Charging module state identification method and device based on sample entropy

Abstract

The invention provides a charging module state identification method and device based on sample entropy, and belongs to the technical field of fault identification. The method comprises the following steps: acquiring a plurality of working performance parameter sequences of a charging module; wherein each working performance parameter sequence is a value of one working performance parameter at a plurality of moments; aiming at each working performance parameter sequence, normalizing each working performance parameter in the working performance parameter sequence to obtain a plurality of parameter normalized sequences of the charging module; performing empirical mode decomposition on each parameter normalized sequence to obtain a plurality of parameter component sequences of the charging module; respectively calculating a sample entropy of each parameter component sequence to obtain a plurality of parameter sample entropies of the charging module, and combining the parameter sample entropies as a fault feature vector of the charging module; and inputting the fault feature vector into a trained state recognition model to obtain a fault state of the charging module. According to the invention, robustness and accuracy of state identification of the charging module can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of fault diagnosis, and in particular to a charging module state recognition method and device based on sample entropy. BACKGROUND

[0002] The emergency backup power supply system of a power station charges storage batteries using a direct current charging power module. The performance state of the direct current charging power module has a significant impact on the safety of charging the storage battery, the insulation deterioration of the module, and the conversion efficiency. Therefore, monitoring the performance state of the direct current charging power module is an important link for the safe operation of the emergency backup power supply system of the power station.

[0003] Currently, methods for recognizing the state of a direct current charging power module include: 1. a method based on an analytical model, which simulates the working state of the circuit, predicts the voltage or current, and compares it with a reference value; 2. a data-based modeling method, which is usually based on an offline modeling framework and uses expert experience methods to correct the learned model parameters; and 3. a signal processing-based method, which decomposes and transforms data according to a signal analysis strategy, extracts feature information therefrom, and can achieve fault diagnosis and recognition by comparing the change rules of the feature information before and after the fault. However, the robustness and accuracy of the above methods still need to be improved. SUMMARY

[0004] Embodiments of the present application provide a sample entropy method and device to solve the problem of improving the robustness and accuracy of charging module state recognition.

[0005] In a first aspect, embodiments of the present application provide a charging module state recognition method based on sample entropy, comprising:

[0006] obtaining a plurality of working performance parameter sequences of the charging module; wherein each working performance parameter sequence is the value of a kind of working performance parameter at a plurality of time points;

[0007] normalizing each working performance parameter in each working performance parameter sequence to obtain a plurality of parameter normalized sequences of the charging module;

[0008] respectively performing empirical mode decomposition on each parameter normalized sequence to obtain a plurality of parameter component sequences of the charging module;

[0009] respectively calculating the sample entropy of each parameter component sequence to obtain a plurality of parameter sample entropies of the charging module, and combining each parameter sample entropy as a fault feature vector of the charging module;

[0010] inputting the fault feature vector into a trained state recognition model to obtain the fault state of the charging module.

[0011] In a second aspect, an embodiment of the present application provides a charging module state recognition device based on sample entropy, comprising:

[0012] An acquisition module is configured to acquire a plurality of working performance parameter sequences of the charging module, wherein each working performance parameter sequence is a value of a kind of working performance parameter at a plurality of time points;

[0013] A normalization module is configured to normalize each working performance parameter in each working performance parameter sequence to obtain a plurality of parameter normalized sequences of the charging module.

[0014] A mode decomposition module is configured to perform empirical mode decomposition on each parameter normalized sequence to obtain a plurality of parameter component sequences of the charging module.

[0015] A sample entropy calculation module is configured to calculate the sample entropy of each parameter component sequence to obtain a plurality of parameter sample entropies of the charging module, and combine each parameter sample entropy as a fault feature vector of the charging module.

[0016] A state recognition module is configured to input the fault feature vector into a trained state recognition model to obtain a fault state of the charging module.

[0017] The embodiment of the present application provides a charging module state recognition method and device based on sample entropy, which removes the change classification in the working performance parameter sequence of the charging module through normalization, guarantees the state recognition accuracy, selects the component that can significantly reflect the state difference of the charging module, such as normal, abnormal and fault, through empirical mode decomposition, improves the accuracy of the state recognition method, obtains the fault feature vector through sample entropy combination, takes into account the data compression ratio and the rapidity of the state recognition method, and thus improves the robustness and accuracy of the charging module state recognition. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without any creative effort on the basis of these drawings.

[0019] Figure 1 is an application scenario diagram of the charging module state recognition method based on sample entropy provided by the embodiment of the present application;

[0020] Figure 2 is an implementation flowchart of the charging module state recognition method based on sample entropy provided by the embodiment of the present application;

[0021] Figure 3is a schematic diagram of an experimental platform provided by an embodiment of the present application;

[0022] Figure 4A is a working temperature simulation diagram of a secondary LLC converter switching power supply MOSFET open circuit fault provided by an embodiment of the present application without protection;

[0023] Figure 4B is a working temperature simulation diagram of a secondary LLC converter switching power supply MOSFET open circuit fault provided by an embodiment of the present application with protection;

[0024] Figure 4C is a three-phase input current and output current simulation diagram after a single MOSFET short circuit fault provided by an embodiment of the present application;

[0025] Figure 5A is a fault detection result diagram when S0 and S0* are shorted provided by an embodiment of the present application;

[0026] Figure 5B is a VD1 open circuit fault detection result diagram provided by an embodiment of the present application;

[0027] Figure 5C is a D3 short circuit fault detection result diagram provided by an embodiment of the present application;

[0028] Figure 5D is a S1 short circuit fault detection result diagram provided by an embodiment of the present application;

[0029] Figure 6 is a fault experiment precision diagram provided by an embodiment of the present application;

[0030] Figure 7 is a structure schematic diagram of a charging module state recognition device based on sample entropy provided by an embodiment of the present application. DETAILED DESCRIPTION

[0031] In the following description, for the purpose of explanation and not limitation, specific details are set forth, such as particular system configurations, techniques, etc., in order to provide a thorough understanding of the embodiments of the present application. However, it will be apparent to those skilled in the art that the present application can be practiced in other embodiments that depart from these specific details. In other instances, detailed descriptions of well-known systems, devices, circuits, and methods are omitted so as not to obscure the description of the present application with unnecessary detail.

