An inverter fault-tolerant control method, device and medium

Abstract

The application discloses an inverter fault fault-tolerant control method, equipment and medium, and is applied to the technical field of inverter fault monitoring. The method comprises the following steps: collecting voltage and current signals corresponding to an inverter, performing Hilbert-Huang transformation on the voltage and current signals to obtain a center frequency and an amplitude corresponding to the inverter; if the inverter fault type corresponding to the center frequency and the amplitude is determined according to a fault database, then determining the inverter fault degree according to the energy, the duration and the amplitude corresponding to the inverter fault; and determining the processing mode corresponding to the inverter according to the inverter fault degree and the inverter fault type. As can be seen, the application performs double determination of the fault type on the center frequency and the amplitude determined through Hilbert-Huang transformation, thereby improving the fault identification accuracy. In addition, different fault types and fault degrees correspond to different processing operations in the application, thereby improving the sustainability and working efficiency of the inverter.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of inverter fault monitoring, in particular to an inverter fault tolerance control method, device and medium. BACKGROUND

[0002] As the core equipment of new energy power generation system, the operation stability of inverter directly affects the safety of the entire power grid. In actual application, inverters often face complex power grid conditions and environmental disturbances, and thus are prone to faults. In order to ensure the safe and stable operation of the inverter, the design of the fault monitoring system of the inverter is crucial.

[0003] At present, the fault monitoring system of the inverter mostly adopts a centralized data processing architecture. After the voltage, current and other parameters are collected by sensors, they are transmitted to a central control system, and a threshold detection method is used to judge faults. However, in this type of method, the delay in the data transmission process is likely to cause a lag in the fault response, and the threshold detection method has weak ability to distinguish complex faults, which is prone to misjudgment. In addition, the updating of the fault library in the fault monitoring system relies on manual input, and cannot adapt to the variation of fault characteristics of the inverter under different working conditions. When a fault occurs, the inverter is often shut down for troubleshooting and maintenance for safety considerations. However, the non-detailed classification and response of faults and the subsequent indiscriminate shutdown processing method lead to the fact that once a module in the entire machine fails, the entire machine cannot perform normal control actions, which greatly affects the sustainability of the inverter and greatly reduces the work efficiency.

[0004] In view of the above-mentioned technology, it is an urgent problem for those skilled in the art to seek an inverter fault tolerance control method with high precision, low delay and high efficiency. SUMMARY

[0005] The purpose of the present application is to provide an inverter fault tolerance control method, device and medium. The problem of lag in fault response and misjudgment caused by the threshold detection method in the prior art can be solved, and the problem of reduced sustainability and work efficiency of the inverter caused by shutdown for maintenance when a module in the entire machine fails can also be solved.

[0006] To solve the above-mentioned technical problems, the present application provides an inverter fault tolerance control method, comprising:

[0007] Collecting voltage and current signals corresponding to the inverter, and performing Hilbert-Huang transform on the voltage and current signals to obtain the center frequency and amplitude corresponding to the inverter;

[0008] If the inverter fault type corresponding to the center frequency and the amplitude is determined according to the fault database, the inverter fault degree is determined according to the energy, the duration and the amplitude corresponding to the inverter fault; wherein the fault database is constructed according to each inverter fault type and corresponding fault parameters, and is updated according to federated learning;

[0009] The processing mode corresponding to the inverter is determined according to the inverter fault degree and the inverter fault type.

[0010] Preferably, the voltage and current signals are subjected to Hilbert Huang transform to obtain the center frequency and the amplitude corresponding to the inverter, comprising:

[0011] Each signal component corresponding to the voltage and current signals is determined based on an empirical mode decomposition algorithm;

[0012] The Hilbert transform is performed on each signal component to obtain the instantaneous frequency and the instantaneous amplitude corresponding to each signal component;

[0013] The signal component corresponding to the instantaneous frequency in the abnormal frequency range is obtained from each signal component, and is taken as a target signal component;

[0014] The center frequency and the amplitude corresponding to the inverter are determined based on the instantaneous frequency and the instantaneous amplitude corresponding to the target signal component.

[0015] Preferably, it further comprises:

[0016] If the inverter fault type corresponding to the center frequency and the amplitude is not determined according to the fault database, the operating parameters corresponding to the inverter are obtained;

[0017] It is determined whether the inverter corresponding to the operating parameters is a fault inverter according to the operating parameters;

[0018] If the inverter is a fault inverter, the new fault feature structured parameters corresponding to the operating parameters are determined according to the federated fault learning model;

[0019] The new inverter fault type and the new fault parameters corresponding to the new fault feature structured parameters are determined, and the new inverter fault type and the new fault parameters are updated into the fault database.

[0020] Preferably, after the new inverter fault type and the new fault parameters are updated into the fault database, it further comprises:

[0021] It is determined whether the new inverter fault type and the new fault parameters meet the similarity requirements with each historical inverter fault type and each historical fault parameter stored in the fault database;

[0022] If the similarity requirements are met, the historical inverter fault types and historical fault parameters that meet the similarity requirements are obtained and used as the target historical inverter fault types and target historical fault parameters.

