Fault test method, system and device suitable for fan converter

Abstract

The application relates to the technical field of electrical sensor measurement, and discloses a fault test method, system and equipment suitable for a fan converter, which comprises the following steps: collecting current data in an IGBT module of the fan converter and taking the current data as original current signals; screening a plurality of main frequencies of the original current signals; decomposing the original current signals to obtain a plurality of component signals and corresponding main frequency signals; screening a reference signal; obtaining variation abnormality of the component signals according to variation trend differences between the component signals and the reference signal; analyzing energy differences between the component signals and the reference signal in a time-frequency energy matrix to obtain energy difference indexes of the component signals; obtaining abnormality indexes of the component signals; and performing fault detection on the fan converter based on the abnormality indexes of the component signals. The application aims to solve the problem that high-frequency electromagnetic noise can affect signal performance of a current sampling loop and cover real fault characteristics.

Description

Technical Field

[0001] This invention relates to the field of electrical sensor measurement technology, and specifically to fault testing methods, systems and equipment applicable to wind turbine converters. Background Technology

[0002] As a core component of wind power generation systems, wind turbine converters are responsible for converting the alternating current (AC) generated by the wind turbine into direct current (DC) or synchronous AC that conforms to grid standards. Their performance directly impacts the efficiency and safety of the entire system. With the continuous increase in wind power generation capacity, the risk of converter failures also rises. Common failures include overload, short circuit, abnormal temperature, and electrical interference. Existing methods combine real-time monitoring, data acquisition, and intelligent algorithms. By establishing an operating model of the converter, its working status is monitored in real time. The system uses sensors to collect data such as current, voltage, and temperature, and uses data analysis techniques, such as machine learning and signal processing, to identify potential fault modes in real time. Furthermore, fault testing systems often possess self-learning capabilities, enabling them to optimize fault diagnosis algorithms based on historical fault data, improving the accuracy and response speed of fault detection.

[0003] In existing technologies, fault detection of wind turbine converters mainly involves analyzing changes in current signals. By analyzing the current state change patterns, it is possible to determine whether there are any abnormalities during the operation of the wind turbine converter. However, since the current changes of wind turbine converters exhibit multiple modes, and the high-frequency electromagnetic noise generated by the switching action of the converter's IGBTs can couple into the current sampling circuit, resulting in spikes and glitches in the current waveform (with amplitudes exceeding 20% ​​of the normal value), these spikes can mask the true fault characteristics. This interferes with the analysis results of the current signal change patterns, thereby affecting the accuracy of the fault monitoring results of the wind turbine converter. Summary of the Invention

[0004] This invention provides a fault testing method, system, and equipment suitable for wind turbine converters, to solve the problem that high-frequency electromagnetic noise can affect the signal performance of the current sampling circuit and mask the true fault characteristics. The specific technical solution adopted is as follows:

[0005] This invention proposes a fault testing method suitable for wind turbine converters, which includes the following steps:

[0006] Collect current data from the IGBT module of the wind turbine converter and use it as the raw current signal;

[0007] Analyze the amplitude performance of each frequency in the spectrum of the original current signal to select several dominant frequencies of the original current signal; decompose the original current signal to obtain several component signals and correspond them with each dominant frequency to obtain several dominant frequency signals; analyze the differences between each dominant frequency signal and the original current signal to select reference signals.

[0008] Based on the difference in the variation trend between each component signal and the reference signal, the variation anomaly of each component signal is obtained; the energy difference between each component signal and the reference signal in the time-frequency energy matrix is ​​analyzed to obtain the energy difference index of each component signal; and the anomaly index of each component signal is obtained by combining the variation anomaly of the component signals and the energy difference index.

[0009] Fault detection of wind turbine converters is performed based on the anomaly index of each component signal.

[0010] Optionally, the specific method for analyzing the amplitude characteristics of each frequency in the spectrum of the original current signal and filtering several dominant frequencies of the original current signal includes:

[0011] Based on the amplitude performance of each frequency in the spectrum of the original current signal and the amplitude difference between adjacent frequencies, the probability of the dominant frequency of each frequency is obtained.

[0012] The frequencies in the spectrum of the original current signal are arranged in descending order of their dominant frequency probability to obtain a dominant frequency probability sequence. The dominant frequency probability sequence is then clustered by hierarchical clustering to obtain several clusters.

[0013] The mean of the frequency probabilities of all frequencies in any cluster is obtained as the average frequency probability of that cluster; all frequencies in the cluster with the highest average frequency probability are taken as several frequencies of the original current signal.

[0014] Optionally, the specific method for obtaining the probability of the dominant frequency for each frequency is as follows:

[0015]

[0016] in, Indicates the first The possible frequencies of the main frequency. This represents the number of frequencies in the spectrum of the original current signal. Indicates the first The amplitude of a frequency in the spectrum Indicates the first The amplitude of a frequency in the spectrum and They represent the first Passing the exam The amplitude of each frequency in the spectrum; Represents the absolute value function. This represents an exponential function with the natural constant as its base.

