Over-current protection monitoring method and system for frequency converter

Abstract

The invention belongs to the field of circuit protection, and provides an overcurrent protection monitoring method and system for a frequency converter. The method comprises the following steps: continuously acquiring current time sequence data of three-phase current of each frequency converter in a workstation, sampling to form an original data set, and standardizing the original data set; fault segmentation is carried out on the standardized original data set, so that abnormal data and normal data in the original data set are separated; converting an original data set representing abnormal data from a time domain to a frequency domain based on discrete Fourier transform to obtain a new diagnosis data set; based on a principal component analysis technology, features in the diagnosis data set are extracted and dimensionally reduced to obtain a feature set, the square prediction error of each feature in the feature set is calculated, the total contribution degree of the features is obtained according to the influence of the square prediction error on other features, and whether a detected object fails or not is determined; therefore, the problem that in the prior art, irregular overcurrent faults which do not reach the threshold value are difficult to diagnose can be solved.

Description

Technical Field

[0001] This invention relates to the field of circuit protection, and in particular to an overcurrent protection monitoring method and system for frequency converters. Background Technology

[0002] A frequency converter is a power electronic device, also known as a variable frequency drive or variable frequency speed controller, used to regulate the speed of an AC motor. It controls the motor's speed and torque by adjusting the frequency and voltage of the input AC power. A frequency converter typically consists of a rectifier, inverter, filter, and control circuitry.

[0003] The inverter section of a frequency converter typically consists of a series of power semiconductor devices (such as transistors, power MOSFETs, IGBTs, etc.) and is equipped with control circuitry to control the switching states of these semiconductor devices. In industrial applications, inverter section failures are mainly caused by damage to the power switching transistors, such as burnout due to overcurrent.

[0004] Of course, most frequency converters in the current technology have overcurrent protection, which shuts down the system and protects the circuit when the current reaches the threshold. However, not every overcurrent will reach the threshold, and the overcurrent time is not fixed. Therefore, the frequency converter will work under overload conditions unpredictably. At the same time, such uncertain overcurrent faults are difficult to diagnose and it is difficult to find the root cause of the fault. Summary of the Invention

[0005] In view of the above technical problems, the present invention provides an overcurrent protection monitoring method and system for frequency converters to solve the problem of difficulty in diagnosing intermittent overcurrent faults that do not reach the threshold in the prior art.

[0006] Other features and advantages of this disclosure will become apparent from the following detailed description, or may be learned in part from practice of this disclosure.

[0007] According to one aspect of the present invention, an overcurrent protection monitoring method for a frequency converter is disclosed, the method comprising: The current time series data of the three-phase current of each inverter in the workstation is continuously acquired, sampled to form a raw dataset, and the raw dataset is standardized. Fault segmentation is performed on the standardized original dataset to separate abnormal data from normal data. Based on the discrete Fourier transform, the original dataset representing the abnormal data is transformed from the time domain to the frequency domain to obtain a new diagnostic dataset; Based on principal component analysis, features in the diagnostic dataset are extracted and dimensionality reduced to obtain a feature set. The squared prediction error of each feature in the feature set is calculated. The total contribution of the features is obtained based on the influence of the squared prediction error on other features. Based on the total contribution, it is determined whether the detected object is faulty.

[0008] Furthermore, after acquiring the current time series data, the current time series data is sampled at intervals with a fixed sampling length. In the Xth sampling, the interval between the starting position of the Xth sampling and the starting position of the (X-1)th sampling is less than the fixed sampling length.

[0009] Furthermore, standardization includes: Set an average error threshold and a standardization amplitude, calculate the average value of any set of current time series data in the dataset, if the average value is greater than the average error threshold, calculate the maximum value of the absolute value of the dataset, obtain a first standardization coefficient based on the maximum value of the absolute value of the dataset and the standardization amplitude, and calculate the standardized current time series data based on the first standardization coefficient. If the average value is less than the average error threshold, calculate the current component in each sampling period and the sampling amplitude of the current component in each sampling period. Based on the sampling amplitude and the standardized amplitude, obtain the second standardization coefficient. Based on the second standardization coefficient, calculate the standardized current time series data. Repeat the operation until each of the current time series data in the dataset is standardized.