[0023] The new inverter fault type, new fault parameters, target historical inverter fault type, and target historical fault parameters are fitted to generate inverter fault fitting type and fitted fault parameters, and then updated to the fault database.

[0024] Preferably, the degree of inverter failure is determined based on the energy, duration, and amplitude corresponding to the inverter failure, including:

[0025] Obtain the total energy, short-time energy, and abnormal mode energy corresponding to the duration of an inverter fault;

[0026] The corresponding abnormal mode energy ratio is determined based on the abnormal mode energy and the total energy;

[0027] The corresponding time-frequency entropy is determined based on the short-time energy and the total energy.

[0028] The time ratio is determined based on the duration of the inverter failure and the safety tolerance time corresponding to the inverter failure type.

[0029] The amplitude ratio is determined based on the amplitude corresponding to the inverter fault and the amplitude threshold corresponding to the inverter fault type.

[0030] The degree of inverter failure is determined based on the abnormal mode energy ratio, time-frequency entropy, time ratio, and amplitude ratio.

[0031] Preferably, obtaining the duration of the inverter fault includes:

[0032] Determine the first time when the instantaneous frequency of each signal component first appears within the abnormal frequency range;

[0033] Determine the corresponding target first time based on each first time;

[0034] Determine the second time corresponding to the current moment;

[0035] The duration is determined based on the first and second time points of the target.

[0036] Preferably, the inverter fault degree is determined based on the abnormal mode energy ratio, time-frequency entropy, time ratio, and amplitude ratio, including:

[0037] Obtain the energy weight parameters, time weight parameters, and amplitude weight parameters corresponding to the abnormal mode energy ratio, time ratio, and amplitude ratio, respectively;

[0038] The time-frequency entropy is normalized to obtain the energy distribution entropy;

[0039] acquire a distribution weight parameter corresponding to the energy distribution entropy;

[0040] determine a fault severity index according to the abnormal modal energy ratio, the time ratio, the amplitude ratio, the energy distribution entropy, the energy weight parameter, the time weight parameter, the amplitude weight parameter, and the distribution weight parameter;

[0041] determine the inverter fault degree corresponding to the fault severity index.

[0042] Preferably, determining the inverter fault degree corresponding to the fault severity index comprises:

[0043] when the fault severity index is less than a first fault threshold, the inverter fault degree is a light fault;

[0044] when the fault severity index is not less than the first fault threshold and less than a second fault threshold, the inverter fault degree is a moderate fault;

[0045] when the fault severity index is not less than the second fault threshold, the inverter fault degree is a severe fault.

[0046] In another aspect, the present application also provides an electronic device, comprising a memory for storing a computer program;

[0047] a processor for executing the computer program to implement the steps of the inverter fault tolerance control method described above.

[0048] In another aspect, the present application also provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the steps of the inverter fault tolerance control method described above.

[0049] The inverter fault tolerance control method provided by the present application comprises: collecting voltage and current signals corresponding to an inverter, performing Hilbert-Huang transform on the voltage and current signals to obtain a center frequency and an amplitude corresponding to the inverter; if the inverter fault type corresponding to the center frequency and the amplitude is determined according to a fault database, then determining the inverter fault degree according to the energy, the duration, and the amplitude corresponding to the inverter fault; wherein the fault database is constructed according to each inverter fault type and corresponding fault parameters, and is updated according to federated learning; and determining the processing mode corresponding to the inverter according to the inverter fault degree and the inverter fault type. As can be seen, the present application performs double determination of the fault type of the center frequency and the amplitude determined by Hilbert-Huang transform, thereby improving the fault recognition accuracy. In addition, different fault types and fault degrees in the present application correspond to different processing operations, thereby improving the sustainability and efficiency of inverter operation. Moreover, the fault database in the present application can be updated according to federated learning, thereby shortening the new fault recognition period in subsequent inverter fault judgment. BRIEF DESCRIPTION OF DRAWINGS

[0050] In order to more clearly illustrate the embodiments of the present application, the drawings required to be used in the embodiments will be briefly introduced. Obviously, the drawings described below are only some of the embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0051] Figure 1 A flow chart of an inverter fault tolerance control method provided by an embodiment of the present application;

[0052] Figure 2 A structural diagram of an electronic device provided by another embodiment of the present application. DETAILED DESCRIPTION

[0053] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the protection scope of the present application.

[0054] The core of the present application is to provide an inverter fault tolerance control method, device and medium.

[0055] In order to enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the drawings and specific embodiments.

[0056] Figure 1 A flow chart of an inverter fault tolerance control method provided by an embodiment of the present application, as shown in Figure 1 includes the following steps:

[0057] S10: Collecting voltage and current signals corresponding to the inverter, performing Hilbert Huang transform on the voltage and current signals to obtain the center frequency and amplitude corresponding to the inverter.