[0017] Optionally, the specific method for obtaining several main frequency signals includes:

[0018] Construct a bipartite graph by taking the frequencies corresponding to each component signal as left nodes and each main frequency as right nodes. Use the difference between the frequency of the left node and the main frequency of the right node as the edge between the left and right nodes. Perform KM matching on the bipartite graph. The matching rule is that the shorter the edge value, the better the match. Obtain the matched frequency and the corresponding component signal for each main frequency. Take the component signal corresponding to the matched frequency of each main frequency as the main frequency signal of each main frequency.

[0019] Optionally, the specific method for analyzing the differences between each main frequency signal and the original current signal to obtain a reference signal includes:

[0020] For the The first main frequency signal is used to obtain the standard deviation of all its amplitudes, which is then used as the first... The amplitude standard deviation of the main frequency signal is obtained, and the amplitude standard deviation of the original current signal is also obtained; and the amplitude standard deviation of the first main frequency signal is also obtained. The DTW distance between the main frequency signal and the original current signal; Reference salience of the main frequency signal The calculation method is as follows:

[0021]

[0022] in, Indicates the first Each main frequency signal corresponds to a possible main frequency. Indicates the first The DTW distance between the main frequency signal and the original current signal L. Indicates the first The standard deviation of the amplitude of the main frequency signal, This represents the standard deviation of the amplitude of the original current signal L; Represents the absolute value function. Represents an exponential function with the natural constant as its base;

[0023] Obtain the reference salience of each main frequency signal, and take the main frequency signal with the maximum reference salience as the reference signal.

[0024] Optionally, the specific method for obtaining the anomalies in the changes of each component signal includes:

[0025] The reference signal is divided into segments according to the window length, resulting in several data segments. For any data segment of the reference signal, a straight line is fitted using the least squares method, and the fitted slope is obtained as the slope of change for that data segment. For the ... The component signal is used to acquire several data segments based on the window length, and the slope of change for each data segment is obtained; Abnormality of changes in component signals The calculation method is as follows:

[0026]

[0027] in, This indicates the number of data segments in the reference signal and each component signal; Indicates reference signal The Middle The slope of the change in the data segment; Indicates the first of the component signals The slope of the change in the data segment; This represents the absolute value function.

[0028] Optionally, the specific method for obtaining the energy difference index of each component signal includes:

[0029] By using short-time Fourier transform, the reference signal and the... The time-frequency energy matrix is ​​obtained from each component signal. Energy difference index of component signals The calculation method is as follows:

[0030]

[0031] in, This indicates the number of time points corresponding to the reference signal and each component signal; Indicates reference signal The time-frequency energy matrix of the first Elements at each point in time; Indicates the first The time-frequency energy matrix of the reference signal of the first line Elements at each point in time; This represents the absolute value function.

[0032] Optionally, the specific method for obtaining the anomaly index of each component signal includes:

[0033] Using the reference weight as the weight for anomalies and the difference weight as the weight for the energy difference index, the results are obtained by applying the first... The weighted sum of the variation anomalies and energy difference coefficients of the component signals is used as the result of the first weighted sum. Anomaly index of component signals.

[0034] This invention also proposes a fault testing system suitable for wind turbine converters, the system comprising:

[0035] The current signal acquisition module is used to acquire current data from the IGBT module of the wind turbine converter and use it as the raw current signal.

[0036] The current signal analysis module is used to analyze the amplitude performance of each frequency in the spectrum of the original current signal, and to filter out several main frequencies of the original current signal; to decompose the original current signal into several component signals, and to correspond them with each main frequency to obtain several main frequency signals; to analyze the differences between each main frequency signal and the original current signal, and to filter out reference signals.

[0037] Based on the difference in the variation trend between each component signal and the reference signal, the variation anomaly of each component signal is obtained; the energy difference between each component signal and the reference signal in the time-frequency energy matrix is ​​analyzed to obtain the energy difference index of each component signal; and the anomaly index of each component signal is obtained by combining the variation anomaly of the component signals and the energy difference index.

[0038] The converter fault detection module is used to detect faults in the wind turbine converter based on the anomaly index of each component signal.

[0039] The present invention also proposes a fault testing device suitable for wind turbine converters, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of the above method.

[0040] The beneficial effects of this invention are as follows: This invention analyzes the amplitude performance of the original current signal spectrum, filters the dominant frequency in each frequency range, and, combined with the component signals obtained from EMD decomposition, obtains several dominant frequency signals. These dominant frequency signals reflect the various change modes generated by the IGBT module. Based on the similarity of the overall change trend between the dominant frequency signals and the original current signal, the amplitude fluctuation, and the probability of the dominant frequency, a reference signal is selected to reflect the regular change trend of the original current signal without being affected by high-frequency noise, providing a basis for subsequent anomaly analysis and comparison of component signals. Based on the reference signal, the local change trends of each component signal are compared with the reference signal. The analysis of the time-frequency energy matrix and the local change trend can reflect the stability of the local change slope and the distribution law of the extreme points, accurately capturing the micro-waveform distortion caused by IGBT faults. The time-frequency energy matrix reflects the abnormal concentration or loss of energy in the time domain caused by the fault, thus comprehensively quantifying the abnormality index of the component signal. By identifying the anomalies in the current signal of the wind turbine converter IGBT module, and reducing the influence of the high-frequency noise coupling current loop generated by the switching action, the waveform changes generated by various faults in the current signal are analyzed, so as to monitor the faults of various current change modes respectively, and finally realize the accurate fault testing of the wind turbine converter. Attached Figure Description

