CN112327200B - Health characterization and health baseline construction method of DC-AC inverter circuit - Google Patents
Health characterization and health baseline construction method of DC-AC inverter circuit Download PDFInfo
- Publication number
- CN112327200B CN112327200B CN202011136503.3A CN202011136503A CN112327200B CN 112327200 B CN112327200 B CN 112327200B CN 202011136503 A CN202011136503 A CN 202011136503A CN 112327200 B CN112327200 B CN 112327200B
- Authority
- CN
- China
- Prior art keywords
- inverter circuit
- health
- parameters
- characteristic
- characteristic parameters
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/40—Testing power supplies
- G01R31/42—AC power supplies
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/24—Querying
- G06F16/245—Query processing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/067—Enterprise or organisation modelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
Abstract
The invention discloses a health characterization and health baseline construction method of a DC-AC inverter circuit, which comprises the following steps: extracting time domain, frequency domain, time-frequency domain signals of each measuring point in the DC-AC inverter circuit and each performance index of the DC-AC inverter circuit as characteristic parameters; bringing the extracted characteristic parameters into a candidate fault characteristic library; optimizing the characteristic parameters in the fault characteristic library, and taking the finally selected characteristic parameters as health characterization parameters; the method comprises the steps of mapping health characterization parameters of a DC-AC inverter circuit under actual working conditions to standard working conditions to obtain fault characteristic parameters; and constructing a Mahalanobis space by calculating the Mahalanobis distance between the fault characteristic parameter and the healthy sample, and taking the Mahalanobis space as a healthy baseline of the DC-AC inverter circuit. The health characterization and the health baseline constructed by the method finally provide technical support for the flight safety of the airplane.
Description
Technical Field
The invention belongs to the field of airplane fault diagnosis and health management, and particularly relates to a method for representing the health of a DC-AC inverter circuit and constructing a health baseline.
Background
In the field of aviation, a DC-AC inverter circuit is an airplane secondary power supply device, belongs to a typical high-power analog circuit of a power tube switch control working mode, and plays an important role in improving the safety of an airplane. Power electronic circuit faults are classified in severity into hard faults and soft faults. Hard faults are also called sudden faults or complete faults, namely parameters of components in a circuit are changed suddenly and greatly, and the hard faults often cause the loss of circuit functions, mainly comprise short-circuit faults and open-circuit faults. Soft faults are also called gradual faults or partial faults, namely parameter values of components in a circuit deviate to an unallowable extent along with time or environmental conditions and exceed the tolerance range of the parameters of the components, and the functions of the components are not completely lost but only change circuit performance indexes. Because a hard fault generally causes a circuit topology to change, the circuit cannot work, and the finding and diagnosis are simple, the invention mainly researches soft fault diagnosis of the DC-AC inverter circuit, wherein the DC-AC inverter circuit topology is as shown in FIG. 1, and FIGS. 2 and 3 are a voltage-doubling flyback DC-DC booster circuit and a DC-AC inverter circuit respectively.
When the DC-AC inverter circuit has soft faults, effective acquisition processing is carried out based on the change of corresponding characteristic indexes, so that fault characteristics are accurately extracted for analysis, corresponding health characteristics and health baselines are constructed, corresponding technical support is provided for health assessment of the airplane, and the using safety of the airplane is finally ensured.
Disclosure of Invention
The invention aims to provide a method for constructing health representation and a health baseline of a DC-AC inverter circuit, when the aviation DC-AC inverter circuit has soft faults, the circuit state changes, and related characteristic parameters can be accurately acquired and compared with the constructed health representation and health baseline, so that the health state of the circuit is judged, and technical support is provided for aircraft operators and maintenance personnel.
The invention aims to be realized by the following technical scheme:
a health characterization and health baseline construction method for a DC-AC inverter circuit comprises the following steps:
the method comprises the following steps: extracting a time domain signal and a frequency domain signal of a detection signal of each measuring point in the DC-AC inverter circuit by utilizing characteristic analysis as characteristic parameters;
step two: performing wavelet packet decomposition on detection signals of each measuring point in the DC-AC inverter circuit, and extracting signal energy from low frequency to high frequency to be used as characteristic parameters;
step three: calculating each performance index of the DC-AC inverter circuit as a characteristic parameter;
step four: the characteristic parameters extracted in the first step, the second step and the third step are brought into a candidate fault characteristic library;
step five: optimizing the characteristic parameters in the fault characteristic library, and taking the finally selected characteristic parameters as health characterization parameters;
step six: the method comprises the steps of mapping health characterization parameters of a DC-AC inverter circuit under actual working conditions to standard working conditions to obtain fault characteristic parameters;
step seven: and constructing a Mahalanobis space by calculating the Mahalanobis distance between the fault characteristic parameter and the healthy sample, and taking the Mahalanobis space as a healthy baseline of the DC-AC inverter circuit.