[0032] In order to make the objects, technical solutions and advantages of the present application clearer, the following will be described by specific embodiments in conjunction with the accompanying drawings.

[0033] With the rapid development of new energy power systems, lithium-ion batteries have been widely used in emergency backup power supplies for power stations. However, the charging safety requirements of the charging module have high reliability, so it is particularly necessary to identify the fault type and location in time and accurately. The output characteristics of the converter are controlled by changing the duty cycle. The change of duty cycle will cause small fluctuations or significant changes in voltage and current, which will lead to the degradation of system stability and its equivalent parameters, and even open and short circuit faults. Surge current caused by faults can cause serious damage to other components. Power electronic switching devices are the weakest link in the reliability of the charging module, and early diagnosis of abnormalities or faults helps fault-tolerant operation or fault protection, which is crucial to the safety of battery charging. Fault diagnosis methods mainly use analytical models, data-driven techniques, and signal processing methods.

[0034] Using the method based on the analytical model, the working state of the circuit is simulated, the voltage or current is predicted, and it is compared with the reference value. Then we can find the abnormal state of the circuit operation.

[0035] Jlassi compares the phase current measurements before and after the fault, and realizes the diagnosis function according to the characteristics of the obtained phase current residual value. However, it is difficult to define the diagnostic variable. Pecina proposed a sliding mode PI observer, which can realize accurate tracking of the phase current, and can realize diagnosis of multiple tube open circuit faults according to the characteristics of the fault profile vector and the current residual value. Although this method can achieve faster fault diagnosis and accurate fault location, the fault evolution mechanism requires a high-precision model, which leads to modeling difficulties.

[0036] Data-based modeling methods are usually based on modeling frameworks under offline conditions, and use expert experience methods to correct the learned model parameters. A portable data-driven fault diagnosis method based on extreme learning machine is proposed for open circuit fault diagnosis of three inverters. The current signal is used as the original input to generate fault features. The learning machine model is trained using original system data to form a diagnostic model, but this method has high requirements for training data. Chen applied artificial neural networks to fault diagnosis of multi-level inverters, and used genetic algorithms to optimize the structure of artificial neural networks, further improving the diagnostic performance.

[0037] The signal processing-based method decomposes and transforms the data according to the signal analysis strategy, and extracts the feature information from it. By comparing the change rule of the feature information before and after the fault, the fault diagnosis and identification can be realized. Abari proposed a non-invasive diagnostic strategy, which detects the near-field voltage signal of the DC bus through the antenna, and uses FFT as the basis for fault classification to extract the frequency spectrum characteristics of the obtained signal, but the diagnostic information is less. A microgrid inverter fault feature extraction strategy based on phase current signal symmetry reconstruction is proposed, and multi-scale feature extraction is performed on the signal. However, the selection of wavelet basis will affect the extraction effect and increase the application difficulty.

[0038] The emergency backup power system of the power station uses a DC charging power module to charge the battery. The performance state of the DC charging power module has a significant impact on the safety of battery charging, module insulation degradation, and conversion efficiency. For charging modules, short-circuit faults and phase loss faults are protected, but MOSFET and diode degradation and open-circuit are not protected. According to the real-time identification of the state of the DC charging power module based on the sampling sequence of transient input and output electrical parameters and temperature, it depends on the customized state feature extraction method. Therefore, to obtain the state identification feature quantity of the charging power module in real time, it is necessary to consider the fluctuation range of the input side power supply voltage of the charging power module and the change range of the output side battery voltage to improve the robustness of the state identification method; it is necessary to consider selecting components that can significantly reflect the differences between normal, abnormal, and fault states of the charging power module from the collected sequences of performance parameters to improve the accuracy of the state identification method; it is necessary to consider selecting the capacity of the sequence and extracting the sequence feature parameters as the identification state features to balance the data compression ratio and the speed of the state identification method.

[0039] To solve this problem, this paper simulates the device faults at different positions of the charging module and proposes three risk levels. The main contributions of this paper are as follows

[0040] 1. A fault model of the charging module is established based on simulation, and the changes in input and output characteristics of the charging module before and after the fault are studied.

[0041] 2. In order to accurately identify the fault type and fault location of the charging module, the charging module faults are classified and coded in detail, totaling 32 fault types.

[0042] 3. For different features of different faults, the fault feature vector is extracted from the input and output parameters, and a state feature normalization method SN-EMD-SSE (Normalization-Empirical Mode Decomposition-Sequence Sample Entropy Extraction) is proposed, which, combined with the state recognition model, greatly improves the accuracy and speed of fault diagnosis.

[0043] 4. The accuracy and robustness of the EMD-SSE-HKELM-DBO (Empirical Mode Decomposition-Sequence Sample Entropy Extraction-Hybrid Kernel Extreme Learning Machine-Dung Beetle Optimizer) model are verified by a large number of experiments.