[0041] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0042] Figure 1 This is a schematic diagram of a fault testing method for wind turbine converters provided in one embodiment of the present invention;

[0043] Figure 2 The diagram below shows a structural block diagram of a fault testing system for wind turbine converters, provided as another embodiment of the present invention. Detailed Implementation

[0044] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0045] Please see Figure 1 The diagram illustrates a fault testing method for wind turbine converters according to an embodiment of the present invention, which includes the following steps:

[0046] Step S001: Collect the current data in the IGBT module of the wind turbine converter and use it as the raw current signal.

[0047] The purpose of this embodiment is to identify abnormalities in the current by monitoring the current in the IGBT switch of the wind turbine converter, thereby judging the fault of the wind turbine converter. Therefore, it is necessary to collect current data first.

[0048] Specifically, this embodiment uses a Hall effect closed-loop sensor. Specifically, a closed-loop Hall current sensor (accuracy ±0.2%, bandwidth ≥200kHz) is directly connected in series with the IGBT module output circuit to measure the collector current (±2000A range) in real time. Heat loss is reduced through a primary-side through-hole design. Simultaneously, to capture the IGBT switching transients (≤2μs), the sensor response time must be <100ns, and a magnetically shielded shell is required to suppress 25kHz switching electromagnetic interference (EMI). This allows for real-time acquisition of current data from the IGBT module. The sampling time interval is determined by the Hall current sensor settings, which will not be elaborated upon in this embodiment. The current data constitutes the original current signal.

[0049] It should be noted that the wind turbine converter is the core component of the wind power generation system, mainly undertaking the key functions of power conversion and grid connection control. Its role is to convert the unstable AC power output from the wind turbine (such as 0~50Hz frequency conversion power) into stable AC power (such as 50Hz / 380V or 690V) that matches the grid through power electronic conversion such as rectification and inversion. At the same time, it achieves maximum power point tracking (MPPT) to improve power generation efficiency, and has grid adaptability functions such as low voltage ride-through (LVRT) and harmonic suppression to ensure the safe and efficient grid-connected operation of the wind turbine. During operation, the wind turbine converter may experience various faults, mainly including IGBT module failures (such as open circuit / short circuit of the switch, abnormal drive signal), which can lead to current waveform distortion or overcurrent burnout.

[0050] It should be further explained that the IGBT module is the core power switching device of the wind turbine converter. Its faults mainly manifest as open circuit, short circuit, and abnormal drive signal. Open circuit faults will cause the current path to be interrupted, resulting in output current loss or three-phase imbalance (such as a sudden drop in current of more than 50% in a certain phase). Short circuit faults will cause instantaneous overcurrent (up to 10 times the rated value), which will burn out the module if the protection is not timely. Abnormal drive signal (such as pulse loss or timing error) will cause the IGBT switching to be out of sync, resulting in the output voltage harmonic distortion rate (THD) exceeding 5% or even causing bridge arm shoot-through. By analyzing current data and extracting waveform characteristics, IGBT module faults can be monitored. When an IGBT has an open circuit fault, the three-phase current will be significantly unbalanced, and the current data will show obvious distortion characteristics. Short circuit faults manifest as instantaneous overcurrent (and high-frequency oscillation harmonics). Therefore, by analyzing the changes in current data, the current status of the IGBT module can be determined, and then the presence of abnormalities in the IGBT module can be analyzed.

[0051] Step S002: Analyze the amplitude performance of each frequency in the spectrum of the original current signal and select several main frequencies of the original current signal; decompose the original current signal to obtain several component signals and correspond them with each main frequency to obtain several main frequency signals; analyze the differences between each main frequency signal and the original current signal and select reference signals.

[0052] It should be noted that when analyzing current data, because current data exhibits multiple variation modes, and different variation modes present different data characteristics, in order to accurately analyze the current data, Fourier transform is used to obtain the spectrum of the original current data, and then the dominant frequency fluctuation range of the current data is obtained. The dominant frequency in the spectrum usually shows a large amplitude, and the amplitude difference between dominant frequencies is small. This is used to quantify the probability of the dominant frequency of each signal in the spectrum, and cluster them based on the probability of the dominant frequency. Based on the clustering results, dominant frequency is selected, which provides a basis for the corresponding dominant frequency signal in the component signals after EMD decomposition of the original current signal.