Preferably, the first step further includes determining whether the operating condition of the DC-AC inverter circuit is idealized by using the relative variation of the characteristic parameter, if not ideal, establishing regression modeling of the operating condition and the characteristic parameter, and inputting the characteristic parameter into the regression modeling to obtain an idealized characteristic parameter: wherein, the relative variation of the characteristic parameters is as follows:
θ 0 the characteristic parameter is a characteristic parameter of the DC-AC inverter circuit at an initial time or in a healthy state, θ is a characteristic parameter of the DC-AC inverter circuit after accumulated operation for a period of time, and Δ θ is ideally equal to 0.
Preferably, step two comprises the steps of:
(1) performing n layers of wavelet packet decomposition on the time sequence signal X (n) of the detection signal x, and extracting wavelet decomposition coefficients of the nth layer of each frequency band from low frequency to high frequency, wherein
Represents the jth component of the ith layer;
(2) reconstructing the wavelet packet decomposition coefficient to obtain a reconstructed signal;
(3) and calculating the total energy of the reconstructed signals of each frequency band to form a signal vector.
Preferably, the step five is specifically: and optimizing the characteristic parameters in the candidate fault characteristic library by adopting a multi-evaluation index characteristic optimization model, selecting the characteristic parameters of the first 7 ranked in each single fault mode, then calculating the spearman correlation coefficient for further screening, and finally selecting 7 characteristic parameters as health characterization parameters.
Preferably, the multi-evaluation index feature optimization model adopts three evaluation indexes of trend, monotonicity and robustness to optimize feature parameters in the candidate fault feature library.
Preferably, in the sixth step, an equivalent weight analysis method is adopted to map the health characterization parameters of the DC-AC inverter circuit under the actual working condition to the standard working condition to obtain the fault characteristic parameters.
Preferably, the seventh step further comprises determining the baseline threshold according to whether the probability distribution of mahalanobis distance of the healthy sample satisfies normality: (1) if the Mahalanobis distance of the healthy sample meets the normal distribution, determining a baseline threshold value directly according to the 3 sigma criterion of the normal distribution; (2) if the Mahalanobis distance of the healthy sample does not meet normal distribution, BOX-COX transformation is required to be performed firstly to enable the healthy sample to meet the normal distribution, and finally a baseline threshold is determined according to a 3 sigma criterion.
The invention has the beneficial effects that: when the DC-AC inverter circuit has soft faults, the circuit state changes, and related characteristic parameters can be accurately acquired and compared with the established health characterization and health baseline, so that the health state of the circuit is judged, and technical support is provided for aircraft operators and maintenance personnel.
Drawings
Fig. 1 is a DC-AC inverter circuit topology.
Fig. 2 is a voltage doubling flyback DC-DC booster circuit.
Fig. 3 shows a DC-AC inverter circuit.
Fig. 4 is a test point of the DC-DC flyback boost main circuit.
Fig. 5 is a test point of the DC-AC full bridge inversion main circuit.
Fig. 6 is a flow chart of feature parameter extraction when the operating condition changes.
Fig. 7 is a structural diagram of wavelet packet decomposition.
FIG. 8 is a flow chart of equivalent weight analysis.
Fig. 9 is a flow chart of a healthy baseline configuration.
Fig. 10 is a baseline characterization of DC-AC inverter circuit health under different operating conditions.
Fig. 11 is a schematic flow chart diagram of a DC-AC inverter circuit health characterization and health baseline construction method.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings and examples.
Referring to fig. 11, in order to accurately and effectively construct a health characteristic and a health baseline of an aviation DC-AC inverter circuit, in this embodiment, data acquisition and analysis are performed on an aviation DC-AC inverter circuit when the circuit changes under different working conditions, and a DC-AC inverter circuit fault characteristic analysis is performed on a detected circuit state signal, where the data acquisition and analysis includes time domain and frequency domain characteristic parameters of a detected signal, a parameter relative change characteristic value, and a wavelet transform-based time-frequency domain characteristic parameter. Firstly, forming a candidate fault feature library according to feature parameters extracted by a fault feature extraction method; then, optimizing the characteristic parameters of the candidate fault characteristic library by adopting a multi-evaluation index characteristic optimization model, selecting the characteristic parameters of the first 7 in the sequence under each single fault mode, and calculating the spearman correlation coefficient for further screening, thereby determining the final 7 parameters for representing the health state of the whole circuit; then, an equivalent weight analysis method is adopted to eliminate the influence of working conditions on health characterization parameters, because the health characterization parameters of the circuit change to different degrees in the process from normal to fault, contain component parameter degradation fault information, and are influenced by the working condition conditions (such as input voltage and load); and finally constructing a Mahalanobis space by calculating the Mahalanobis distance of the healthy sample characteristic sequence, and using the Mahalanobis space as a healthy baseline of the DC-AC inverter circuit.