[0044] Figure 1 The application scenario diagram of the sample entropy-based charging module state recognition method provided by the embodiment of the application is shown in FIG. 1. Figure 1 As shown in FIG. 1, the main circuit of the charging module includes a rectification and voltage boosting pre-stage circuit and an electrically isolated current regulation secondary circuit. The pre-stage circuit is composed of six diodes VDi (i = 1, 2, 3, 4, 5, 6), in which a 3-p rectifier is provided with a filter capacitor CVD, two MOSFETs (S0 and S0*) are provided with anti-parallel diodes (D0 and D0*) and coupling capacitors (C0 and C0*), a filter capacitor Cb, a diode Db and an inductor Lb for continuous current in the voltage boosting circuit. The secondary circuit is composed of four MOSFETs (S1 and S1*, S2 and S2*), anti-parallel diodes (D1 and D1*, D2 and D2*) and coupling capacitors (C1 and C1*, C2 and C2*) in the switching power supply, a resonant capacitor Cr, a resonant inductor Lr and an exciting inductor Lm in the LLC converter, an ideal transformer and two diodes (D3 and D4) and a filter capacitor C3 for a 1-p full-wave rectifier at the output end thereof. uA, uB and uC are three-phase input voltages. iA, iB and iC are three-phase input currents. Io is the working current at the output end of the module. Vo is the output voltage, equal to the value of the battery. In addition, the ripple Δio of the input and output (= max(io(k)-min(io(k), k = 1, 2...n) is calculated, the working temperatures of each MOSFET (T0 and T0*, T1 and T1*, T2 and T2*) are monitored, and the output load is tested under independent working conditions by increasing the input and output load.

[0045] Figure 2 The implementation flowchart of the sample entropy-based charging module state recognition method provided by the embodiment of the application is shown in FIG. 2, and will be described in detail as follows.

[0046] In step 201, a plurality of working performance parameter sequences of the charging module are acquired; wherein each working performance parameter sequence is the value of a kind of working performance parameter at a plurality of time points.

[0047] In the embodiment, the selectable working performance parameters include three three-phase input currents, output currents and six MOSFET working temperatures, one or more of which can be selected according to actual needs, and the corresponding working performance parameter sequence is obtained. Specifically, the transistors for temperature detection can be selected according to the historical damage data of the transistors at each position, for example, the number of damage, and the severity of the influence of the transistors after damage on the circuit. In feature extraction, the selection of MOSFETs from the use in the final circuit and the measurable temperature conditions can be combined with the data of stable operation and fault experiments to select the temperature features.

[0048] In step 202, for each working performance parameter sequence, each working performance parameter in the working performance parameter sequence is normalized to obtain a plurality of parameter normalized sequences of the charging module.

[0049] In the embodiment, the working performance parameters such as transient current have a large number of variation components under different conditions, which leads to a decrease in state recognition accuracy. Normalizing the working performance parameters can remove interference information in the parameters to ensure state recognition accuracy.

[0050] In step 203, each parameter normalized sequence is respectively subjected to empirical mode decomposition to obtain a plurality of parameter component sequences of the charging module.

[0051] In the embodiment, empirical mode decomposition can decouple the original signal, decompose the signal into a plurality of intrinsic mode functions (IMF) based on the time scale characteristics of the original signal itself, and each IMF component represents the local characteristics of the original signal, which facilitates the analysis of the time sequence characteristics of each parameter normalized sequence.

[0052] In step 204, the sample entropy of each parameter component sequence is calculated respectively to obtain a plurality of parameter sample entropies of the charging module, and each parameter sample entropy is combined as a fault feature vector of the charging module.

[0053] In the embodiment, sample entropy (SE) measures the complexity of a time sequence by measuring the probability of generating a new pattern in the signal. The greater the probability of generating a new pattern, the greater the complexity of the sequence. Compared with approximate entropy, sample entropy has two advantages: the calculation of sample entropy does not depend on the length of the data; and sample entropy has better consistency. Each parameter sample entropy specifically describes the time sequence complexity of each working performance parameter of the charging module, and the combination thereof can describe the running state of the charging module in multiple dimensions to achieve fault state recognition.

[0054] In step 205, the fault feature vector is input into a trained state recognition model to obtain the fault state of the charging module.

[0055] In the embodiment, the state recognition model can be a hybrid kernel extreme learning machine (HKELM) regression model. The HKELM model is trained by using the fault feature vectors and the fault states, so that the HKELM model learns the relationship between the fault feature vectors and the fault states, classifies the input fault feature vectors, determines the corresponding fault states, and realizes the fault state recognition of the charging module.

[0056] The embodiment of the application removes the change classification in the working performance parameter sequence of the charging module through normalization, ensures the state recognition accuracy, selects the component that can significantly reflect the state differences of the charging module such as normal state, abnormal state and fault state from the performance parameters through empirical mode decomposition, improves the accuracy of the state recognition method, obtains the fault feature vector through sample entropy combination, takes into account the data compression ratio and the rapidity of the state recognition method, and thus improves the robustness and accuracy of the charging module state recognition.

[0057] In a possible implementation manner, the working performance parameter sequence includes three-phase input current sequences, output current sequences and working temperature sequences, each three-phase input current sequence corresponds to one phase of the three-phase input current as the working performance parameter, and each working temperature sequence corresponds to the working temperature of one transistor as the working performance parameter.