[0053] Preferably, in one embodiment of the present invention, analyzing the amplitude characteristics of each frequency in the spectrum of the original current signal and filtering several dominant frequencies of the original current signal includes the following specific method:

[0054] Performing a Fourier transform on the original current signal yields its spectrum, which contains several frequencies. The amplitude of each frequency in the spectrum is then calculated. The main frequency of each frequency is possible. The calculation method is as follows:

[0055]

[0056] in, This represents the number of frequencies in the spectrum of the original current signal. Indicates the first The amplitude of a frequency in the spectrum Indicates the first The amplitude of a frequency in the spectrum and They represent the first Passing the exam The amplitude of each frequency in the spectrum; Represents the absolute value function. This embodiment uses an exponential function with the natural constant as its base. The model is used to represent the inverse proportional relationship and for normalization processing. As input to the model, implementers can set inverse proportional functions and normalization functions according to the actual situation.

[0057] It should be noted that the larger the amplitude of the current frequency accounts for in the entire signal spectrum, the greater the likelihood that it corresponds to the dominant frequency. At the same time, the smaller the amplitude difference between the current frequency and the adjacent frequencies, the larger the proportion of the current frequency and the adjacent frequencies in the original current signal, and the more likely it is to be the dominant frequency signal.

[0058] Furthermore, the frequencies in the spectrum of the original current signal are arranged in descending order of their dominant frequency probability to obtain a dominant frequency probability sequence. This sequence is then clustered using hierarchical clustering to obtain several one-dimensional clusters. In this embodiment, the clustering maturity is set to 2, resulting in several clusters. The mean of the dominant frequency probability of all frequencies in any cluster is obtained as the average dominant frequency probability of that cluster. All frequencies in the cluster with the highest average dominant frequency probability are taken as several dominant frequencies of the original current signal.

[0059] It should be further explained that the dominant frequency represents the fluctuations in the current signal caused by various interferences. The more complex the frequency changes, the more patterns of change in the current signal. During the operation of the wind turbine converter, the power changes generated by the IGBT module are caused by multiple factors, resulting in multiple states in the current data. Therefore, in order to accurately analyze the current change state, it is necessary to analyze the causes of the current changes. In order to analyze the current signal under multiple interferences, multiple IMF components are obtained through EMD decomposition. Different IMF components represent the signal change state of different frequencies. The earlier the component is decomposed in the EMD process, the higher its frequency, and the greater the possibility that it is a noise signal. It is more likely to be high-frequency electromagnetic noise generated by the switching action of the converter IGBT. Therefore, it is necessary to map the dominant frequency to the IMF components to obtain the dominant frequency signal, and further analyze the difference between it and the original current signal to obtain a reference signal that better highlights the change trend of the original current signal.

[0060] Preferably, in one embodiment of the present invention, the original current signal is decomposed to obtain several component signals, and corresponding with each main frequency to obtain several main frequency signals. The specific method includes:

[0061] The original current signal is decomposed into several IMF components using EMD, with each IMF component serving as a component signal. Each component signal corresponds to a frequency. A bipartite graph is constructed using the frequency corresponding to each component signal as the left node and each dominant frequency as the right node. The difference between the frequency of the left node and the dominant frequency of the right node (the absolute value of the frequency difference) is used as the edge between the left and right nodes. KM matching is performed on the bipartite graph, with the matching rule being that the shorter the edge value, the better the match. For each dominant frequency, the matched frequency and its corresponding component signal are obtained. The component signal corresponding to the matched frequency of each dominant frequency is taken as the dominant frequency signal of that dominant frequency. It should be noted that there may be a difference in the number of left and right nodes in the bipartite graph, i.e., the number of dominant frequencies is different from the number of component signals. If the number of dominant frequencies is greater than the number of component signals, only the dominant frequency signal of the successfully matched dominant frequency is obtained. If the number of component signals is greater than the number of dominant frequencies, only the component signal of the successfully matched frequency is taken as the corresponding dominant frequency signal.

[0062] It should be noted that after obtaining the main frequency signal, it is necessary to obtain a reference signal that can better represent the changing trend of the original current signal. Based on the analysis of the similarity between the overall changing trends of the main frequency signal and the original current signal, further consideration should be given to the probability of the main frequency and the differences in amplitude fluctuation between the main frequency signal and the original current signal. The greater the probability of the main frequency, the more features of the original current signal are contained in the main frequency signal, the closer the overall changing trend is, and the smaller the difference in amplitude fluctuation, the better the main frequency signal can reflect the regular changing trend of the original current signal, and it is not affected by high-frequency noise. Therefore, it is more likely to be used as a reference signal for subsequent comparative analysis.