The method for health characterization and health baseline construction of a DC-AC inverter circuit shown in this embodiment specifically includes the following steps:
the method comprises the following steps: and extracting a time domain signal and a frequency domain signal of the detection signal x of each measuring point in the DC-AC inverter circuit as characteristic parameters by utilizing characteristic analysis. By way of example, the time domain signal may be represented by any of the following characteristics:
(1) m-order moment:
where N denotes the time-series length of the detection signal x, m is a positive integer, and m-1 and m-2, whose moments are the average value and the mean square, respectively, and both < > and E denote averaging.
Average value m is 1
Mean square value m is 2
(2) Moment of deviation
(ii) variance m is 2
Third order moment of deviation
Third four order deviation moments
Standard deviation or root mean square deviation (Standard deviation):
skewness degree
Skewness is a measure of the direction and extent of skew of a statistical data distribution. The frequency distribution of the statistical data is symmetrical, and is asymmetrical, namely, shows a bias state. In a skewed distribution, when the skewness is positive, the distribution is positively skewed, i.e., the mode is to the left of the arithmetic mean; when the skewness is negative, the distribution is negatively biased, i.e., the mode is to the right of the arithmetic mean. We can use the relationship between the mode, median and arithmetic mean to determine whether the distribution is left-biased or right-biased, but to measure the degree of distribution skewness, we need to calculate skewness. This amount represents the degree of upper and lower asymmetry of the observed data (time series) from the mean, and also the degree of asymmetry (inconsistency) or shape factor of the rise time and fall time of the X-t curve.
In the formula, σ 3 Is to make gamma 1 Becomes the amount of dimension 1.
Sixthly, Kurtosis
When gamma is 2 3 is defined as a profile with normal kurtosis (i.e., zero kurtosis); when gamma is 2 >At 3, the profile has a positive kurtosis. As can be seen from equation (10), when the standard deviation σ is smaller than the standard deviation in the normal state, that is, the degree of dispersion of the observed values is small, γ 2 The height of the peak top of the normal distribution curve is increased and is higher than that of the normal distribution curve, so that the degree of the kurtosis is called as the kurtosis. When gamma is 2 <When 3, the distribution curve has a negative kurtosis, and it can be seen from equation (10) that when the standard deviation σ is larger than the standard deviation in the normal state, i.e., the degree of dispersion of the observed values is large, γ is 2 The height of the peak top of the normal distribution curve is reduced and is called negative kurtosis.
(3) Other correlated time domain signals
Form factor (SF):
crest factor (crest factor, CF):
pulse factor (IF):
mean amplitude (absolute mean, PF):
square root amplitude (Fg):
margin factor (LF):
in this project, each measurement point is selected as shown in fig. 4 and 5, and the time domain signal and the frequency domain signal (mean, peak-to-peak value, variance, standard deviation, skewness, and kurtosis) of each measurement point should be analyzed by selecting as many measurement points as possible. Wherein the frequency domain signal is obtained by performing a Fast Fourier Transform (FFT) on the time domain signal. The change of the time domain signal and the frequency domain signal of the DC-AC inverter circuit is directly extracted to be used as fault diagnosis and fault prediction, and two problems are ignored:
(1) the change rate of each circuit characteristic parameter is different from normal to fault
(2) The influence of the operating conditions on the characteristic parameters is not taken into account.
The parameter relative variation is used as a new circuit fault evaluation index and is defined as follows: under the same working condition, the characteristic parameter at the initial moment (in health) of the circuit is theta 0 If the characteristic parameter is θ after the circuit is operated for a period of time, the relative variation of the characteristic parameter is:
because each DC-AC inverter circuit has strong nonlinearity, different circuit topological structures and different types of components, the relationship between the circuit parameters and the component parameters is difficult to directly establish. The relative variation of the characteristic parameters does not need to establish the relationship between the circuit parameters and the component parameters, and can reflect the health condition of the circuit in the state. When the circuit component parameter is a health value, Δ θ is 0, which is not affected by the circuit operating condition and is only related to the health condition of the circuit itself.