[0058] S11: If the inverter fault type corresponding to the center frequency and amplitude is determined according to the fault database, the inverter fault degree is determined according to the energy, duration and amplitude corresponding to the inverter fault.

[0059] S12: Determining the processing mode corresponding to the inverter according to the inverter fault degree and the inverter fault type.

[0060] In specific embodiments, a fault database is first constructed, and the construction steps of the fault database are as follows:

[0061] 1、In the inverter system experiment, a large number of simulation is generated for multiple typical faults, such as: A type fault: IGBT module open circuit fault (simulated by disconnecting the IGBT drive signal); B type fault: output end electrolytic capacitor aging fault (simulated by using a capacity decay of 30% capacitor); C type fault: input end film capacitor breakdown fault (simulated by parallel 10kΩ resistance local breakdown); D type fault: three-phase output short circuit fault (simulated by star connection 0.1Ω resistance).

[0062] 2、In the output end of the inverter, a voltage sensor (precision ±0.5%) and a current sensor (bandwidth DC-10kHz) with a sampling frequency of 20kHz are set to collect the voltage signals and current signals under different typical fault conditions.

[0063] 3、Perform Hilbert Huang transform on the collected signals: through empirical mode decomposition algorithm (Empirical Mode Decomposition, EMD), the voltage signal and the current signal are decomposed into intrinsic mode functions (IMF), since the processing method and corresponding calculation formula of the current signal and the voltage signal are the same, the voltage signal is taken as an example, the corresponding formula is:

[0064] ;

[0065] Where, is the residual component, representing the trend item of the signal; is the intrinsic mode function; is any one of the intrinsic mode functions.

[0066] 4、Perform Hilbert transform on each IMF component to get the instantaneous frequency , the formula is as follows:

[0067] ;

[0068] ;

[0069] Where, is the Hilbert transform operator.

[0070] 5、Extract the feature frequency of each fault (take the above four types of faults as an example): A type fault feature frequency : 150Hz±10Hz (fundamental frequency) + 500Hz harmonic; B type fault feature frequency : 80 Hz ± 5 Hz (linearly reduced with aging degree); Class C fault characteristic frequency : 220 Hz ± 15 Hz; Class D fault characteristic frequency : 300 Hz or more wide frequency oscillation (amplitude > 2 V).

[0071] 6. A fault database (three-dimensional database) is constructed based on the parameters obtained in the above steps, and the storage structure corresponding to each type of fault in the fault database is shown in Table 1. The fault database supports dynamic expansion of fields, and reserves feature storage bits for new fault types.

[0072] Table 1

[0073]

[0074] In the fault database, the stored faults are not fixed, and they will be updated with new fault types in the fault database according to federated learning of faults, that is, new fault types can be prompted to operation and maintenance personnel through a human-computer interaction interface, and then recorded and updated in the fault database.

[0075] After the above fault database is constructed, it can be determined whether a fault occurs and the type of the fault in real time according to the operation of the inverter.

[0076] An edge computing unit (for example, STM32H743 microprocessor) is deployed at the output end of the inverter to collect voltage and current signals in real time, with a sampling frequency of 20 kHz and a sampling accuracy of 16 bits. Then, the voltage and current are subjected to Hilbert-Huang transform to obtain the corresponding center frequency and amplitude of the inverter.

[0077] If the current center frequency and amplitude can be successfully matched in the fault database, it is clear that the inverter has failed, and the type of the fault can be determined. At this time, the degree of inverter failure is further determined according to the three-dimensional parameters of energy, duration and amplitude corresponding to the inverter failure, and then corresponding operations are performed according to different fault types and different fault degrees.

[0078] For example, taking a class B fault and a mild fault as an example, a derating control is triggered, the output power is reduced to 80% of the rated value through a proportion-integration-differentiation (PID) controller, the pulse width modulation (PWM) duty cycle range is limited from 0-100% to 0-80%, and the switching frequency is kept stable (20 kHz).

[0079] It should be noted that the specific implementation of obtaining the duration of the inverter fault is to determine each first time when the instantaneous frequency corresponding to each signal component first appears in the abnormal frequency range; determine the corresponding target first time according to each first time; determine the second time corresponding to the current time; and determine the duration according to the target first time and the second time. That is, record the first time when each instantaneous frequency first exceeds the normal range, and then determine the corresponding target first time using the average value, and finally determine the duration according to the current second time.

[0080] Therefore, the present application determines the center frequency and amplitude determined by the Hilbert Huang transform for double fault type judgment, thereby improving the fault recognition accuracy. In addition, different fault types and fault degrees in the present application correspond to different processing operations, which improves the sustainability and efficiency of the inverter operation. In addition, the fault database in the present application can be updated according to federated learning, thereby shortening the new fault recognition period in subsequent inverter fault judgment.