[0063] Preferably, in one embodiment of the present invention, the method for analyzing the differences between each main frequency signal and the original current signal to obtain a reference signal includes:

[0064] For the The first main frequency signal is used to obtain the standard deviation of all its amplitudes, which is then used as the first... The amplitude standard deviation of the main frequency signal is obtained, and the amplitude standard deviation of the original current signal is also obtained; and the amplitude standard deviation of the first main frequency signal is also obtained. The DTW distance between the main frequency signal and the original current signal; Reference salience of the main frequency signal The calculation method is as follows:

[0065]

[0066] in, Indicates the first Each main frequency signal corresponds to a possible main frequency. Indicates the first The DTW distance between the main frequency signal and the original current signal L. Indicates the first The standard deviation of the amplitude of the main frequency signal, This represents the standard deviation of the amplitude of the original current signal L; Represents the absolute value function. This embodiment uses an exponential function with the natural constant as its base. The model is used to represent the inverse proportional relationship and for normalization processing. As input to the model, implementers can set inverse proportional functions and normalization functions according to the actual situation.

[0067] Furthermore, the reference salience of each main frequency signal is obtained according to the above method, and the main frequency signal corresponding to the maximum reference salience is taken as the reference signal.

[0068] Thus, by analyzing the amplitude performance of the original current signal spectrum, the dominant frequency in each frequency range is selected. Combined with the component signals obtained from EMD decomposition, several dominant frequency signals are obtained to reflect the various change modes generated by the IGBT module. Based on the similarity of the overall change trend between the dominant frequency signal and the original current signal, the amplitude fluctuation, and the possibility of the dominant frequency, reference signals are selected to reflect the regular change trend of the original current signal without being affected by high-frequency noise, providing a basis for subsequent anomaly analysis and comparison of component signals.

[0069] Step S003: Based on the difference in the variation trend between each component signal and the reference signal, obtain the variation anomaly of each component signal; analyze the energy difference between each component signal and the reference signal in the time-frequency energy matrix to obtain the energy difference index of each component signal; combine the variation anomaly of the component signals and the energy difference index to obtain the anomaly index of each component signal.

[0070] It should be noted that as a core power switching device, the switching action of the IGBT module directly affects the waveform characteristics of the output current. When the IGBT is working normally, the local fluctuation trend of the original current signal (such as the rise and fall slope, the distribution of extreme points, etc.) will show regular changes and remain consistent with the reference signal. However, once a fault occurs (such as open circuit, short circuit, or abnormal drive), the regularity will be disrupted, causing the local change trend to deviate significantly.

[0071] It should be further noted that when an IGBT operates at a fixed switching frequency (e.g., 2kHz) under PWM (Pulse Width Modulation) drive, the current signal exhibits the following behavior: when the IGBT is turned on, the current increases linearly with the DC bus voltage, and the slope is determined by the load inductance. L The voltage difference determines the current flow; when the IGBT is turned off, the current flows through the freewheeling diode, and the slope is reversed; therefore, different change patterns will be reflected in the decomposed signal, that is, each component signal. By calculating the change relationship between the reference signal and each component signal, the more similar the overall change trend formed by the local fluctuation trend is, the more consistent the overall trend is, thus quantifying the change anomaly.

[0072] Preferably, in one embodiment of the present invention, the abnormality of the change of each component signal is obtained based on the difference in the change trend between each component signal and the reference signal, including the following specific method:

[0073] A preset window length is used; in this embodiment, a window length of 5 is used. The reference signal is segmented according to this window length to obtain several data segments. Each data segment contains 5 data points from the reference signal, and there is no overlap between the data segments. It is worth noting that if the number of remaining data points is insufficient to form a data segment, the remaining data points are directly treated as a single data segment for subsequent processing. A linear fit is performed on any data segment of the reference signal using the least squares method, and the fitted slope is used as the slope of change for that data segment. Similarly, for the... The component signal is used to acquire several data segments based on the window length, and the slope of change of each data segment is obtained; then the first... Abnormality of changes in component signals The calculation method is as follows:

[0074]

[0075] in, This indicates the number of data segments in the reference signal and each component signal (the reference signal is also a component signal, and if the component signals are of equal length, the number of data segments is the same). Indicates reference signal The Middle The slope of the change in the data segment; Indicates the first of the component signals The slope of the change in the data segment; This represents the absolute value function.

[0076] It should be noted that by summing the differences in the slope of the same data segment and obtaining the ratio with the sum of the values ​​of the slopes, the closer the ratio is to 0, the closer the slopes of the same data segment are, the higher the consistency of the trend, and the smaller the abnormality of the corresponding component signal. However, IGBT failure causes a sudden change in current, which increases the difference in the slope of the change, making the ratio closer to 1, and the greater the abnormality of the change.

[0077] It should be further explained that by analyzing the stability of the local change slope of the current signal and the distribution law of extreme points, the micro-waveform distortion caused by IGBT faults can be accurately captured; then, by quantifying the difference in slope (change trend) between the actual current signal (each component signal) and the reference signal in the local time window, it can be determined whether the IGBT is working properly.