[0058] For each working performance parameter sequence, each working performance parameter in the working performance parameter sequence is normalized to obtain a plurality of parameter normalized sequences of the charging module, including:

[0059] For each three-phase input current sequence, the normalized input current corresponding to each three-phase input current in the three-phase input current sequence is calculated based on a normalized input current formula to obtain the parameter normalized sequence corresponding to the three-phase input current sequence; wherein the normalized input current formula is:

[0060]

[0061] wherein, i k is the instantaneous value of the three-phase input current, is the normalized input current corresponding to the three-phase input current, k=A, B, C, U k is the effective value of the input end phase voltage, U o is the output voltage, I o is the output current;

[0062] For the output current sequence, the normalized output current corresponding to each output current in the output current sequence is calculated based on a normalized output current formula to obtain the parameter normalized sequence corresponding to the output current sequence; wherein the normalized output current formula is:

[0063]

[0064] where I o is the output current, is the normalized output current corresponding to the output current, i oref is the expected value of the output current;

[0065] For each working temperature sequence, the normalized working temperature corresponding to each working temperature in the working temperature sequence is calculated based on the normalized working temperature formula, and the parameter normalization sequence corresponding to the working temperature sequence is obtained; wherein the normalized working temperature formula is:

[0066]

[0067] where T i is the working temperature of the transistor, is the normalized working temperature of the transistor, is the average value of the working temperature of the transistor.

[0068] In this embodiment, the output voltage and output current are also needed for the normalization processing of the three-phase input current sequence, because in actual application, the output signals of sensors or other devices often have different maximum and minimum values. For example, in some designs, the output voltage value of the sensor needs to be processed by a formula to keep it between 0 and 100. This processing not only involves voltage values, but also may involve current values, because current is also an important parameter for measuring sensor performance.

[0069] In addition, the limitations of hardware will also affect the normalization processing. For example, if the hardware requires that the collected voltage can only be between 0 and 5V, then normalization processing needs to be performed before the voltage output, adjusting the actual voltage value to within this range. This shows that when performing normalization processing, not only the characteristics of the data itself need to be considered, but also the limitations of the hardware device.

[0070] The output voltage and output current are needed for normalization processing because these parameters directly affect the processing and analysis results of the data. By appropriately normalizing these parameters, the consistency and comparability of the data can be ensured, thereby improving the performance and accuracy of the model.

[0071] Normalization of the output current sequence based on the expected value of the output current can significantly improve the state recognition accuracy of the algorithm, simplify the prediction of non-stationary time series data, and improve the performance of the model.

[0072] Based on the average temperature of the transistor with temperature coupling, each working temperature sequence is normalized, which can improve the comparability of data (normalization is a data processing method that limits data within a certain fixed range or converts data with dimensions to dimensionless data, which can eliminate the dimensional influence between different working temperature sequences and ensure the comparability of data. This is particularly important for thermal coupling problems in complex chips, as the temperature distribution between transistors on different layers can be significantly different); simplify model training (normalizing the average temperature of the transistor can make it easier for the model to capture key features, thereby improving the performance and stability of the model); enhance model robustness; optimize thermal management strategy (by normalizing each working temperature sequence, the behavior of the transistor at different temperatures can be better understood and predicted. This helps to optimize the thermal management strategy); improve the speed of solving the optimal solution (it can improve the speed of solving the optimal solution, as it reduces the comparability problem between data, making the algorithm converge to the optimal solution faster).

[0073] In one possible implementation, each parameter normalized sequence is subjected to empirical mode decomposition to obtain a plurality of parameter component sequences of the charging module, including:

[0074] For each parameter normalized sequence, empirical mode decomposition is performed to obtain a plurality of IMF components corresponding to the parameter normalized sequence;

[0075] For each parameter normalized sequence, the first n IMF components in the plurality of IMF components corresponding to the parameter normalized sequence are selected as the plurality of parameter component sequences corresponding to the parameter normalized sequence; wherein n is a preset value.

[0076] In this embodiment, empirical mode decomposition is an adaptive time-frequency processing method for nonlinear non-stationary signals x(t). It decouples the signal into n intrinsic mode functions (IMF) IMFi(i=1,2...) and a trend component rn(T),

[0077]

[0078] The IMF component should satisfy the following conditions: (1) the sum of extreme points should be equal to the number of zero points or at most one difference; (2) the upper and lower envelopes are locally symmetric on the time axis.

[0079] The general EMD decomposition process of a non-smooth signal x(t) is:

[0080] 1) Determine all local extreme points of x(t), connect the local extreme points into upper envelope v1(t) and lower envelope v2(t) with a cubic spline function, and calculate the average value m(t):

[0081]

[0082] 2) Calculate h 10 (t):

[0083] h 10 (t) = x(t) - m(t)

[0084] If h 10 (t) is an IMF, then it is the first IMF component IMF1 of x(t); if h 10 (t) is not an IMF, then it is used as the original signal to repeat steps 1) and 2) to calculate h 1k (t), (k = 1, 2,...), until the IMF condition is met, obtaining IMF1;

[0085] 3) Calculate r n (t):

[0086]

[0087] Take r n (t) as the initial signal, repeat steps 1-2 until r n (t) becomes a monotonic function, and the criterion for judging is that the sd of two consecutive IMFs is less than a given value:

[0088]

[0089] where T is the time span of the signal, and h 1(k-1) (t) and h 1k(t) are two consecutive IMFs in the decomposition process.

[0090] The number n of IMF components can be determined according to actual conditions. For example, when 10 performance parameters (three input currents, one output current, and six operating temperatures) are normalized and used for modal analysis, 6 component samples of IM1, IM2, IM3, IM4, IM5, and IM6 can be selected for each sequence to form 60 sequences for state characteristic parameter extraction.