The flow chart of extracting the fault characteristic parameters when the working conditions are changed is shown as 6, and the key problems to be solved by determining the relative variation of the circuit characteristic parameters are as follows: firstly, determining characteristic parameters of a DC-AC inverter circuit; secondly, how to predict the relative variation of the characteristic parameters of the DC-AC inverter circuit at a certain future moment is specifically how to obtain theta and theta 0 . The implementation steps are divided into two steps:
step 1: obtaining theta 0
Setting parameters of circuit components to initial time values (health values) and keeping the initial time values unchanged, and simultaneously respectively setting different working conditions (V) i 、i o )。
Circuit parameters (output ripple voltage and output voltage ripple ratio) of the circuit under various working conditions are obtained and used as training samples of LSSVM regression fitting, and a working condition and characteristic parameter regression modeling (health model) is established.
③ working conditions (V) at any time i 、i o ) As the input of the trained LSSVM, the output of the LSSVM is the circuit parameter (output ripple voltage and output voltage ripple ratio) with small corresponding working condition, namely theta when the circuit is healthy 0 。
Step 2: obtaining theta
Firstly, according to the time-dependent degradation rule of slow-varying type faults of all components of the circuit, the change trend of parameters of the components along with time is set, and meanwhile, the fluctuation change of working conditions is set, so that circuit simulation is carried out.
② selecting V i 、V out And i o And acquiring corresponding voltage and current waveform data as monitoring signals, and calculating a circuit parameter, namely theta.
Fourthly, the circuit parameter at a certain moment in the circuit degradation process is obtained, and theta under the same working condition is obtained according to the step 1 0 From equation (17), the relative change amount Δ θ of the characteristic parameter can be obtained as 0.
Step two: wavelet packet decomposition is performed on a detection signal x at each measurement point in the DC-AC inverter circuit, and signal energy from a low frequency to a high frequency is extracted as a characteristic parameter. The detailed steps are as follows:
(1) performing n-layer wavelet packet decomposition on the time series signal X (n) of the detection signal x, and extracting wavelet decomposition coefficients of the nth layer from low frequency to high frequency in each frequency band, wherein the decomposition structure is shown in FIG. 7
Represents the jth component of the ith layer;
(2) and reconstructing the wavelet packet decomposition coefficient. Decomposing n layers of wavelet packet to obtain 2 n And reconstructing the sequences in the frequency bands to obtain 2n reconstructed signals.
(3) Calculating the total energy of the signals of each frequency band to form a signal vector [ E 1 ,E 2 ,…,E j ]. The formula is as follows:
for an uncertainty system, if a random variable X with a finite value is used to represent the state characteristics, the value is X j Has a probability of P j ＝P{X＝X j 1,2, …, L, and
W j ＝(w j1 ,w j2 ,…w jn ) Then the information obtained from a certain result of X can be used as I j ＝log(1/P j ) Denotes that the information entropy of X is
When P is present j When equal to 0, P j log(P j ) The entropy H is an information measure of the positioning system under certain conditions, and is a measure of the degree of sequence unknowns, which can be used to estimate the complexity of the signal. Performing wavelet packet decomposition on the signal, and obtaining the total energy of the signal
According to the characteristics of orthogonal wavelet transformation, the total energy E of a signal is equal to the energy E of each component in a certain time i And (4) summing. If it is
Then
Entropy of the wavelet of
Meanwhile, parameters such as average frequency (MF), frequency center (center of gravity frequency: FC), frequency Root Mean Square (RMSF), frequency standard deviation (RVF) and the like are used as one of characteristic parameters in frequency domain characteristic analysis. Its definition is as follows:
mean Frequency (MF):
frequency center (center of gravity frequency, FC):
root Mean Square Frequency (RMSF):
standard deviation of frequency (RVF):
step three: and calculating each performance index of the DC-AC inverter circuit as a characteristic parameter. With input voltage signal V of DC-DC conversion circuit i And an output voltage signal V out Input current signal i 1 Output current signal i o For example, the calculation method of the DC-DC inverter performance index is described as shown in table 1 and table 2. Wherein N is the number of sampling points, u o (k) Sampling a kth output voltage for the circuit; u. of o (max) is the maximum value in the output voltage sample signal; u. of o (min) is the minimum value in the output voltage sampling signal; i.e. i i (k) Sampling a kth input current for the circuit; i.e. i o (k) The kth output current sampled for the circuit.
TABLE 1 Circuit-level Performance index parameters for DC-DC converters
TABLE 2 Circuit-level Performance index parameters for DC-AC converters
Step four: and (4) bringing the feature parameters extracted in the first step, the second step and the third step into a candidate fault feature library. In the present embodiment, 45 characteristic parameters are extracted, wherein 33 characteristic parameters of the inverter circuit state signal exist, and 12 performance indexes exist. As shown in table 3.