[0010] Furthermore, after standardizing the dataset, the standardized dataset is preprocessed using linear filters and wavelet transform methods to remove noise.

[0011] Furthermore, during signal segmentation, a segmentation model function is established based on partial differential equations. The segmentation model function sets the original dataset to consist of normal data and abnormal data. The segmentation model function includes an evolution curve and an energy fitting function. Based on the segmentation model function and the Dirac function, the original dataset is segmented when the difference between normal data and abnormal data tends to a normal value.

[0012] Furthermore, when performing the Discrete Fourier Transform, it includes: Based on the number of samplings, sampling length, and defined frequency components during the sampling of the original dataset, the amplitude and phase of each frequency component are calculated. The frequency components are rearranged so that the frequency gradually increases from zero to obtain frequency domain data. The frequency domain data is then integrated into the diagnostic dataset.

[0013] According to another aspect of this disclosure, an overcurrent protection monitoring system for a frequency converter is provided, the system comprising: The acquisition module is used to continuously acquire the three-phase current time series data of each inverter in the workstation, sample and form a raw dataset, and standardize the raw dataset. The fault separation module is used to perform fault segmentation on the standardized original dataset, so that abnormal data and normal data are separated. The frequency domain transformation module is used to transform the original dataset representing the abnormal data from the time domain to the frequency domain based on the discrete Fourier transform, so as to obtain a new diagnostic dataset. The fault analysis module is used to extract and reduce the dimensionality of features in the diagnostic dataset based on principal component analysis technology to obtain a feature set, calculate the squared prediction error of each feature in the feature set, obtain the total contribution of the features based on the influence of the squared prediction error on other features, and determine whether the detected object is faulty based on the total contribution.

[0014] The technical solution disclosed herein has the following beneficial effects: By standardizing the current time series, the original characteristics of the current can be preserved, whether it is in a fault, transient or steady state. Based on discrete Fourier transform and principal component analysis, the standardized current data can be diagnosed, making faults easier to detect and allowing for further analysis of the root cause of the fault. Attached Figure Description

[0015] Figure 1 This is a flowchart of an overcurrent protection monitoring method for a frequency converter, as described in the embodiments of this specification. Figure 2 This is a block diagram of the overcurrent protection monitoring system for frequency converters in the embodiments of this specification. Detailed Implementation

[0016] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided to make this disclosure more comprehensive and complete, and to fully convey the concept of example embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are provided to give a full understanding of embodiments of this disclosure. However, those skilled in the art will recognize that the technical solutions of this disclosure may be practiced with one or more of the specific details omitted, or other methods, components, systems, steps, etc., may be employed. In other instances, well-known technical solutions are not shown or described in detail to avoid obscuring various aspects of this disclosure.

[0017] Furthermore, the accompanying drawings are merely illustrative of this disclosure. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor systems and / or microcontroller systems.

[0018] like Figure 1 As shown in the embodiments of this specification, an overcurrent protection monitoring method for frequency converters is provided. The execution subject of this method can be a computer. Specifically, the method may include the following steps S101~S104: In step S101, the current time series data of the three-phase current of each inverter in the workstation are continuously acquired, sampled to form an original dataset, and the original dataset is standardized.

[0019] To detect IGBT (Integrated Power Supply) faults in real time, real-time data acquisition of the switch can be performed, such as sampling the current through a sampling resistor. During sampling, to avoid serious false alarms and missed alarms, after acquiring the current time series data, it is sampled at fixed intervals with a fixed sampling length. In the Xth sampling, the interval between the starting position of the Xth sampling and the starting position of the (X-1)th sampling is less than the fixed sampling length.