[0091] In one possible implementation, the sample entropy of each parameter component sequence is calculated respectively to obtain multiple parameter sample entropies of the charging module, including:

[0092] For each parameter component sequence, the sample entropy of the parameter component sequence is calculated based on the sample entropy formula; wherein the sample entropy formula is:

[0093]

[0094]

[0095] d[X m (i), Xm (j)] = max[|x(i+k)-x(j+k)|]

[0096] where En Sample (m,r) is the sample entropy of component sequence, m is the dimension of component sequence, r is the similarity tolerance, B m (r) is the probability of component sequence matching m points with component sequence of increasing dimension to m+1, is the number of distance between any two elements in component sequence less than the similarity tolerance, X m (i) is the i-th element of component sequence, d[X m (i),X m (j)] is the distance between element X m (i) and element X m (j).

[0097] In this embodiment, the sample entropy of sequence d is a good tool for feature extraction of nonlinear time series XN(1) = x(1), x(2),..., x(N). The calculation results are as follows:

[0098] 1) Extract N-m+1 dimensional vector sequence using ordinal number Xm(i), (i = 1, 2... N-m+1), and calculate the distance between Xm(i) and Xm(j):

[0099] d[X m (i),X m (j)] = max[x(i+k)-x(j+k)]

[0100] 2) Given r (r>0), calculate the number of d[Xm(i), Xm(j)] less than r for each I, and calculate the ratio of it to the total number of distance N-m, marked as

[0101]

[0102] Calculate the average B m (r) of N-m+1 ratio :

[0103]

[0104] 3) N-m vector sequence, m+1 dimensional Xm+1(i), (i = 1, 2... N-m), has:

[0105]

[0106] 4) The entropy of sequence sample EnSample(m,r) is calculated as follows:

[0107]

[0108] The sample entropy is closely related to the sequence dimension M and the similarity tolerance r. Some studies show that m is 1 or 2, and r is 0.1-0.25std (std is the standard deviation of the original data).

[0109] In a possible implementation, before the sample entropy of each parametric component sequence is calculated based on the sample entropy formula, the method further includes the following steps:

[0110] For each parametric component sequence, a part of a preset time period is intercepted as a subsequence of the parametric component sequence.

[0111] Correspondingly, for each parametric component sequence, the sample entropy of the parametric component sequence is calculated based on the sample entropy formula, including:

[0112] For each parametric component sequence, the sample entropy of the subsequence of the parametric component sequence is calculated based on the sample entropy formula, and is taken as the sample entropy of the parametric component sequence.

[0113] In the embodiment, the same time period is intercepted for each parametric component sequence. The intercepted time period can be one power frequency cycle or more complete power frequency cycles. This way can ensure that the intercepted part contains the parametric characteristics of the charging module in the fault state, while shortening the sequence length and reducing the calculation amount of the sample entropy.

[0114] In a possible implementation, the method further includes the following steps:

[0115] The multiple working performance parametric sequences of the charging module are obtained in cycles of a preset time length.

[0116] In the embodiment, the length of the preset time length should be less than the length of the working performance parametric sequence. For example, the working performance parametric sequence with a length of 100 ms is refreshed every 5 ms. The embodiment can ensure that no sampling value of the working performance parameter is missed, and realize real-time and close state recognition of the charging module.

[0117] In a possible implementation, before the fault feature vector is input into the trained state recognition model, the method further includes the following steps:

[0118] An initial state recognition model is obtained.

[0119] The initial state recognition model is trained by using a training sample set to obtain the trained state recognition model. The training sample set includes multiple training samples, each training sample is a group of fault feature vectors, and the label of each training sample is a fault state code.

[0120] In this embodiment, different types of faults can be set in the simulation circuit, and corresponding fault feature vectors can be collected to construct a training sample set to train the initial state recognition model, enabling the state recognition model to learn the relationship between fault feature vectors and fault states. To accurately identify the fault type and location of the charging module, this application has meticulously classified and coded the fault states of the charging module, resulting in a total of 32 fault types.

[0121] In one possible implementation, the fault states of the charging module include no-risk level, low-risk level, and high-risk level.

[0122] In this embodiment, to visually represent the severity of the charging module's fault state, three risk levels, G, Y, and R, are defined, corresponding to no risk, low risk, and high risk levels, respectively. Risk level R corresponds to out-of-range failure, where a variable exceeds the range determined by protection values ​​or overload limits. Risk level g indicates no or slight deterioration of parts and electrical contacts, corresponding to normal operating conditions, with each variable varying within a certain range, generally determined by rated values ​​and usage reduction factors. Level Y lies between G and R, typically caused by a gradual, non-negligible deterioration. Different levels of alarms and responses can be implemented based on the different risk levels of the charging module.

[0123] In one specific embodiment, to verify the effectiveness and feasibility of the fault identification method proposed in this paper, a system was built. Figure 3 The charging module experimental platform shown conducted open-circuit fault experiments on diodes and switches at different locations. Input and output temperature data were collected using an oscilloscope as the training sample set for the DBO-HKELM model, divided into a training set and a test set, each accounting for 50% of the total. The main components used in the platform are: a PFC inductor (model TH6-9021.3, size 940MF), a diode (model C4D15120D), an IGBT (model IKW40N120H3), and an NTC thermistor (model CWF4B103j3950). Data analysis and processing were implemented in computer software.

[0124] To improve the robustness of the diagnostic method, the input voltage was set within ±15% (i.e., 187V to 264V) of the rated value (i.e., 220VAC), and the output voltage was set within ±10% (i.e., 198V to 286V) of the rated value (i.e., 220VDC), thus simulating the full operating conditions.