TABLE 3 library of candidate fault signatures
Step five: and optimizing the characteristic parameters in the candidate fault characteristic library by adopting a multi-evaluation index characteristic optimization model, selecting the characteristic parameters of the first 7 ranked characteristic parameters in each single fault mode, calculating the spearman correlation coefficient for further screening, and finally selecting 7 characteristic parameters as a health characterization parameter map for characterizing the health state of the whole circuit.
This example illustrates that 3 evaluation indexes are used for optimization, and the multi-evaluation index feature optimization model is represented by formula (23):
wherein, W is a comprehensive evaluation index and has a value range of [0,1]]，w i Tre (g) is trend, Mon (g) is monotonicity, rob (g) is robustness, and the specific calculation formula is shown in formulas (24) to (27).
(1) Tendency of
Wherein F ═ F 1 ,f 2 ,…,f N ) For the signature sequence, N is the total number of samples, and T ═ T 1 ,t 2 ,…,t N ) For the corresponding sample sequence, t i The characteristic value corresponding to the time is f i 。
The calculated output value of the trend index is [0,1], and the higher the trend is, the higher the linear correlation degree of the characteristic sequence and the corresponding sampling sequence is.
(2) Monotonicity
Wherein, the expression of the unit order function delta (x) is:
the output value of the monotonicity index is [0,1], and the monotonicity index reflects the strength of the characteristic monotonously increasing or monotonously decreasing. A larger monotonicity of a feature indicates that the feature has a more stringent monotonically increasing or decreasing trend.
(3) Robustness
Wherein, the first and the second end of the pipe are connected with each other,
is characterized by i And (4) obtaining a trend part through smoothing processing.
The calculation output range of the robustness index is [0,1]]And is used for measuring the volatility of the characteristic sequence. The more severe the fluctuation of the feature sequence or the more singular points, the weaker the robustness. Selecting DC-AC inverter circuit health characterization parameters by a multi-evaluation index characteristic parameter selection model, selecting the characteristic parameters of the first 7 in the sequence under each single fault mode through comprehensive evaluation, then calculating the spearman correlation coefficient for further screening, and finally selecting 7 characteristic parameters for characterizing the health state of the whole circuit, for example, V q201 The square root amplitude is a square root amplitude of the voltage of the measurement component Q201.
TABLE 4 DC-AC INVERTER CIRCUIT HEALTH CHARACTERIZATION PARAMETERS
| V d102 Square root amplitude | DC output voltage ripple |
| V 4 Maximum value | V q201 Square root amplitude |
| I acsample Frequency of center of gravity | V dc Entropy of wavelet energy |
| V d103 Square root amplitude |
Step six: and mapping the health characterization parameters of the DC-AC inverter circuit under the actual working condition to be used as fault characterization parameters under the standard working condition.
In this embodiment, for example, the equivalent weight analysis method is selected to map the health characterization parameters of the DC-AC inverter circuit under the actual operating conditions to the standard operating conditions. The method specifically comprises the following steps: in the no-fault mode, 6 different operating conditions are considered as shown in table 5:
TABLE 5 operating condition mode Table
The health characterization parameters under the circuit working condition and the non-standard working condition are used as input, the health characterization parameters under the standard working condition (such as +28V input and 200VA output) are used as output to train the neural network model, the relation between the system input and the system output is fitted, the function expression is shown as the formula (28), and the flow of performing equivalent weight analysis is shown as the figure 8.
Wherein u is in 、R L Is the actual input voltage, load resistance, c i I is 1,2, …, n is a health characterization parameter under actual working conditions,
is under standard working conditionsA health characterization parameter.
The method comprises the following specific steps:
selecting health characterization parameters capable of reflecting the performance degradation rule of the circuit when the working condition is not changed;
setting the variation trend of device parameters along with time according to the time-dependent degradation rule of each component of the DC-AC inverter circuit, simultaneously setting the fluctuation change of working conditions, performing circuit simulation, and calculating and obtaining health characterization parameters under non-standard working conditions;
setting the variation trend of the same device parameters with time as the step two, performing circuit simulation when the circuit works under the standard working condition, and calculating and acquiring health characterization parameters under the standard working condition;
selecting an intelligent algorithm, training a fitting model based on the acquired nonstandard working condition, the health characterization parameters under the nonstandard working condition and the health characterization parameters under the standard working condition, and establishing a relation model of the health characterization parameters under the standard working condition, the actual working condition and the health characterization parameters under the actual working condition;
fifthly, inputting any actual working condition and corresponding health characterization parameter value into the trained fitting model in the fourth step, wherein the calculation output of the fitting model is the equivalent value of the health characterization parameter under the standard working condition, namely the fault characteristic parameter of the circuit.