[0020] Specifically, Fs is the number of samples, T is the period of the detected current time series data, H is the interval length, and N is the fixed sampling length, resulting in... The dataset is represented as p (p=a,b,c), where a, b, and c represent different phases. The k-stage dataset is represented as: ; in, , Indicates time t=* Then, the kth data Depend on Composition, represented as: ; During the testing process, continuous sampling is conducted to improve the timeliness of fault detection and prevent serious faults from occurring.

[0021] Furthermore, different fault types generate different data, typically represented as continuous and lengthy one-dimensional time series data. Significant differences between different fault samples can affect subsequent fault identification. To address this issue, data standardization is employed, such as using Z-score normalization to adjust the data, setting its mean to zero and its standard deviation to 1.

[0022] In step S102, the standardized original dataset is subjected to fault segmentation, which separates abnormal data from normal data.

[0023] In signal segmentation, a segmentation model function is established based on partial differential equations. The segmentation model function sets the original dataset to consist of normal data and abnormal data. The segmentation model function includes an evolution curve and an energy fitting function. According to the segmentation model function and the Dirac function, the original dataset is segmented when the difference between normal data and abnormal data tends to a normal value.

[0024] Fault segmentation can improve data quality and yield more accurate structures in subsequent fault analysis. During segmentation, the segmentation model function can be expressed as: ; Where C is the evolution curve, As the weighting coefficients, the energy fitting function can be expressed as: ; To get closer to the target signal, the formula is set as follows: ; According to the Dirac function: ; When the difference between normal and abnormal data approaches normal values, segmentation can begin. A smaller constant defined.

[0025] In step S103, based on the discrete Fourier transform, the original dataset representing the abnormal data is converted from the time domain to the frequency domain to obtain a new diagnostic dataset.

[0026] Although the three-phase currents have significant differences, it is difficult to identify when the fault occurs in the time domain. Therefore, the discrete Fourier transform is used to transform the data from the time domain to the frequency domain. Specifically, based on the number of samples, the sampling length and the defined frequency components when the original dataset is sampled, the amplitude and phase of each frequency component are calculated. The frequency components are rearranged so that the frequency gradually increases from zero to obtain frequency domain data. The frequency domain data is then integrated into the diagnostic dataset.

[0027] Specifically, calculate: ; It is the amplitude of the i-th fundamental frequency. Then, the frequency domain data is calculated, and the frequency components are rearranged: ; A new diagnostic dataset was obtained: .

[0028] In step S104, based on principal component analysis, features in the diagnostic dataset are extracted and dimensionality reduced to obtain a feature set. The squared prediction error of each feature in the feature set is calculated. Based on the influence of the squared prediction error on other features, the total contribution of the features is obtained. Based on the total contribution, it is determined whether the detected object is faulty.

[0029] After the dataset undergoes discrete Fourier feature extraction, it needs to be dimensionality reduced through principal component analysis.

[0030] The training data for the principal component model is constructed based on data sampled using the same method as in step S101: ; By retaining the previous principal component, most of the features in the dataset are preserved: ; Where T is the score matrix, P is the loading matrix, and E is the residual matrix; then, the squared prediction error (S) and Use statistics to check for errors in the process: ; in, Indicates the control limits for the squared prediction error; ; express The control limit, when When both the sum and squared prediction errors exceed the control limits, it can be used to diagnose whether there is a current fault. Then, based on the impact of the squared prediction error on other characteristics, the total contribution of the characteristics is obtained. Based on the total contribution, the specific fault location is located, such as through the formula: ; among, It is the contribution of the i-th variable in the dataset. It represents the i-th column of the identity matrix T.

[0031] Based on the above method, when an open circuit or short circuit (overcurrent) fault occurs in any phase of the frequency converter, the specific phase fault can be located through the above steps S101-S104, so that timely warning can be given, the fault can be eliminated, and the fault can be prevented from escalating and damaging the circuit.