[0081] On the basis of the above-mentioned embodiments, as a preferred embodiment, the specific implementation of performing Hilbert Huang transform on the voltage and current signals to obtain the center frequency and amplitude corresponding to the inverter is:

[0082] Based on the empirical mode decomposition algorithm, determine each signal component corresponding to the voltage and current signals;

[0083] Perform Hilbert transform on each signal component to obtain the instantaneous frequency and instantaneous amplitude corresponding to each signal component;

[0084] Obtain the signal component corresponding to the instantaneous frequency in the abnormal frequency range in each signal component, and use it as the target signal component;

[0085] Determine the center frequency and amplitude corresponding to the inverter based on the instantaneous frequency and instantaneous amplitude corresponding to the target signal component.

[0086] In a specific embodiment, an edge computing unit is deployed at the output end of the inverter, and real-time voltage and current signals are collected. Then, the original voltage and current signals need to be preprocessed, that is, the original voltage and current signals are subjected to 50Hz notch filtering to remove power grid fundamental wave interference; then, the empirical mode decomposition algorithm (EMD algorithm) is used to decompose the voltage and current signals into signal components IMF components (single frequency components) denoted as 、 、 …, and the termination condition of decomposition is that the energy difference of two consecutive IMF components is less than 1%. Then, time-frequency feature extraction is realized, that is, Hilbert transform is performed on each IMF component to calculate the instantaneous frequency and the instantaneous amplitude . The instantaneous amplitude It equals the amplitude.

[0087] Determining the instantaneous frequency and instantaneous amplitude Next, fault identification and judgment are required. This involves acquiring the signal components whose instantaneous frequencies fall within the abnormal frequency range and using them as target signal components. This can be understood as: normal signals are eliminated. For example, if the normal operating frequency range is set to 49.5-50.5Hz, stable frequency components within this range (duration > 100ms) are eliminated. Then, based on the remaining frequency components (target signal components), their center frequency is calculated. (This can also be understood as a weighted average frequency), the formula is as follows:

[0088] ;

[0089] Finally, the center frequency Compare with the feature frequency range in the database; if a match is found and the instantaneous amplitude... If the value is greater than the corresponding threshold, it is determined to be the corresponding fault type.

[0090] This invention provides a specific implementation method for performing Hilbert-Huang transform on voltage and current signals to obtain the center frequency and amplitude of the inverter. This method includes signal filtering and extraction processes, which further ensures the accuracy of the center frequency and amplitude of the inverter.

[0091] Based on the above embodiments, as a preferred embodiment, it further includes:

[0092] If the inverter fault type corresponding to the center frequency and amplitude cannot be determined from the fault database, then obtain the corresponding operating parameters of the inverter.

[0093] Determine whether the corresponding inverter is a faulty inverter based on the operating parameters;

[0094] If the inverter is a faulty inverter, then the new fault feature structured parameters corresponding to the operating parameters are determined according to the federated fault learning model.

[0095] Based on the structured parameters of the new fault characteristics, the corresponding new inverter fault type and new fault parameters are determined, and the new inverter fault type and new fault parameters are updated to the fault database.

[0096] In an embodiment, if the center frequency and the amplitude corresponding to the inverter fault type are not determined according to the fault database, the current inverter corresponds to two cases. The first case is that the inverter itself is not faulty; the second case is that the fault corresponding to the current inverter does not exist in the fault database. Therefore, it is necessary to determine whether the inverter is a faulty inverter according to the operating parameters of the inverter. If not, it belongs to the first case, and the inverter can operate normally; if yes, it belongs to the second case.

[0097] When it belongs to the second case, it is necessary to analyze and summarize the current situation, so as to add the corresponding fault to the fault database, so as to timely judge and handle when such a fault occurs subsequently.

[0098] That is, the new fault feature structured parameter corresponding to the operating parameter is determined according to the federal fault learning model, and the new inverter fault type and the new fault parameter corresponding to the new fault feature structured parameter are determined, and the new inverter fault type and the new fault parameter are updated to the fault database. The specific steps are as follows:

[0099] 1. New fault feature detection and labeling: in real-time analysis, if the instantaneous frequency satisfies: not in the existing fault database feature frequency range; the instantaneous amplitude is continuously higher than the noise threshold (0.1V), it is labeled as a potential new fault feature. The associated parameters (which can also be understood as the operating parameters of the current inverter) of the feature are automatically recorded, including the load rate (accuracy ±2%) when the fault occurs, the environmental temperature (sampling interval 1s), the inverter running time, to form a new fault feature package : .

[0100] 2. Local model training and parameter extraction: the edge node calls the local lightweight industrial neural network (using a 3-layer CNN structure, the input is the time-frequency spectrum, and the output is the fault probability distribution) to train the new fault feature package as a training sample, fine-tune training on the local data set (containing historical fault data), the iteration number is set to 50 times, and the learning rate is 0.001. After training, the convolution layer weight parameter W and the bias term b of the model are extracted as the local model parameters to be uploaded (the parameter size is compressed to within 500KB to avoid transmission redundancy).