[0078] Preferably, in one embodiment of the present invention, the energy difference between each component signal and the reference signal in the time-frequency energy matrix is ​​analyzed to obtain the energy difference index of each component signal, including the following specific method:

[0079] It should be noted that when an IGBT malfunctions (such as open circuit, short circuit, or drive abnormality), the energy distribution of the current signal in the time and frequency domains will change significantly. Open circuit fault: energy concentrates at lower frequencies (due to the lack of high-frequency switching action components); Short circuit fault: energy suddenly increases in the high-frequency range (due to transient overcurrent and resonant oscillation); Drive abnormality: energy exhibits abnormal peaks in the switching frequency sideband (e.g., 2kHz±50Hz). Therefore, further analysis of the component signal and reference signal distribution in the time and frequency domains is necessary. Based on the time-frequency energy matrix, with the frequency domains of the component and reference signals fixed, the energy difference between the reference and component signals in the time domain is analyzed. The time-frequency energy matrix represents the signal's energy distribution over time. and frequency Energy density at a given location; in wind turbine converter fault detection, the time-frequency energy matrix reveals how the energy of the current signal dynamically distributes with time and frequency, quantifies the intensity of each time component, and the fault will cause abnormal concentration or absence of energy in a specific time period under a fixed frequency band.

[0080] Specifically, through short-time Fourier transform, the reference signal and the... The time-frequency energy matrix is ​​obtained for each component signal, then the... Energy difference index of component signals The calculation method is as follows:

[0081]

[0082] in, This indicates the reference signal and the number of time points corresponding to each component signal (if each component signal is of equal length, then the number of time points is equal). Indicates reference signal The time-frequency energy matrix of the first Elements at each time point (the frequency of the reference signal is fixed, and the elements are determined according to the time); Indicates the first The time-frequency energy matrix of the reference signal of the first line Elements at each point in time; This represents the absolute value function.

[0083] It should be noted that the sum of the accumulated energy differences between the component signal and the reference signal in the main frequency band is used. The larger the sum of the differences, the greater the energy shift caused by the fault. The total energy of the reference signal in the main frequency band is then normalized.

[0084] Preferably, in one embodiment of the present invention, the abnormality index of each component signal is obtained by combining the abnormality of the component signal changes and the energy difference index, including the following specific method:

[0085] For the Single component signal, preset reference weight This embodiment adopts To narrate; to As the difference weight, the reference weight is used as the weight for the anomaly of change, and the difference weight is used as the weight for the energy difference index, through the first... The weighted sum of the variation anomalies and energy difference coefficients of the component signals is used as the result of the first weighted sum. Anomaly index of component signals.

[0086] It should be noted that, since the dispersion and concentration of energy are more significant in indicating the degree of fault manifestation than regular trends, more weight is assigned to the energy difference index to obtain the anomaly index of the component signals.

[0087] Based on the reference signal, the local variation trends of each component signal and the reference signal are compared, and the time-frequency energy matrix is ​​analyzed. The local variation trend can reflect the stability of the local variation slope and the distribution law of the extreme points, accurately capturing the micro-waveform distortion caused by IGBT faults. The time-frequency energy matrix reflects the abnormal concentration or loss of energy in the time domain caused by the fault, thereby comprehensively quantifying the abnormality index of the component signals.

[0088] Step S004: Based on the anomaly index of each component signal, perform fault detection on the wind turbine converter.

[0089] Specifically, in this embodiment, the IGBT module is set with fault-based thresholds: an open-circuit threshold of 0.8, a short-circuit threshold of 0.5, and a drive anomaly threshold of 0.3. For any component signal, if its anomaly index is greater than the open-circuit threshold, then the wind turbine converter has an IGBT module open-circuit fault; if its anomaly index is within the open-circuit threshold, then the wind turbine converter has an IGBT module open-circuit fault. If the range is within a certain range, then the wind turbine converter has an IGBT module short-circuit fault. If its abnormality index is within a certain range... If the range is within the specified range, then the wind turbine converter has an abnormal drive signal fault. If its abnormality index is less than or equal to the drive abnormality threshold, then the current change mode corresponding to this component signal does not have a fault.

[0090] Furthermore, for each component signal of the original current signal, if at least one component signal has an anomaly index within its threshold range under each fault, then the wind turbine converter has a corresponding fault. If the anomaly index of all component signals is less than or equal to the drive anomaly threshold, then the wind turbine converter has no fault.

[0091] It should be noted that each component signal reflects multiple current change modes of the IGBT module corresponding to the original current signal. By performing fault judgment on the abnormality index of each current change mode, the fault detection of the wind turbine converter can be carried out comprehensively.

[0092] Thus, by identifying anomalies in the current signal of the wind turbine converter IGBT module and reducing the impact of high-frequency noise coupling current loops generated by switching actions, the waveform changes in the current signal caused by various faults can be analyzed to monitor faults in multiple current change modes, ultimately achieving accurate fault testing of the wind turbine converter.

[0093] Please see Figure 2 This illustrates a structural block diagram of a fault testing system for wind turbine converters provided by another embodiment of the present invention. The system includes:

[0094] Current signal acquisition module 101: Acquires current data from the IGBT module of the wind turbine converter and uses it as the raw current signal;

[0095] Current signal analysis module 102: Analyzes the amplitude performance of each frequency in the spectrum of the original current signal, and selects several main frequencies of the original current signal; decomposes the original current signal to obtain several component signals, and corresponds them with each main frequency to obtain several main frequency signals; analyzes the difference between each main frequency signal and the original current signal, and selects reference signals.