Step seven: and constructing a Mahalanobis space, namely a normal reference space, by calculating the Mahalanobis distance between the fault characteristic parameter and the healthy sample, and taking the Mahalanobis space as a healthy baseline of the DC-AC inverter circuit.
The baseline threshold can be determined in two cases according to whether the probability distribution of mahalanobis distance of the healthy sample satisfies normality: (1) and if the Mahalanobis distance of the healthy sample meets the normal distribution, determining a baseline threshold value directly according to the 3 sigma criterion of the normal distribution. (2) If the Mahalanobis distance of the healthy sample does not meet normal distribution, BOX-COX transformation is required to be performed firstly to enable the healthy sample to meet the normal distribution, and finally, a baseline threshold value is determined according to a 3 sigma criterion. A healthy baseline configuration is shown in fig. 9.
Wherein, the mahalanobis distance calculation formula is as follows:
wherein Z is ij For normalized feature values, p is the number of extracted fault feature parameters, and m is the total number of samples.
Wherein z is j T ＝[z 1j ,z 2j ,…,z pj ]And C is a covariance matrix,
is the mean value of the samples, S i Is the standard deviation.
And (4) carrying out 50 Monte Carlo analyses on the DC-AC inverter circuit without faults, setting parameters of related components as shown in a table 6, and obtaining a healthy baseline of the DC-AC inverter circuit.
TABLE 6 FAULT-FREE DC-AC CIRCUIT 50 SET PARAMETER SETTING TABLE FOR MONOCAROL ANALYSIS DEVICE
The method comprises the steps of conducting 50 Monte Carlo analyses on a DC-AC inverter circuit without faults, wherein Monte Carlo is a simulation statistical method, a numerical calculation method based on probability theory and statistical theory and method is adopted, health characterization parameters shown in a table 4 are adopted, health base lines of the DC-AC inverter circuit based on Mahalanobis distance measurement under different working conditions are obtained and are shown in a table 7, and the change range of each health characterization parameter under different working conditions is shown in a table 8.
TABLE 7 MAMMER DISTANCE MEASUREMENT-BASED DC-AC INVERTER CIRCUIT HEALTH Baseline FOR VARYING OPERATING CONDITIONS
TABLE 8 characterization parameters for DC-AC inverter circuit health under different operating conditions
Under the no-fault mode, 6 different working conditions (see table 5) are considered, the health characterization parameters of the circuits in table 4 are respectively extracted, and the mahalanobis distances based on the seven fault characterization parameters and the health baseline of the DC-AC inverter circuit under the different working conditions are obtained by using the mahalanobis distance-based health baseline construction method in step seven, as shown in fig. 10.
It should be understood that equivalents and modifications of the technical solution and inventive concept thereof may occur to those skilled in the art, and all such modifications and alterations should fall within the scope of the appended claims.
Claims (7)
1. A health characterization and health baseline construction method for a DC-AC inverter circuit comprises the following steps:
the method comprises the following steps: extracting a time domain signal and a frequency domain signal of a detection signal of each measuring point in the DC-AC inverter circuit by utilizing characteristic analysis as characteristic parameters;
step two: wavelet packet decomposition is carried out on detection signals of all measuring points in the DC-AC inverter circuit, and signal energy from low frequency to high frequency is extracted to be used as characteristic parameters;
step three: calculating each performance index of the DC-AC inverter circuit as a characteristic parameter;
step four: the feature parameters extracted in the first step, the second step and the third step are brought into a candidate fault feature library;
step five: optimizing the characteristic parameters in the fault characteristic library, and taking the finally selected characteristic parameters as health characterization parameters;
step six: the method comprises the steps of mapping health characterization parameters of a DC-AC inverter circuit under actual working conditions to standard working conditions to obtain fault characteristic parameters;
step seven: and constructing a Mahalanobis space by calculating the Mahalanobis distance between the fault characteristic parameter and the healthy sample, and taking the Mahalanobis space as a healthy baseline of the DC-AC inverter circuit.