[0101] 3. Federal server parameter aggregation: the federal server receives the model parameters uploaded by each edge node (each inverter), and aggregates them using a weighted average algorithm: ; wherein, is the training sample size of the i-th node, For the total sample size, For each edge node uploaded model parameters, ensure that the sample size is higher. After aggregation, the accuracy of the global model is verified (10% of the reserved fault samples are used), and if the accuracy is ≥95%, the next step is entered, otherwise the nodes are returned to retraining (the number of iterations is increased to 80 times).

[0102] 4, Global database update and model distribution: the server writes the aggregated global model parameters Convert to new fault feature structure parameters (including center frequency, amplitude range, and associated working condition parameters), write to the fault database, and assign a unique fault ID (such as E-class fault).

[0103] In addition, in a specific application scenario, the operation of a scenario includes multiple nodes (inverters), and when a new fault occurs in an inverter, a corresponding signal needs to be sent to each inverter, so that the remaining inverters can be aware of the fault in time, and the specific way is to update the new fault to the fault database and then reissue it to the corresponding inverter (node), or use an incremental update method to issue the new fault feature and the optimized model to each edge node. When the incremental update method is used to issue the new fault feature and the optimized model to each edge node, only the newly added field (such as the 180Hz±15Hz frequency range of the E-class fault) is transmitted, reducing the transmission amount.

[0104] At the same time, it should also be noted that there are certain differences in the operation of each inverter, so when a new fault is received and a new fault is encountered, it needs to be compared with historical faults to determine whether it is the same fault or the same type of fault, that is, whether the new inverter fault type and the new fault parameter meet the similarity requirements with the historical inverter fault type and the historical fault parameter stored in the fault database; if the similarity requirements are met, the historical inverter fault type and the historical fault parameter that meet the similarity requirements are obtained and used as the target historical inverter fault type and the target historical fault parameter; the new inverter fault type, the new fault parameter, the target historical inverter fault type, and the target historical fault parameter are fitted to generate an inverter fault fitting type and a fitting fault parameter, and are updated to the fault database.

[0105] In a specific embodiment, after the edge node (the remaining inverters) receives the update, it automatically compares the new fault feature with the local historical data, and if there is a similar feature (frequency deviation <5Hz), it generates a personalized correction coefficient (such as a capacitor fault frequency offset of +2Hz in a high-altitude area). Local verification: simulate the generation of 3 groups of new fault features, verify the recognition accuracy ≥98%, and complete the update, otherwise trigger local secondary training (combine the global model and local data).

[0106] Among them, it needs to be pointed out that the operation and maintenance personnel can be prompted to confirm the new fault type through the man-machine interaction interface, and then the fault database is entered and updated.

[0107] The application provides a method for updating a fault database driven by federated learning, which realizes safe sharing and global database iteration of multi-node fault data through federated learning, improves the identification ability of the system for new fault types on the premise of protecting the data privacy of each inverter node.

[0108] On the basis of the above-mentioned embodiments, as a preferred embodiment, the specific implementation manner of determining the inverter fault degree according to the corresponding energy, duration and amplitude when the inverter fails is:

[0109] Obtaining the total energy, short-time energy and abnormal modal energy corresponding to the duration when the inverter fails;

[0110] Determining the abnormal modal energy ratio according to the abnormal modal energy and the total energy;

[0111] Determining the time-frequency entropy according to the short-time energy and the total energy;

[0112] Determining the time ratio according to the duration when the inverter fails and the safe tolerance time corresponding to the inverter fault type;

[0113] Determining the amplitude ratio according to the amplitude when the inverter fails and the amplitude threshold value corresponding to the inverter fault type;

[0114] Determining the inverter fault degree according to the abnormal modal energy ratio, the time-frequency entropy, the time ratio and the amplitude ratio.

[0115] The specific implementation manner of determining the inverter fault degree according to the abnormal modal energy ratio, the time-frequency entropy, the time ratio and the amplitude ratio is:

[0116] Obtaining the energy weight parameter, the time weight parameter and the amplitude weight parameter corresponding to the abnormal modal energy ratio, the time ratio and the amplitude ratio respectively;

[0117] Normalizing the time-frequency entropy to obtain the energy distribution entropy;

[0118] Obtaining the distribution weight parameter corresponding to the energy distribution entropy;

[0119] Determining the fault severity index according to the abnormal modal energy ratio, the time ratio, the amplitude ratio, the energy distribution entropy, the energy weight parameter, the time weight parameter, the amplitude weight parameter and the distribution weight parameter;

[0120] When the fault severity index is less than the first fault threshold value, the inverter fault degree is a mild fault;

[0121] When the fault severity index is not less than the first fault threshold and less than the second fault threshold, the inverter fault degree is moderate fault;

[0122] When the fault severity index is not less than the second fault threshold, the inverter fault degree is severe fault.