[0125] The most common fault types are considered in this embodiment, namely open circuit fault, short circuit fault and abnormal operating temperature of MOSFET or filter diode in operation. For the charging module, short circuit, overcurrent, overvoltage faults are protected, but the deterioration of MOSFET or rectifier diode and open circuit fault have not fully played their protection function. The simulation results of the fault are shown in Figures 4A-4C Figure 4A Figure 4B Figure 4C

[0126] According to the time domain characteristics of these variables, three risk levels G, Y and R are designed, corresponding to no risk level, low risk level and high risk level respectively. Risk level R corresponds to out-of-limit failure, that is, a certain variable exceeds the range determined by the protection value or overload state limit value. Risk level G is that the part and electrical contact are not or slightly deteriorated, corresponding to normal operating condition, within a certain range of each variable variation, generally determined by the rated value and use reduction coefficient. Y level is between G and R, which is usually caused by gradual deterioration that cannot be ignored. In this embodiment, the risk level threshold of state characteristics is set as shown in Table 1.

[0127] Table 1

[0128]

[0129] According to the risk level threshold in Table 1, the fault mode code corresponding to the state characteristics and position of MOSFET or diode listed in Table 2 can be determined.

[0130] Table 2

[0131]

[0132]

[0133]

[0134] In this embodiment, empirical mode decomposition is used to decompose the three-phase input current and DC output current of the charging module. According to the characteristics of the three-phase input current and DC output current of the charging module, the IMF component with obvious fault characteristics is extracted, the parameter m=2, R=0.1std, and the sample entropy is calculated as the fault feature vector S=[EnSample1, EnSample2...].

[0135] ​​​​A fault feature vector is extracted from the current and temperature simulated under all working conditions with faults occurring at a given time point, a state recognition model is selected and constructed, data set division, model selection and optimization, model training and model verification are performed, finally the HKELM-DBO model is adopted, and the root mean square error (RMSE) and the mean absolute relative error (MARE) are used for precision to obtain the best model parameters. Full working condition unprotected fault state recognition is carried out on the experimental platform, and part of the results are shown in FIG. 5, wherein Figure 5A is a fault detection result diagram for S0 and S0* short circuit, Figure 5B is a VD1 open circuit fault detection result diagram, Figure 5C is a D3 short circuit fault detection result diagram, Figure 5D is a S1 short circuit fault detection result diagram. The accuracy of fault diagnosis is counted, and the counting results are shown in Figure 6 , wherein the abscissa is the sample number, the ordinate is the diagnosis result, the red * in the figure is the algorithm diagnosis result, and the blue O is the actual fault experiment diagnosis result. The comparison of the two can represent the accuracy of the diagnosis algorithm.

[0136] The experimental results show that the recognition method has good source-load adaptability, high robustness and high precision. The performance variables of the method are real-time measurable current and temperature, which aims to solve the recognition problem at the application level, can accurately recognize the module state and risk level, locate the power electronic device fault, and bring the unprotected power electronic device fault into the protection category. It can also identify protected faults, provide evidence support for intelligent control and operation.

[0137] It should be understood that the size of the serial number of each step in the above embodiment does not mean the order of execution, and the execution order of each process should be determined according to its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiment of the present application.

[0138] The following is a device embodiment of the present application. For details not described in detail, please refer to the corresponding method embodiments described above.

[0139] Figure 7 The structure schematic diagram of the charging module state recognition device based on sample entropy provided by the embodiment of the present application is shown, only the part related to the embodiment of the present application is shown for convenience of description, and the details are as follows:

[0140] As shown in Figure 7 , the charging module state recognition device based on sample entropy 8 comprises:

[0141] The acquisition module 81 is configured to acquire a plurality of working performance parameter sequences of the charging module; wherein each working performance parameter sequence is a value of a kind of working performance parameter at a plurality of time points;

[0142] The normalization module 82 is configured to normalize each working performance parameter in each working performance parameter sequence to obtain a plurality of parameter normalized sequences of the charging module;

[0143] The modal decomposition module 83 is configured to respectively perform empirical mode decomposition on each parameter normalized sequence to obtain a plurality of parameter component sequences of the charging module;

[0144] The sample entropy calculation module 84 is configured to respectively calculate sample entropy of each parameter component sequence to obtain a plurality of parameter sample entropies of the charging module, and combine the parameter sample entropies as a fault feature vector of the charging module.

[0145] The state recognition module 85 is configured to input the fault feature vector into a trained state recognition model to obtain a fault state of the charging module.

[0146] In a possible implementation, the working performance parameter sequence includes a three-phase input current sequence, an output current sequence and a working temperature sequence, each three-phase input current sequence corresponds to one phase of the three-phase input current, and each working temperature sequence corresponds to the working temperature of one transistor.

[0147] The normalization module 82 is specifically configured to:

[0148] For each three-phase input current sequence, the normalization module 82 is configured to calculate a normalized input current corresponding to each three-phase input current in the three-phase input current sequence based on a normalized input current formula to obtain a parameter normalized sequence corresponding to the three-phase input current sequence; wherein the normalized input current formula is:

[0149]

[0150] wherein, i k is an instantaneous value of the three-phase input current, is the normalized input current corresponding to the three-phase input current, k=A, B, C, U k is a root mean square value of the input end phase voltage, U o is an output voltage, I o is an output current;

[0151] For the output current sequence, the normalization module 82 is configured to calculate a normalized output current corresponding to each output current in the output current sequence based on a normalized output current formula to obtain a parameter normalized sequence corresponding to the output current sequence; wherein the normalized output current formula is:

[0152]

[0153] wherein, I o is the output current, is a normalized output current corresponding to the output current I oref is an expected value of the output current;

[0154] For each working temperature sequence, a normalized working temperature corresponding to each working temperature in the working temperature sequence is calculated based on a normalized working temperature formula, to obtain a parameter normalization sequence corresponding to the working temperature sequence; wherein the normalized working temperature formula is:

[0155]

[0156] wherein T i is a working temperature of the transistor, is a normalized working temperature of the transistor, is an average value of the working temperature of the transistor.