2. The method according to claim 1, wherein the first step further comprises determining whether the operating conditions of the DC-AC inverter circuit are ideal by using the relative variation of the characteristic parameters, if not, establishing regression modeling of the operating conditions and the characteristic parameters, and inputting the characteristic parameters into the regression modeling to obtain the ideal characteristic parameters: wherein, the relative variation of the characteristic parameters is as follows:
θ 0 the characteristic parameter is a characteristic parameter of the DC-AC inverter circuit at an initial time or in a healthy state, θ is a characteristic parameter of the DC-AC inverter circuit after accumulated operation for a period of time, and Δ θ is ideally equal to 0.
3. The method for health characterization and health baseline construction of a DC-AC inverter circuit according to claim 1, wherein the second step comprises the steps of:
(1) to pairPerforming n-layer wavelet packet decomposition on the time series signal X (n) of the detection signal x, and extracting wavelet decomposition coefficients of the nth layer from low frequency to high frequency in each frequency band, wherein
Represents the jth component of the ith layer;
(2) reconstructing the wavelet packet decomposition coefficient to obtain a reconstructed signal;
(3) and calculating the total energy of the reconstructed signals of each frequency band to form a signal vector.
4. The method for health characterization and health baseline construction of a DC-AC inverter circuit according to claim 1, wherein step five is specifically: and (3) optimizing the characteristic parameters in the candidate fault characteristic library by adopting a multi-evaluation index characteristic optimization model, selecting the characteristic parameters of the top 7 in the sequence under each single fault mode, then calculating the spearman correlation coefficient for further screening, and finally selecting 7 characteristic parameters as health characterization parameters.
5. The method for health characterization and health baseline construction of the DC-AC inverter circuit according to claim 4, wherein a multi-evaluation index feature optimization model optimizes feature parameters in a candidate fault feature library by using three evaluation indexes of trend, monotonicity and robustness.
6. The method for health characterization and health baseline construction of a DC-AC inverter circuit according to claim 1, wherein in the sixth step, an equivalent weight analysis method is adopted to map the health characterization parameters of the DC-AC inverter circuit under actual working conditions to standard working conditions to obtain fault characteristic parameters.
7. The method of claim 1, wherein the step seven further comprises determining the baseline threshold according to whether the probability distribution of mahalanobis distance of the healthy sample satisfies normality: (1) if the Mahalanobis distance of the healthy sample meets the normal distribution, determining a baseline threshold value directly according to the 3 sigma criterion of the normal distribution; (2) if the Mahalanobis distance of the healthy sample does not meet normal distribution, BOX-COX transformation is required to be performed firstly to enable the healthy sample to meet the normal distribution, and finally a baseline threshold is determined according to a 3 sigma criterion.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011136503.3A CN112327200B (en) | 2020-10-22 | 2020-10-22 | Health characterization and health baseline construction method of DC-AC inverter circuit |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011136503.3A CN112327200B (en) | 2020-10-22 | 2020-10-22 | Health characterization and health baseline construction method of DC-AC inverter circuit |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112327200A CN112327200A (en) | 2021-02-05 |
| CN112327200B true CN112327200B (en) | 2022-08-30 |
Family
ID=74312233
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011136503.3A Active CN112327200B (en) | 2020-10-22 | 2020-10-22 | Health characterization and health baseline construction method of DC-AC inverter circuit |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112327200B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113219374A (en) * | 2021-03-05 | 2021-08-06 | 国网江苏省电力有限公司徐州供电分公司 | Transformer voltage fault detection method |
| CN112785092A (en) * | 2021-03-09 | 2021-05-11 | 中铁电气化局集团有限公司 | Turnout residual life prediction method based on self-adaptive deep layer feature extraction |
| CN113515811B (en) * | 2021-05-18 | 2022-04-05 | 中国航空工业集团公司上海航空测控技术研究所 | Aviation DC-AC inverter degradation trend prediction method |
| CN113985217B (en) * | 2021-07-26 | 2023-10-13 | 中国电力科学研究院有限公司 | Method and system for determining optimal stunning parameters |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6291987B1 (en) * | 1999-10-28 | 2001-09-18 | General Electric Company | Method and system for detecting incipient failures in a power inverter |