[0123] In the embodiment, in order to improve the robustness and interpretability of the central frequency and amplitude determination extracted based on the Hilbert Huang Transform (HHT), the application proposes a comprehensive severity index-fault severity index which takes the abnormal modal energy ratio, time-frequency entropy, duration and amplitude into account. The index comprehensively considers the amplitude, duration ratio, energy ratio and spectral entropy to reflect the harmfulness of the fault and the energy concentration and chaos degree in the time-frequency domain, so as to improve the misjudgment and omission problem under the condition of grid interference and noise.

[0124] It decomposes the voltage signal and the current signal into intrinsic mode functions (IMF) by empirical mode decomposition algorithm, and each mode can obtain instantaneous frequency and instantaneous amplitude (or can be understood as amplitude) , , same below.

[0125] In a given fault window , the fault window can be understood as the target first time in the above, can be understood as the second time, so the duration .

[0126] The amplitude ratio is: The ratio of the current amplitude (also can be expressed as function ) and the amplitude threshold value (corresponding to the fault type of the inverter, such as the amplitude threshold value of IGBT fault) of the inverter fault type (the maximum threshold value of the fault type).

[0127] The time ratio is: The ratio of the duration and the safe tolerance time corresponding to the fault type of the inverter (such as the safe tolerance time of short circuit fault).

[0128] The abnormal modal energy ratio is: The ratio of the abnormal modal energy and the total energy , wherein the abnormal modal energy , The frequency range of abnormal signals (abnormal frequency range); total energy ;That For short-time energy, the calculation formula is as follows: ,That for The voltage signal corresponding to any one of the intrinsic mode functions.

[0129] Time frequency: For short-term energy and total energy The ratio of the values ​​is then normalized to obtain the normalized energy distribution entropy. To ensure scale consistency, the energy distribution entropy is defined as... ,That . ( There are equal probability distributions, and the maximum entropy is . Therefore, divide by (Obtain the normalized result)

[0130] In this embodiment, the severity of the fault is quantified and assessed by constructing a fault severity index, the formula of which is as follows:

[0131] ;

[0132] ;

[0133] in, For magnitude weighting parameters; For time weighting parameters; For energy weighting parameters; These are the distribution weight parameters. Preferably, their... , , , .

[0134] Classification based on S-value: (The first fault threshold) indicates a minor fault. This is a moderate fault. The second fault threshold is considered a severe fault. Updates to the thresholds (first and second fault thresholds) trigger a semi-automatic manual review strategy to prevent noise from being mistakenly labeled into the fault database.

[0135] The corresponding handling method for the inverter is determined based on the degree and type of inverter failure, as follows:

[0136] Take the B-class fault, mild fault as an example: trigger derating control, reduce the output power to 80% of the rated value through the PID controller, limit the pulse width modulation (PWM) duty cycle range from 0-100% to 0-80%, while keeping the switching frequency stable (20 kHz). Real-time monitoring: collect every 50 ms, if the S value drops for 3 consecutive cycles, slowly increase the power (5% each time); if the S value rises to 0.3 or above, upgrade to moderate fault handling.

[0137] Take the A and C class fault, moderate fault as an example: standby module switching, disconnect the fault IGBT module through the drive circuit, and close the relay of the standby module (switching time < 50 ms), use soft start strategy during switching process (current overshoot ≤10%). Adjust the operating parameters, recalculate the three-phase current balance, compensate for phase deviation through space vector pulse width modulation (SVPWM), and ensure that the total harmonic distortion (THD) is ≤5% after switching.

[0138] Take the D-class fault, severe fault as an example: emergency protection, immediately trigger the soft shutdown program, linearly reduce the PWM duty cycle from the current value to 0 within 10 ms, and disconnect the grid contactor (break time < 20 ms) to avoid grid impact. Send fault code (such as "D-001" for three-phase short circuit) to the central monitoring system through the CAN bus, and store the time-frequency spectrum (sampling rate 20 kHz) of the 10 seconds before the fault for subsequent maintenance.

[0139] In addition, in the mild fault recovery, when the fault disappears and there is no abnormality for 10 consecutive cycles, automatically remove the derating mode and restore the rated operation. After the module is repaired, maintenance personnel replace the faulty module, send the reset command through the infrared remote control, the system re-calibrates the parameter consistency of the standby module and the main module (error ≤1%), and restores the normal operation mode.

[0140] And need to save the corresponding time-frequency spectrum (recorded for 10 seconds) and the corresponding inverter operating parameters (load rate, temperature, etc.) at all times.

[0141] In summary, the hardware configuration of the inverter fault tolerance control method provided by the application is: signal acquisition module: Hall voltage sensor (precision 0.1%), current sensor (bandwidth 10 kHz), deployed at the output end of the inverter, mainly used for acquiring voltage and current signals; edge computing unit: FPGA (Field-Programmable Gate Array) chip (supports parallel EMD decomposition, processing delay < 1 ms). Communication module: supports 5G edge computing industrial gateway, realizes federal learning parameter transmission.