[0157] In a possible implementation, the modal decomposition module 83 is specifically configured to:

[0158] For each parameter normalization sequence, an empirical mode decomposition is performed to obtain a plurality of IMF components corresponding to the parameter normalization sequence;

[0159] For each parameter normalization sequence, the first n IMF components in the plurality of IMF components corresponding to the parameter normalization sequence are selected as a plurality of parameter component sequences corresponding to the parameter normalization sequence; wherein n is a preset value.

[0160] In a possible implementation, the sample entropy calculation module 84 is specifically configured to:

[0161] For each parameter component sequence, a sample entropy of the parameter component sequence is calculated based on a sample entropy formula; wherein the sample entropy formula is:

[0162]

[0163] d[X m (i),X m (j)]=max[|x(i+k)-x(j+k)|]

[0164] wherein En Sample (m,r) is a sample entropy of the component sequence, m is a dimension of the component sequence, r is a similarity tolerance, B m (r) is a probability that the component sequence and the component sequence with the dimension increased to m+1 match m points, is a number of distances between any two elements in the component sequence being less than the similarity tolerance, X m (i) is an i-th element of the component sequence, d[X m (i),X m (j)] is an element X m(i) distance from element X m (j) distance.

[0165] In a possible implementation, the sample entropy calculation module 84 is further configured to:

[0166] Before calculating the sample entropy of each sequence of the parameter component based on the sample entropy formula, for each sequence of the parameter component, a part of a preset time period is intercepted as a subsequence of the sequence of the parameter component.

[0167] Correspondingly, the sample entropy calculation module 84 is specifically configured to:

[0168] For each sequence of the parameter component, the sample entropy of the subsequence of the sequence of the parameter component is calculated based on the sample entropy formula, and is taken as the sample entropy of the sequence of the parameter component.

[0169] In a possible implementation, the acquisition module 81 is specifically configured to:

[0170] The plurality of sequences of the working performance parameters of the charging module are acquired in a preset time period.

[0171] In a possible implementation, the state identification module 85 is further configured to:

[0172] Before inputting the fault feature vector into the trained state identification model, an initial state identification model is acquired.

[0173] The initial state identification model is trained by using a training sample set to obtain the trained state identification model, wherein the training sample set includes a plurality of training samples, each training sample is a group of fault feature vectors, and the label of each training sample is a fault state code.

[0174] In a possible implementation, the fault states of the charging module include a no-risk level, a low-risk level and a high-risk level.

[0175] The embodiments of the present application remove the variation classification in the sequence of the working performance parameters of the charging module through normalization, ensure the state recognition accuracy, select the component that can significantly reflect the state differences of the charging module such as normal, abnormal and fault from the performance parameters through empirical mode decomposition, improve the accuracy of the state recognition method, obtain the fault feature vector through sample entropy combination, take into account the data compression ratio and the rapidity of the state recognition method, and thus improve the robustness and accuracy of the state recognition of the charging module.

[0176] In the above embodiments, the description of each embodiment has its own focus, and the part not described or recorded in detail in a certain embodiment can be referred to the related description of other embodiments.

[0177] Those skilled in the art can appreciate that the templates, units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether the functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. The skilled person can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0178] The modules / units, if implemented in the form of software function units and sold or used as independent products, can be stored in a computer readable storage medium. Based on such understanding, all or part of the processes in the above-mentioned embodiment methods can also be completed by instructing related hardware through a computer program, and the computer program can be stored in a computer readable storage medium. When the processor executes the computer program, the steps of each of the above-mentioned sample entropy-based charging module state recognition method embodiments can be implemented. The computer program includes computer program code, which can be in the form of source code, object code, executable files or some intermediate forms, etc. The computer readable medium can include any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory, random access memory, electrical carrier signal, telecommunication signal and software distribution medium, etc.

[0179] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A charging module state recognition method based on sample entropy, characterized in that, The method includes: Obtain multiple operating performance parameter sequences of the charging module; wherein each operating performance parameter sequence is the value of an operating performance parameter at multiple times; For each sequence of operating performance parameters, each operating performance parameter in the sequence is normalized to obtain multiple parameter normalization sequences of the charging module. Empirical mode decomposition is performed on each parameter normalization sequence to obtain multiple parameter component sequences of the charging module; The sample entropy of each parameter component sequence is calculated separately to obtain multiple parameter sample entropies of the charging module, and the sample entropies of each parameter are combined as the fault feature vector of the charging module; the fault feature vector is input into the trained state recognition model to obtain the fault state of the charging module.

2. The charging module state recognition method based on sample entropy according to claim 1, characterized in that, The operating performance parameter sequence includes a three-phase input current sequence, an output current sequence, and an operating temperature sequence. The operating performance parameter corresponding to each three-phase input current sequence is one phase of the three-phase input current, and the operating performance parameter corresponding to each operating temperature sequence is the operating temperature of a transistor. For each sequence of operating performance parameters, each operating performance parameter in the sequence is normalized to obtain multiple parameter normalization sequences of the charging module, including: For each three-phase input current sequence, the normalized input current corresponding to each three-phase input current in the sequence is calculated based on the normalized input current formula, thus obtaining the parameter normalized sequence corresponding to the three-phase input current sequence; wherein, the normalized input current formula is: Among them, i k This represents the instantaneous value of the three-phase input current. The normalized input currents corresponding to the three-phase input currents, k = A, B, C, U k U is the effective value of the phase voltage at the input terminal. o For the output voltage, I o For output current; For the output current sequence, the normalized output current corresponding to each output current in the output current sequence is calculated based on the normalized output current formula, thus obtaining the parameter normalized sequence corresponding to the output current sequence; wherein, the normalized output current formula is: Among them, I o For output current, I is the normalized output current corresponding to the output current. oref This represents the desired value of the output current. For each operating temperature sequence, the normalized operating temperature corresponding to each operating temperature in the sequence is calculated based on the normalized operating temperature formula, thus obtaining the parameter normalized sequence corresponding to the operating temperature sequence; wherein, the normalized operating temperature formula is: Among them, T i The operating temperature of the transistor. This refers to the normalized operating temperature of the transistor. This represents the average operating temperature of the transistor.