[0032] In one implementation, standardization can be achieved using an alternative method, which includes setting an average error threshold. and standardization magnitude Calculate the k dataset The average value of the current time series data mentioned in the text The summary is as follows If the average value Greater than the average error threshold Calculate the maximum absolute value of the dataset. Based on the maximum absolute value of the dataset and the standardized amplitude, a first standardization coefficient is obtained. Based on the first standardization coefficient, the standardized current time series data is calculated. The first standardization coefficient is derived from the maximum current value. and target standardized amplitude The calculation is obtained, i.e., the calculation formula: ; This process standardizes each element in the dataset and uses a formula to form a new, standardized dataset.

[0033] If the average value Less than the average error threshold Calculate the current component within each sampling period, i.e., calculate using Clark transform. The transformation formula is as follows: ; Simultaneously, the sampling amplitude of the current component within each sampling period is calculated. Sampling amplitude The calculation is as follows: ; Based on the sampling amplitude and the standardized amplitude The second standardization coefficient is obtained, and the standardized current time series data is calculated based on the second standardization coefficient, wherein the calculation formula is as follows: The operation is repeated until each of the current time series data in the dataset is standardized.

[0034] In summary, a complete dataset contains data under different loads. However, to avoid subsequent feature extraction errors caused by load changes, this invention proposes a method for standardizing data based on the instantaneous amplitude of the current vector. The dataset processed by this method is unaffected by load changes and can detect faults more accurately. Specifically, when the system is in steady state, the three-phase current output by the inverter should be a sine wave with an average value of zero over one cycle. Due to factors such as AD sampling accuracy, offset correction, and noise interference, the average current value over one cycle is not strictly zero. In transient states, the output current does not abruptly change to the specified current amplitude but rather undergoes a transition process; typically, increasing or decreasing to the specified current amplitude requires several cycles, during which the average current value is not zero. Therefore, an average error threshold is set. When the average Less than the average error threshold If the system is in a normal state, then it is considered to be in a normal state. In practice, the maximum load of the motor is increased from no-load to full load. Average error threshold. The maximum value is taken as the maximum value in this transient process, and other transient processes will not exceed this maximum value. In addition, when an IGBT experiences an open-circuit fault, its three-phase current will change. Due to partial voltage loss, the current is not sinusoidal, and the average current deviates significantly from zero within one cycle. Based on this, the above-mentioned standardization method is adopted.

[0035] In one embodiment, after the dataset is standardized, it is preprocessed using a linear filter and wavelet transform method to remove noise.

[0036] Based on the same line of thought, such as Figure 2As shown in the exemplary embodiment of this disclosure, an overcurrent protection monitoring system for frequency converters is also provided. The system includes: a data acquisition module 201, used to continuously acquire the three-phase current time series data of each frequency converter in the workstation, sample to form an original dataset, and standardize the original dataset; a fault separation module 202, used to perform fault segmentation on the standardized original dataset, so that abnormal data and normal data are separated; a frequency domain conversion module 203, used to convert the original dataset representing the abnormal data from the time domain to the frequency domain based on discrete Fourier transform, to obtain a new diagnostic dataset; and a fault analysis module 204, used to extract and reduce the dimensionality of the features in the diagnostic dataset based on principal component analysis technology, to obtain a feature set, calculate the squared prediction error of each feature in the feature set, obtain the total contribution of the features based on the influence of the squared prediction error on other features, and determine whether the detected object is faulty based on the total contribution.

[0037] In the above system, the standardization process of the current time series ensures that the original characteristics of the current are preserved, whether it is in a fault, transient or steady state. Based on discrete Fourier transform and principal component analysis, the standardized current data is used for diagnosis, making faults easier to detect and allowing for further analysis of the root cause of the fault.

[0038] The specific details of each module in the above system have been described in detail in the implementation method section. For any undisclosed details, please refer to the implementation method section, and therefore will not be repeated here.

