensive deviation factor is greater than a threshold, and step S 403 : alarming or repeating the above steps;
in step S 401 : the true characteristic sequence is defined as follows:
M={m 1 ,m 2 , . . . m i }  (11)

in Formula (11), M is a true characteristic sequence, in a form of a one-dimensional sequence; i is a value of the time corresponding to the sequence, and the time i increases every 30 minutes according to the collection period of the collected signal; m i is a value corresponding to the characteristic of the i -th time in the sequence M; each value in the sequence M is obtained by the filtering in step S 105 ;
comparing a predicted sequence F i ={f 1 , f 2 , . . . f i } with an actual sequence to be studied M i ={m 1 , m 2 , . . . m i }, and calculating the comprehensive deviation factor h; the comprehensive deviation factor is defined as follows:

h
=

∑

i
=
1

p

⁢



f
i

-

m
i

p
⁢

m
i



(
12
)

in Formula (12), h is the comprehensive deviation factor calculated from the predicted sequence {f 1 , f 2 , . . . f i } and a real sequence {m 1 , m 2 , . . . m i }, without unit and in the form of numerical value; i is subscript of the two sequences, without unit and in the form of numerical value, which are the same as those in Formula (10) and Formula (11); p is a sequence length, without unit and in the form of numerical value;
in step S 402 : if h is greater than an upper limit h max of the comprehensive deviation factor, it indicates that the characteristics of the vibration signal deviate greatly from an ideal value, and internal fastening components of the high-voltage shunt reactor have a certain defect and an alarm is issued; a recommended value range for h max is [0.05, 0.15];
in step S 403 : if the comprehensive deviation factor h is less than the upper limit h max of the comprehensive deviation factor, continuing to collect data, and performing the step S 1 based on the current data.

2. The method for evaluating mechanical state of a high-voltage shunt reactor based on vibration characteristics according to claim 1 , comprising: using fundamental frequency amplitudes of the vibration signals and the noise signals on the surface of the oil tank of the high-voltage shunt reactor as characteristic variables.

3. The method for evaluating mechanical state of a high-voltage shunt reactor based on vibration characteristics according to claim 1 , comprising: taking 30 minutes as a sampling period to collect the vibration signals and the noise signals on the surface of the oil tank of the high-voltage shunt reactor.

4. The method for evaluating mechanical state of a high-voltage shunt reactor based on vibration characteristics according to claim 1 , comprising: predicting vibration characteristic values of the high-voltage shunt reactor for a period of time in the future through the LSTM neural network.

5. The method for evaluating mechanical state of a high-voltage shunt reactor based on vibration characteristics according to claim 1 , comprising: calculating, based on the predicted characteristic value obtained from the prediction of the vibration characteristics of through the LSTM neural network, a comprehensive deviation factor between the predicted characteristic value and the actual characteristic value, and using the comprehensive deviation factor as a state evaluation index.

6. The method for evaluating mechanical state of a high-voltage shunt reactor based on vibration characteristics according to claim 1 , wherein specific steps of step S 1 comprise: step S 101 : arranging measuring points; step S 102 : collecting equipment; step S 103 : collection period; step S 104 : extracting characteristics from the signals; and step S 105 : filtering a characteristic sequence;
specific steps of step S 3 comprise: step S 301 : inputting sequence; step S 302 :

Assignees

• State Grid Hebei Electric Power Res Institute
• State Grid Corp China
• Univ Xi An Jiaotong

Inventors

• Gao Shuguo
• Meng Lingming
• Zeng siming
• Fan Hui
• Zang Qian
• Jia Boyan
• Ji Shengchang
• Liu Hongliang
• Sun Lu
• Xing Chao
• Zhao Jun

Classifications

• G06N3/0442 Primary

  characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU] · CPC title

• G06N3/09

  Supervised learning · CPC title

• G01H1/00

  Measuring {characteristics of} vibrations in solids by using direct conduction to the detector (G01H9/00, G01H11/00 take precedence) · CPC title

• G06N3/048

  Activation functions · CPC title

• G06N3/044

  Recurrent networks, e.g. Hopfield networks · CPC title