3. The charging module state recognition method based on sample entropy according to claim 1, characterized in that, The step of performing empirical mode decomposition on each parameter normalized sequence to obtain multiple parameter component sequences of the charging module includes: For each parameter normalization sequence, empirical mode decomposition is performed to obtain multiple IMF components corresponding to that parameter normalization sequence; For each parameter normalization sequence, the first n IMF components are selected from the multiple IMF components corresponding to the parameter normalization sequence, and these are used as the multiple parameter component sequences corresponding to the parameter normalization sequence; where n is a preset value.

4. The charging module state recognition method based on sample entropy according to claim 1, characterized in that, The step of calculating the sample entropy of each parameter component sequence to obtain multiple parameter sample entropies of the charging module includes: For each parameter component sequence, the sample entropy of that parameter component sequence is calculated based on the sample entropy formula; wherein, the sample entropy formula is: d[X m (i),X m (j)]=max[|x(i+k)-x(j+k)|] Among them, En sample (m,r) represents the sample entropy of the component sequence, m is the dimension of the component sequence, r is the similarity tolerance, and B... m (r) represents the probability that the component sequence matches m points with the component sequence after the dimension is increased to m+1. X is the number of elements in a component sequence whose distance is less than the similarity tolerance. m (i) is the i-th element of the component sequence, d[X m (i),X m (j)] represents element X m (i) with element X m The distance of (j).

5. The charging module state recognition method based on sample entropy according to claim 1, characterized in that, Before calculating the sample entropy of each parameter component sequence based on the sample entropy formula, the method further includes: For each parameter component sequence, a portion of a preset time period is extracted as a subsequence of that parameter component sequence; Accordingly, the step of calculating the sample entropy of each parameter component sequence based on the sample entropy formula includes: For each parameter component sequence, the sample entropy of the subsequence of that parameter component sequence is calculated based on the sample entropy formula, and used as the sample entropy of that parameter component sequence.

6. The charging module state recognition method based on sample entropy according to claim 1, characterized in that, The acquisition of multiple operating performance parameter sequences of the charging module includes: The system continuously acquires multiple performance parameter sequences of the charging module over a preset time period.

7. The charging module state recognition method based on sample entropy according to claim 1, characterized in that, Before inputting the fault feature vector into the trained state recognition model, the method further includes: Obtain the initial state recognition model; The initial state recognition model is trained using a training sample set to obtain a trained state recognition model; wherein, the training sample set includes multiple training samples, each training sample is a set of fault feature vectors, and the label of each training sample is a fault state code.

8. The charging module state recognition method based on sample entropy according to claim 1, characterized in that, The fault states of the charging module include no-risk level, low-risk level, and high-risk level.

9. A charging module state recognition device based on sample entropy, characterized in that, include: The acquisition module is used to acquire multiple operating performance parameter sequences of the charging module; wherein each operating performance parameter sequence is the value of an operating performance parameter at multiple times. The normalization module is used to normalize each working performance parameter in each working performance parameter sequence to obtain multiple parameter normalization sequences of the charging module. The mode decomposition module is used to perform empirical mode decomposition on each parameter normalization sequence to obtain multiple parameter component sequences of the charging module. The sample entropy calculation module is used to calculate the sample entropy of each parameter component sequence separately, obtain multiple parameter sample entropies of the charging module, and combine the sample entropies of each parameter as the fault feature vector of the charging module. The status recognition module is used to input the fault feature vector into the trained status recognition model to obtain the fault status of the charging module.

10. The charging module state recognition device based on sample entropy according to claim 9, characterized in that, The operating performance parameters include a sequence of three-phase input current, an output current, and an operating temperature. The operating performance parameter corresponding to each three-phase input current sequence is one phase of the three-phase input current, and the operating performance parameter corresponding to each operating temperature sequence is the operating temperature of a transistor. The normalized sequence module is specifically used for: For each three-phase input current sequence, the normalized input current corresponding to each three-phase input current in the sequence is calculated based on the normalized input current formula, thus obtaining the parameter normalized sequence corresponding to the three-phase input current sequence; wherein, the normalized input current formula is: Among them, i k This represents the instantaneous value of the three-phase input current. The normalized input currents corresponding to the three-phase input currents, k = A, B, C, U k U is the effective value of the phase voltage at the input terminal. o For the output voltage, I o For output current; For the output current sequence, the normalized output current corresponding to each output current in the output current sequence is calculated based on the normalized output current formula, thus obtaining the parameter normalized sequence corresponding to the output current sequence; wherein, the normalized output current formula is: Among them, I o For output current, I is the normalized output current corresponding to the output current. oref This represents the desired value of the output current. For each operating temperature sequence, the normalized operating temperature corresponding to each operating temperature in the sequence is calculated based on the normalized operating temperature formula, thus obtaining the parameter normalized sequence corresponding to the operating temperature sequence; wherein, the normalized operating temperature formula is: Among them, T i The operating temperature of the transistor. This refers to the normalized operating temperature of the transistor. This represents the average operating temperature of the transistor.

Images