[0142] And the software process is as follows:

[0143] Initialization phase: 1000 sets of typical fault data are generated by the simulation platform to build a fault database and write it to the FPGA storage unit. After the system is powered on, the edge computing unit loads the initial fault database (stored in the SD card); performs EMD algorithm self-checking to ensure that the signal decomposition function is normal; the federal learning client establishes a connection with the server and completes identity authentication.

[0144] Running phase:

[0145] Real-time acquisition of voltage and current signals, EMD decomposition is performed every 10ms;

[0146] Perform Hilbert transform on the decomposed IMF components to extract instantaneous frequency and instantaneous amplitude (amplitude);

[0147] According to the above embodiment, if it is determined that it is a known fault, the corresponding level of measures is triggered according to the severity index S; if it is determined that it is a potential new fault, the federal learning update process is started.

[0148] In summary, the inverter fault tolerance control method provided by the application has the following advantages:

[0149] 1. The edge computing unit is deployed locally in the inverter, the non-stationary voltage / current signal is decomposed into intrinsic mode function (IMF) by EMD decomposition, and the instantaneous frequency and amplitude are extracted by Hilbert transform, realizing local high-precision analysis of fault features.

[0150] 2. The HHT algorithm is localized by edge computing, the processing delay is compressed to within 10ms, and the influence of power grid interference on feature extraction is reduced through 50Hz notch filtering and EMD termination condition optimization (energy difference <1%).

[0151] 3. A fault database of "frequency range-amplitude threshold-working condition parameter" is constructed, and federal learning is used to realize safe sharing of multi-node fault features, that is, each edge node only uploads model parameters (not original data), the federal server aggregates the parameters to update the global fault database, and then distributes it to each node. Through the characteristics of federal learning "data not moving model moving", the new fault identification period is shortened, and the privacy of inverter operation data is protected.

[0152] 4. A fault severity index is proposed, which comprehensively considers fault amplitude and duration, and divides faults into three levels of mild, moderate and severe, and triggers differentiated responses such as reduced capacity operation, standby module switching, and emergency shutdown.

[0153] 5. After federated learning updates, edge nodes automatically compare with local historical data to generate environmental adaptability correction coefficients (such as a +2Hz frequency shift for capacitor faults in high-altitude areas and a 10% reduction in the IGBT fault amplitude threshold in high-temperature environments), thereby achieving regional / environmental adaptation for fault identification and improving the system's adaptability to regional environments.

[0154] 6. Integrate fault detection (HHT), model iteration (federated learning), hierarchical response and automatic recovery mechanism to form a closed loop: minor faults automatically restore rated operation, and after module replacement, parameter calibration is achieved through infrared reset (error ≤1%), shortening fault recovery time.

[0155] Figure 2 A structural diagram of an electronic device provided in another embodiment of this application, such as... Figure 2 As shown, the electronic device includes: a memory 20 for storing computer programs;

[0156] The processor 21 is used to execute computer programs to implement the steps of the inverter fault-tolerant control method mentioned in the above embodiments.

[0157] The electronic devices provided in this embodiment may include, but are not limited to, smartphones, tablets, laptops, or desktop computers.

[0158] The processor 21 may include one or more processing cores, such as a quad-core processor or an octa-core processor. The processor 21 may be implemented using at least one of the following hardware forms: Digital Signal Processor (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). The processor 21 may also include a main processor and a coprocessor. The main processor, also known as the Central Processing Unit (CPU), is used to process data in the wake-up state; the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, the processor 21 may integrate a Graphics Processing Unit (GPU), which is responsible for rendering and drawing the content to be displayed on the screen. In some embodiments, the processor 21 may also include an Artificial Intelligence (AI) processor, which is used to handle computational operations related to machine learning.

[0159] The memory 20 can include one or more computer-readable storage media that can be non-transitory. The memory 20 can also include high-speed random access memory and nonvolatile, computer-readable storage media such as one or more magnetic disk storage devices, flash memory devices. In this embodiment, the memory 20 is at least used to store the following computer program 201, wherein the computer program is loaded and executed by the processor 21, and can realize the related steps of the inverter fault tolerance control method disclosed in any of the preceding embodiments. In addition, the resources stored by the memory 20 can also include an operating system 202 and data 203, etc., and the storage mode can be temporary storage or permanent storage. The operating system 202 can include Windows, Unix, Linux, etc.

[0160] In some embodiments, the electronic device can further include a display screen 22, an input / output interface 23, a communication interface 24, a power supply 25, and a communication bus 26.

[0161] Those skilled in the art can understand that the structure shown in the above embodiments does not constitute a limitation on the electronic device, and can include more or fewer components than those shown in the figure. Figure 2

[0162] The electronic device provided by the embodiments of the present application includes a memory and a processor, and the processor can realize the above-mentioned inverter fault tolerance control method when executing the program stored in the memory, and has the same beneficial effects.