[0039] It should be noted that although several modules or units for the device used to perform actions have been mentioned in the detailed description above, this division is not mandatory. In fact, according to exemplary embodiments of this disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0040] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and embodiments are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the claims.

Claims

1. A method for monitoring overcurrent protection in a frequency converter, characterized in that, The method includes: The current time series data of the three-phase current of each inverter in the workstation is continuously acquired, sampled to form a raw dataset, and the raw dataset is standardized. Fault segmentation is performed on the standardized original dataset to separate abnormal data from normal data. Based on the discrete Fourier transform, the original dataset representing the abnormal data is transformed from the time domain to the frequency domain to obtain a new diagnostic dataset; Based on principal component analysis, features in the diagnostic dataset are extracted and dimensionality reduced to obtain a feature set. The squared prediction error of each feature in the feature set is calculated. The total contribution of the features is obtained based on the influence of the squared prediction error on other features. Based on the total contribution, it is determined whether the detected object is faulty.

2. The overcurrent protection monitoring method for frequency converters according to claim 1, characterized in that, After acquiring the current time series data, the current time series data is sampled at intervals with a fixed sampling length. In the Xth sampling, the interval between the starting position of the Xth sampling and the starting position of the (X-1)th sampling is less than the fixed sampling length.

3. The overcurrent protection monitoring method for frequency converters according to claim 1, characterized in that, Standardization includes: Set an average error threshold and a standardization amplitude, calculate the average value of any set of current time series data in the dataset, if the average value is greater than the average error threshold, calculate the maximum value of the absolute value of the dataset, obtain a first standardization coefficient based on the maximum value of the absolute value of the dataset and the standardization amplitude, and calculate the standardized current time series data based on the first standardization coefficient. If the average value is less than the average error threshold, calculate the current component in each sampling period and the sampling amplitude of the current component in each sampling period. Based on the sampling amplitude and the standardized amplitude, obtain the second standardization coefficient. Based on the second standardization coefficient, calculate the standardized current time series data. Repeat the operation until each of the current time series data in the dataset is standardized.

4. The overcurrent protection monitoring method for frequency converters according to claim 1, characterized in that, After standardizing the dataset, the standardized dataset is preprocessed using linear filters and wavelet transform methods to remove noise.

5. The overcurrent protection monitoring method for frequency converters according to claim 1, characterized in that, When performing signal segmentation, a segmentation model function is established based on partial differential equations. The segmentation model function sets the original dataset to consist of normal data and abnormal data. The segmentation model function includes an evolution curve and an energy fitting function. According to the segmentation model function and the Dirac function, the original dataset is segmented when the difference between normal data and abnormal data tends to a normal value.

6. The overcurrent protection monitoring method for frequency converters according to claim 1, characterized in that, When performing the Discrete Fourier Transform, the following are included: Based on the number of samplings, sampling length, and defined frequency components during the sampling of the original dataset, the amplitude and phase of each frequency component are calculated. The frequency components are rearranged so that the frequency gradually increases from zero to obtain frequency domain data. The frequency domain data is then integrated into the diagnostic dataset.

7. An overcurrent protection monitoring system for frequency converters, characterized in that, The system includes: The acquisition module is used to continuously acquire the three-phase current time series data of each inverter in the workstation, sample and form a raw dataset, and standardize the raw dataset. The fault separation module is used to perform fault segmentation on the standardized original dataset, so that abnormal data and normal data are separated. The frequency domain transformation module is used to transform the original dataset representing the abnormal data from the time domain to the frequency domain based on the discrete Fourier transform, so as to obtain a new diagnostic dataset. The fault analysis module is used to extract and reduce the dimensionality of features in the diagnostic dataset based on principal component analysis technology to obtain a feature set, calculate the squared prediction error of each feature in the feature set, obtain the total contribution of the features based on the influence of the squared prediction error on other features, and determine whether the detected object is faulty based on the total contribution.

Images