[0096] Based on the difference in the variation trend between each component signal and the reference signal, the variation anomaly of each component signal is obtained; the energy difference between each component signal and the reference signal in the time-frequency energy matrix is ​​analyzed to obtain the energy difference index of each component signal; and the anomaly index of each component signal is obtained by combining the variation anomaly of the component signals and the energy difference index.

[0097] Converter fault detection module 103: Based on the anomaly index of each component signal, it performs fault detection on the wind turbine converter.

[0098] Another embodiment of the present invention provides a fault testing device suitable for wind turbine converters, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the above-described method steps S001 to S004.

[0099] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method of fault testing suitable for use in a fan inverter, characterised in that, The method comprises the following steps: Collecting current data in an IGBT module of a fan converter, and taking the current data as an original current signal; Analyzing amplitude performance of each frequency in a spectrum of the original current signal, and screening several main frequencies of the original current signal; decomposing the original current signal to obtain several component signals, and corresponding to each main frequency to obtain several main frequency signals; analyzing difference performance of each main frequency signal and the original current signal, and screening to obtain a reference signal; Obtaining change abnormality of each component signal according to difference in change trend of each component signal and the corresponding reference signal; analyzing energy difference of each component signal and the reference signal in a time-frequency energy matrix to obtain an energy difference index of each component signal; and comprehensively obtaining an abnormality index of each component signal according to the change abnormality and the energy difference index of the component signal; Performing fault detection on the fan converter based on the abnormality index of each component signal; The method for screening the several main frequencies of the original current signal comprises the following steps: Obtaining main frequency possibility of each frequency based on amplitude performance of each frequency in a spectrum of the original current signal and amplitude difference of adjacent frequencies; Arranging each frequency in the spectrum of the original current signal according to the main frequency possibility from large to small to obtain a main frequency possibility sequence, and performing hierarchical clustering on the main frequency possibility sequence to obtain several clusters; Obtaining a mean value of the main frequency possibility of all frequencies in any cluster as an average main frequency possibility of the cluster; and taking all frequencies in the cluster with the maximum average main frequency possibility as the several main frequencies of the original current signal; The method for screening to obtain the reference signal comprises the following steps: For the first order dominant frequency signal, the standard deviation of all amplitudes is obtained as the amplitude standard deviation of the first order dominant frequency signal, and the amplitude standard deviation of the original current signal is obtained; and the DTW distance between the first order dominant frequency signal and the original current signal is obtained; the reference saliency of the first order dominant frequency signal is calculated as follows: ； wherein, represents the likelihood of the dominant frequency of the th dominant frequency signal, represents the DTW distance between the th dominant frequency signal and the original current signal L, represents the standard deviation of the amplitude of the th dominant frequency signal, represents the standard deviation of the amplitude of the original current signal L; represents the absolute value function, represents the exponential function with base of natural constant; Taking a main frequency signal corresponding to a maximum value of reference highlight of each main frequency signal as the reference signal; The method for obtaining the change abnormality comprises the following steps: The reference signal is segmented by window length to obtain a plurality of data segments, any data segment of the reference signal is linearly fitted by least square method, and a fitting slope is obtained as a change slope of the data segment; the change slope of the first data segment is taken as a first change slope of the first component signal. The component signal is segmented according to the window length to obtain a plurality of data segments, and the change slopes of the data segments are obtained; the change slope of the first data segment is taken as a first change slope of the first component signal. The change abnormality of the component signal is calculated by the following formula: ​ ； wherein, denotes the number of data segments in the reference signal and in each component signal, respectively; denotes the reference signal the change slope of the th data segment in the reference signal; denotes the change slope of the th data segment in the th component signal; denotes the absolute value function; The method for obtaining the energy difference index comprises the following steps: The time-frequency energy matrixes of the reference signal and the first component signal are obtained respectively by short-time Fourier transform. The time-frequency energy matrixes of the reference signal and the first component signal are obtained respectively by short-time Fourier transform. The energy difference index of the first component signal is calculated by the following formula: The energy difference index of the first component signal is calculated by the following formula: ； wherein, denotes the number of time points corresponding to the reference signal and the respective component signal; denotes the reference signal at the element of the time-frequency energy matrix at the time point; denotes the element of the time-frequency energy matrix of the reference signal at the time point; denotes the absolute value function.

2. The method for fault testing of a variable frequency drive for a fan as claimed in claim 1, wherein, The method for obtaining the main frequency possibility of each frequency comprises the following steps: ； wherein represents the dominant frequency likelihood of the th frequency, represents the number of frequencies in the spectrum of the original current signal, represents the amplitude of the th frequency in the spectrum, represents the amplitude of the th frequency in the spectrum, and represent the amplitude of the th and th frequency in the spectrum, respectively; represents the absolute value function, represents the exponential function with base of the natural constant.