| KR20080104861A (en) * | 2007-05-29 | 2008-12-03 | 김금수 | Inverter logic for estimating state of health and apparatus for control thereof |
| CN108649600A (en) * | 2018-04-04 | 2018-10-12 | 新疆大学 | A kind of parametric method for diagnosing faults of multi-electrical level inverter based on model |
| CN109508909A (en) * | 2018-12-25 | 2019-03-22 | 上海电气分布式能源科技有限公司 | The appraisal procedure and system of the health degree of photovoltaic DC-to-AC converter |
| CN111190088A (en) * | 2019-12-30 | 2020-05-22 | 西安电子科技大学 | Method for extracting characteristic parameters of IGBT (insulated Gate Bipolar transistor) performance degradation |
-
2020
- 2020-10-22 CN CN202011136503.3A patent/CN112327200B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6291987B1 (en) * | 1999-10-28 | 2001-09-18 | General Electric Company | Method and system for detecting incipient failures in a power inverter |
| KR20080104861A (en) * | 2007-05-29 | 2008-12-03 | 김금수 | Inverter logic for estimating state of health and apparatus for control thereof |
| CN108649600A (en) * | 2018-04-04 | 2018-10-12 | 新疆大学 | A kind of parametric method for diagnosing faults of multi-electrical level inverter based on model |
| CN109508909A (en) * | 2018-12-25 | 2019-03-22 | 上海电气分布式能源科技有限公司 | The appraisal procedure and system of the health degree of photovoltaic DC-to-AC converter |
| CN111190088A (en) * | 2019-12-30 | 2020-05-22 | 西安电子科技大学 | Method for extracting characteristic parameters of IGBT (insulated Gate Bipolar transistor) performance degradation |
Non-Patent Citations (7)
| Title |
|---|
| Mahalanobis Distance Approach for Insulated Gate Bipolar Transistors (IGBT) Diagnostics;Patil N等;《17th ISPE International Conference on Concurrent Engineering - New World Situation: New Directions in Concurrent Engineering》;20101102;全文 * |
| VMD小波能量的光伏逆变器软故障诊断;姜媛媛等;《电力系统及其自动化学报》;20181115(第11期);全文 * |
| 一种基于马氏距离的模拟电路软故障检测算法;秦庆强 等;《计算机测量与控制》;20091025(第10期);全文 * |
| 光伏逆变器软故障特征提取方法研究;程浩;《工程科技Ⅱ辑》;20190115;全文 * |
| 城轨车辆辅助逆变器IGBT状态识别研究;季颖等;《计算机测量与控制》;20180625(第06期);全文 * |
| 基于t-SNE流形学习与快速聚类算法的光伏逆变器故障预测技术;张筱辰等;《中国电力》;20200630(第06期);全文 * |
| 基于极限学习机的光伏逆变器软故障辨识方法;李蓉等;《重庆科技学院学报(自然科学版)》;20200615(第03期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112327200A (en) | 2021-02-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112327200B (en) | Health characterization and health baseline construction method of DC-AC inverter circuit | |
| CN109948597B (en) | High-voltage circuit breaker mechanical fault diagnosis method | |
| US20220222409A1 (en) | Method and system for predicting remaining useful life of analog circuit | |
| CN113219300B (en) | Power distribution network single-phase earth fault sensing method based on phase current transient state steady state | |
| CN102288857B (en) | Fault arc identification and detection method and detection protection device | |
| CN110502777A (en) | IGBT module condition detecting system and method based on neural network prediction | |
| CN113899968A (en) | Voltage transformer monitoring method | |
| CN102879728A (en) | Health evaluation index and failure predication method for DC (Direct Current)-DC convertor | |
| CN110703151A (en) | Transformer fault diagnosis method based on vibration blind source separation and Bayesian model | |
| CN111723684A (en) | Method for identifying transient overvoltage type in offshore wind farm | |
| CN112051468A (en) | Method for evaluating health state of aviation static converter under complex working conditions | |
| CN109635430B (en) | Power grid transmission line transient signal monitoring method and system | |
| Yu et al. | Incipient fault diagnosis method for DC–DC converters based on sensitive fault features | |
| CN112636328B (en) | Medium-voltage distribution network equipment load state identification method | |
| CN114184870A (en) | Non-invasive load identification method and equipment | |
| CN113189513A (en) | Ripple-based redundant power supply current sharing state identification method | |
| CN106651206A (en) | Method for evaluating testability evaluation index system of relay protection | |
| CN113884818A (en) | LSTM-based power distribution network fault traveling wave arrival time accurate estimation method | |
| CN116643091A (en) | Online insulation monitoring method for converter transformer valve side sleeve based on multidimensional time sequence characteristics | |
| Rizvi et al. | Real-time ZIP load parameter tracking using adaptive window and variable elimination with realistic synthetic synchrophasor data | |
| CN113872193B (en) | Multipoint harmonic responsibility division method | |
| Liu et al. | Underdetermined blind source separation algorithm of 220kV substation noise based on SCA | |
| CN116819382A (en) | DC-DC converter state monitoring method based on digital twin | |
| CN113419141B (en) | Direct-current line fault positioning method and device | |
| Li et al. | Time-Frequency domain conversion of UHVDC transmission line Corona current based on Fourier transform |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |