An intrinsic state quantity-based electric drive system state perception and evaluation method

By using a bidirectional long short-time neural network model and hierarchical analysis, the problem of rapid and accurate state assessment of electric drive systems was solved, enabling real-time online monitoring and fault early warning of motor status, thus improving detection efficiency and equipment reliability.

CN118734034BActive Publication Date: 2025-12-16ANHUI UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410779437.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-06-17
Publication Date
2025-12-16
Estimated Expiration
2044-06-17

AI Technical Summary

Technical Problem

Existing technologies struggle to quickly and accurately detect early or uncommon faults in electric drive system condition assessments, and require extensive data analysis and expertise, with limited sensitivity and specificity.

Method used

A bidirectional long-short-time neural network model is adopted to construct a current prediction model through speed, voltage, and current signals. Combined with the analytic hierarchy process and Euclidean distance calculation, the online monitoring and evaluation of motor status can be realized, reducing the dependence on sensors and manual intervention.

Benefits of technology

It enables real-time assessment of motor status and fault early warning, improves detection rate and response speed, reduces cost and maintenance difficulty, and improves equipment reliability and maintenance efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118734034B_ABST
    Figure CN118734034B_ABST
Patent Text Reader

Abstract

The application discloses a kind of based on intrinsic state quantity electric drive system state perception and evaluation method, belong to electric drive system detection technical field, the method includes the following steps: step one, select rotational speed, voltage, current as the signal of describing electric drive system operating state, construct bidirectional long short time neural network, and determine its input data and output data;The application does not need artificial intervention and shutdown maintenance, can realize on-line monitoring and real-time evaluation to motor state, improve the detection rate and response speed of motor fault;The application does not depend on additional sensor or measuring device, by neural network model to the analysis of intrinsic state quantity, reduce cost and maintenance difficulty.Electric drive system state perception and evaluation method based on neural network can in-depth analysis the state change in motor, by training and optimizing neural network model, can realize early warning and diagnosis to motor fault, improve maintenance efficiency and equipment reliability.
Need to check novelty before this filing date? Find Prior Art

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of electric drive system detection, and particularly relates to an electric drive system state perception and evaluation method based on intrinsic state variables. BACKGROUND

[0002] Electric drive systems are widely used in industrial automation, mechanical manufacturing, transportation and other fields. In some high-performance working conditions, accurate evaluation of the state of the electric drive system is of great significance for maintaining equipment reliability and improving production efficiency. Traditional electric motor system state evaluation methods mainly include: 1) parameter statistical method, which is a method of collecting and analyzing parameter data (such as current, voltage, speed, etc.) during motor operation, using statistical indicators such as mean, variance, etc. to monitor and evaluate the state of the motor. By comparing the actual parameters with the statistical indicators of the preset range, it is determined whether the motor is working normally or has potential problems; 2) fault diagnosis method, which judges the fault type and degree of the motor by monitoring the fault characteristics during the operation of the motor. These fault characteristics include abnormal noise, vibration, temperature rise, etc. Sensors and monitoring equipment are needed to collect data in real time, and analysis and comparison are needed to identify faults; 3) fault pattern recognition method, which uses statistical methods and machine learning techniques to establish a correlation model between motor fault patterns and parameters, and evaluates the state of the motor by pattern recognition on real-time data. However, these methods usually require a large amount of data for analysis for accurate fault diagnosis and state evaluation. For some early or uncommon faults, these traditional methods have limitations in sensitivity and specificity, and cannot detect or accurately diagnose in time. Moreover, technical personnel need to use professional knowledge and experience to analyze and interpret the data. Therefore, the present application proposes an electric drive system state perception and evaluation method based on intrinsic state variables, which can accurately evaluate the state of the motor without monitoring a large number of parameters or constructing all fault models. SUMMARY

[0003] Therefore, the present application proposes an electric drive system state perception and evaluation method based on intrinsic state variables, which solves the above problems.

[0004] To achieve the above purpose, the first aspect of the present application provides an electric drive system state perception and evaluation method based on intrinsic state variables, comprising the following steps:

[0005] Step one, select speed, voltage and current as signals to describe the running state of the electric drive system, construct a bidirectional long short-term neural network, and determine the input data and output data thereof;

[0006] Step two, train the bidirectional long short-term neural network with data in normal running state to obtain an electric current prediction model based on the bidirectional long short-term neural network;

[0007] Step three, constructing an analysis index according to the output data of the current prediction model, the actual output data and the residual data;

[0008] And based on the analytic hierarchy process, the weight of the analysis index is calculated, and then the ideal operating state vector v1 is obtained;

[0009] Step four, collecting the current data of the electric drive system, inputting to the current prediction model to obtain the current prediction index, and based on the current prediction index, constructing the current operating state vector v2;

[0010] Step five, calculating the distance D between v1 and v2 based on the Euclidean distance, and then constructing the mathematical relationship between the distance value and the motor evaluation score, obtaining the current operating state score of the electric drive system, realizing the situation awareness and state evaluation of the electric drive system.

[0011] Preferably, the input data in step one includes speed, d-axis voltage and q-axis voltage; the output data includes d-axis current and q-axis current.

[0012] Preferably, the analysis index in step three includes peak error, difference error and absolute error.

[0013] Preferably, in step four, the current operating state vector v2 is constructed by processing the current prediction index based on the mathematical model of the index and the score and the weight of the index value.

[0014] Preferably, the evaluation score of the electric drive system state is Score = 100-D.

[0015] Compared with the prior art, the beneficial effects of the present application are: the present application can realize online monitoring and real-time evaluation of the motor state without manual intervention and shutdown maintenance, which improves the detection rate and response speed of motor fault; the present application does not depend on additional sensors or measuring devices, and the intrinsic state quantity is analyzed by a neural network model, which reduces the cost and maintenance difficulty. The electric drive system state perception and evaluation method based on neural network can deeply analyze the state change of the motor, and through training and optimization of the neural network model, the early warning and diagnosis of motor fault can be realized, which improves the maintenance efficiency and equipment reliability. BRIEF DESCRIPTION OF DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiment or prior art description will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application, and those skilled in the art can obtain other drawings according to these drawings without creative labor.

[0017] Figure 1The state perception and evaluation process of the present application;

[0018] Figure 2 The BiLSTM neural network model structure of the present application;

[0019] Figure 3 The mathematical relationship model of the error index and state score in the present application;

[0020] Figure 4 The AHP flowchart in the present application. DETAILED DESCRIPTION

[0021] The technical solutions of the present application will be described in detail below in conjunction with the embodiments. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0022] Please refer to Figures 1-4 A state perception and evaluation method for an electric drive system based on intrinsic state variables, comprising the following steps:

[0023] Step 1: Selecting rotational speed, voltage and current as signals for describing the running state of the electric drive system, constructing a bidirectional long-short time neural network, and determining the input data (rotational speed, d-axis voltage, q-axis voltage) and output data (d-axis current, q-axis current) thereof;

[0024] The voltage equation of a three-phase permanent magnet synchronous motor in a synchronous rotating coordinate system is

[0025]

[0026] ω r =pω m (2)

[0027] ω m =2πn (3)

[0028] In the formula: u d , u q are the d-axis and q-axis stator voltages, i d , i q are the d-axis and q-axis stator currents, ω r is the rotor electric angular velocity, L d , L q are the d-axis and q-axis stator inductances, D is the differential operator, ψ f is the rotor permanent magnet flux linkage, R s is the stator resistance, p is the number of motor pole pairs, ω m is the rotor angular velocity, and n is the motor rotor speed.

[0029] From the above formula, the dq-axis current of the permanent magnet synchronous three-phase motor has a certain mathematical relationship with the dq-axis voltage and the speed. Therefore, the dq-axis voltage signal and the speed signal of the motor are selected as the input data of the neural network, and the dq-axis current signal of the motor is selected as the output data.

[0030] Step two: train the bidirectional long and short time neural network using data in the normal operating state to obtain a current prediction model based on the bidirectional long and short time neural network;

[0031] Step three: construct three indexes (peak error, difference error, and absolute error) based on the output data of the current prediction model, the actual output data, and the residual data, calculate the weights of the three indexes based on the analytic hierarchy process, and then obtain the ideal operating state vector (v1) according to the above weights;

[0032] (1) The peak error is the average value of the relative difference between the peak and valley of the predicted current curve and the actual current curve, which is defined as follows:

[0033]

[0034] (2) The difference error is the average value of the peak error of S'(t) and S(t) relative to the peak value of S(t), which is defined as follows:

[0035]

[0036] (3) The absolute error is defined as the ratio of S'(t) and S(t) to S(t), which is defined as follows:

[0037]

[0038] The weights of the three are calculated based on the analytic hierarchy process, as shown in the flowchart. The weight vector of the indexes is [a1 a2 a3], and the CR is calculated to pass the consistency test. The ideal vector (v1) is calculated based on the above weights: Figure 4

[0039] v1=[100×a1100×a2100×a3] (8)

[0040] Step four: collect the current data of the electric drive system, input it to the current prediction model to obtain three index values, and process the index values based on the mathematical model of the index and the score and the weights to construct the current operating state vector (v2);

[0041] The scores of the three evaluation indexes [S1 S2 S3] are calculated according to the evaluation index scores, as shown in the mathematical relationship model of the error and the state score. Figure 3

[0042] ​​According to the above weight calculation of the current running state vector (v2):

[0043] v2 = [a1xS1 a2xS2 a3xS3] (9)

[0044] Step five: calculate the distance between the current running state vector and the ideal vector based on the Euclidean distance, and then build a mathematical relationship between the distance value and the motor evaluation score to obtain the current running state score of the electric drive system, and realize the situation awareness and state evaluation of the electric drive system.

[0045] Score = 100-D (11)

[0046] Finally, the evaluation score (S) of the electric drive system state is obtained, wherein the range of S is 0 to 100. The larger S is, that is, the smaller the distance between the sample point and the ideal characteristic point is, indicating that the electric drive system is closer to the normal operating state. The smaller S is, that is, the larger the distance between the sample point and the ideal characteristic point is, indicating that the electric drive system has a greater possibility of abnormal situation or failure.

[0047] Some data in the above formula is calculated by removing the dimension, and the formula is obtained by software simulation of a large amount of collected data to be closest to the real situation; the preset parameters and the preset threshold in the formula are set by the person skilled in the art according to the actual situation or obtained by a large amount of data simulation.

[0048] The above embodiments are only used to illustrate the technical method of the present application and are not limited. Although the present application has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical method of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical method of the present application.

Claims

1. A method for state perception and evaluation of an electric drive system based on intrinsic state variables, characterized in that, Includes the following steps: Step 1: Select speed, voltage, and current as signals to describe the operating state of the electric drive system, construct a bidirectional long short-time neural network, and determine its input and output data; Step 2: Train the bidirectional long short-time neural network using data from normal operation to obtain a current prediction model based on the bidirectional long short-time neural network; Step 3: Based on the output data of the current prediction model, the actual output data, and the residual data, construct three indicators: peak error, difference error, and absolute error. Calculate the weights of these three indicators using the analytic hierarchy process (AHP), and then obtain the ideal operating state vector based on these weights. ; Peak error is the average of the relative differences between the peaks and troughs of the predicted current curve and the actual current curve, defined as follows: The difference error is the average error between the peak values ​​of S'(t) and S(t) and the peak value of S(t), defined as follows: The absolute error is defined as the ratio of S'(t) to S(t), and is given by the following formula: The weights of the three factors are calculated using the analytic hierarchy process (AHP), resulting in the following weight vector for the indicators: Furthermore, the CR calculation passes the consistency test, and the ideal operating state vector is calculated based on the aforementioned weights. : Step 4: Collect current data from the electric drive system, input it into the current prediction model to obtain three index values, and process the index values ​​based on the mathematical model of the indexes and scores and their weights to construct the current operating state vector. ; Calculate the scores for the three evaluation indicators based on their scores. ; Calculate the current running state vector based on the above weights. : Step 5: Calculation based on Euclidean distance and distance D This allows for the construction of a mathematical relationship between distance values ​​and motor evaluation scores, obtaining the current operating status score of the electric drive system, and realizing situational awareness and status assessment of the electric drive system.

2. The method for state perception and evaluation of an electric drive system based on intrinsic state variables according to claim 1, characterized in that: The input data in step one includes rotational speed, d-axis voltage, and q-axis voltage; the output data includes d-axis current and q-axis current.

3. The method for state perception and evaluation of an electric drive system based on intrinsic state variables according to claim 2, characterized in that: The evaluation score of the electric drive system status .