3. The method for fault testing of a variable frequency drive for a fan as claimed in claim 1, wherein, The method for obtaining the several main frequency signals comprises the following steps: Constructing a bipartite graph by taking frequencies corresponding to the component signals as left nodes and taking the main frequencies as right nodes, taking difference between the frequencies of the left nodes and the main frequencies of the right nodes as edges between the left and right nodes, performing KM matching on the bipartite graph, and taking a shorter edge value as a matching rule; obtaining matched frequencies and corresponding component signals for each main frequency; and taking the component signals corresponding to the matched frequencies of each main frequency as the main frequency signals of each main frequency.

4. The method for fault testing of a variable frequency drive for a fan as defined in claim 1, wherein, The method for obtaining the abnormality index of each component signal comprises the following steps: The reference weight is used as the weight of the change abnormality, the difference weight is used as the weight of the energy difference index, and the weighted sum of the change abnormality and the energy difference coefficient of the m-th component signal is obtained as the abnormality index of the m-th component signal. The weighted sum of the change abnormality and the energy difference coefficient of the m-th component signal is obtained as the abnormality index of the m-th component signal. The weighted sum of the change abnormality and the energy difference coefficient of the m-th component signal is obtained as the abnormality index of the m-th component signal.

5. A fault testing system for a fan inverter, characterized by, The system comprises: A current signal collecting module configured to collect current data in an IGBT module of a fan converter, and take the current data as an original current signal; A current signal analyzing module configured to analyze amplitude performance of each frequency in a spectrum of the original current signal, and screen several main frequencies of the original current signal; decompose the original current signal to obtain several component signals, and corresponding to each main frequency to obtain several main frequency signals; analyze difference performance of each main frequency signal and the original current signal, and screen to obtain a reference signal; According to the difference between the change trend of each component signal and the corresponding reference signal, the change abnormality of each component signal is obtained; the energy difference index of each component signal is obtained by analyzing the energy difference between each component signal and the reference signal in the time-frequency energy matrix; the abnormality index of each component signal is obtained by comprehensively considering the change abnormality and the energy difference index of each component signal; The converter fault detection module is used for detecting the fault of the fan converter based on the abnormality index of each component signal; The method for screening the several main frequencies of the original current signal is: Based on the amplitude of each frequency in the spectrum of the original current signal and the amplitude difference between adjacent frequencies, the main frequency possibility of each frequency is obtained; Arranging each frequency in the spectrum of the original current signal according to the main frequency possibility from large to small, obtaining the main frequency possibility sequence, and clustering the main frequency possibility sequence by hierarchical clustering to obtain several clusters; Obtaining the mean value of the main frequency possibility of all frequencies in any cluster as the average main frequency possibility of the cluster; all frequencies in the cluster with the maximum average main frequency possibility are regarded as the several main frequencies of the original current signal; The method for screening the reference signal is: For the first order dominant frequency signal, the standard deviation of all amplitudes is obtained as the amplitude standard deviation of the first order dominant frequency signal, and the amplitude standard deviation of the original current signal is obtained; and the DTW distance between the first order dominant frequency signal and the original current signal is obtained; the reference saliency of the first order dominant frequency signal is calculated as follows: ； wherein, represents the pitch of the th pitch signal, represents the DTW distance between the th pitch signal and the original current signal L, represents the standard deviation of the amplitude of the th pitch signal, represents the standard deviation of the amplitude of the original current signal L; represents the absolute value function, represents the exponential function with base of natural constant; Obtaining the reference highlight of each main frequency signal, and taking the main frequency signal corresponding to the maximum reference highlight as the reference signal; The method for obtaining the change abnormality is: The reference signal is segmented by window length to obtain a plurality of data segments. Any data segment of the reference signal is linearly fitted by a least square method, and a fitting slope is obtained as a change slope of the data segment. The change slopes of the data segments of the first component signal are obtained according to the window length. The change abnormality of the first component signal is calculated by the following formula: The change abnormality of the first component signal is calculated by the following formula:​ ； wherein, denotes the number of data segments in the reference signal and in each component signal, respectively; denotes the reference signal the change slope of the th data segment in the reference signal; denotes the change slope of the th data segment in the th component signal; denotes the absolute value function; The method for obtaining the energy difference index is: The time-frequency energy matrixes of the reference signal and the first component signal are obtained respectively by short-time Fourier transform. The time-frequency energy matrixes of the reference signal and the first component signal are obtained respectively by short-time Fourier transform. The energy difference index of the first component signal is calculated by the following formula: The energy difference index of the first component signal is calculated by the following formula: ； wherein denotes the number of time points corresponding to the reference signal and the respective component signal; denotes the reference signal element of the time-frequency energy matrix of the time point of the denotes the element of the time-frequency energy matrix of the reference signal of the time point of the denotes the absolute value function.

6. A fault testing device for a fan inverter, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, The processor executes the computer program to realize the steps of the fault test method for the fan converter according to any one of claims 1-4.

Images