[0163] Finally, the present application also provides an embodiment of a computer-readable storage medium. The computer-readable storage medium stores a computer program, and the computer program is executed by the processor to realize the steps recorded in the above method embodiments.

[0164] It can be understood that if the method in the above embodiments is realized in the form of a software function unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and executes all or part of the steps of the methods described in the embodiments of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (Read-Only Memory, ROM), a random access memory (Random Access Memory, RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0165] ​The above introduces in detail a kind of inverter fault tolerance control method, device and medium provided by the present application. Each embodiment in the specification is described in a progressive manner, and each embodiment focuses on the difference from other embodiments, and the same or similar parts of each embodiment can be referred to each other. For the device disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, it is described simply, and the related parts are described in the method part. It should be pointed out that for ordinary skilled in the art, without departing from the principles of the present application, the present application can be improved and modified, and these improvements and modifications also fall within the protection scope of the claims of the present application.

[0166] It should also be noted that in the present specification, relational terms such as first and second are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply that there is any such actual relationship or order between these entities or operations. Moreover, the terms "include", "contain" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or device. Without more limitations, the element defined by the statement "including a" does not exclude the presence of other identical elements in the process, method, article or device including the element.

Claims

1. A method for inverter fault-tolerant control, characterized in that, The method comprises the following steps: Collecting voltage and current signals corresponding to the inverter, and performing Hilbert-Huang transform on the voltage and current signals to obtain the center frequency and amplitude corresponding to the inverter; If the inverter fault type corresponding to the center frequency and the amplitude is determined according to the fault database, then the inverter fault degree is determined according to the energy, duration and amplitude corresponding to the inverter fault; wherein the fault database is constructed according to each inverter fault type and corresponding fault parameters, and is updated according to federated learning; The processing mode corresponding to the inverter is determined according to the inverter fault degree and the inverter fault type; Further comprising: If the inverter fault type corresponding to the center frequency and the amplitude is not determined according to the fault database, then the operating parameters corresponding to the inverter are obtained; According to the operating parameters, it is determined whether the corresponding inverter is a fault inverter; If the inverter is the fault inverter, then the new fault feature structured parameters corresponding to the operating parameters are determined according to the federated fault learning model; The new inverter fault type and the new fault parameters corresponding to the new fault feature structured parameters are determined, and the new inverter fault type and the new fault parameters are updated into the fault database; After updating the new inverter fault type and the new fault parameters into the fault database, it is judged whether the new inverter fault type and the new fault parameters meet the similarity requirements with each historical inverter fault type and each historical fault parameter stored in the fault database; If the similarity requirements are met, the historical inverter fault type and the historical fault parameter meeting the similarity requirements are obtained and taken as the target historical inverter fault type and the target historical fault parameter; The new inverter fault type, the new fault parameters, the target historical inverter fault type and the target historical fault parameter are fitted to generate inverter fault fitting type and fitting fault parameters, and are updated into the fault database.

2. The method of claim 1, wherein, The Hilbert-Huang transform is performed on the voltage and current signals to obtain the center frequency and amplitude corresponding to the inverter, comprising: Determining each signal component corresponding to the voltage and current signals based on the empirical mode decomposition algorithm; Performing Hilbert transform on each signal component to obtain the instantaneous frequency and instantaneous amplitude corresponding to each signal component; Obtaining the signal component corresponding to the instantaneous frequency within the abnormal frequency range in each signal component, and taking it as the target signal component; Determining the center frequency and amplitude corresponding to the inverter based on the instantaneous frequency and instantaneous amplitude corresponding to the target signal component.

3. The method of claim 2, wherein, The inverter fault degree is determined according to the energy, duration and amplitude corresponding to the inverter fault, comprising: Obtaining the total energy, short-time energy and abnormal modal energy corresponding to the inverter fault within the duration; Determining the abnormal modal energy ratio corresponding to the abnormal modal energy and the total energy; Determining the time-frequency entropy corresponding to the short-time energy and the total energy; determine a time ratio according to the duration corresponding to the inverter fault and a safe tolerance time corresponding to the inverter fault type; determine an amplitude ratio according to the amplitude corresponding to the inverter fault and an amplitude threshold value corresponding to the inverter fault type; determine the inverter fault degree according to the abnormal modal energy ratio, the time-frequency entropy, the time ratio and the amplitude ratio.

4. The method of claim 3, wherein, acquire the duration of the inverter fault, including: determine a first time when the instantaneous frequency corresponding to each signal component first appears in the abnormal frequency range; determine a target first time corresponding to each first time; determine a second time corresponding to the current time; determine the duration according to the target first time and the second time.

5. The method of claim 3 or 4, wherein, determine the inverter fault degree according to the abnormal modal energy ratio, the time-frequency entropy, the time ratio and the amplitude ratio, including: acquire an energy weight parameter, a time weight parameter and an amplitude weight parameter corresponding to the abnormal modal energy ratio, the time ratio and the amplitude ratio respectively; perform normalization processing on the time-frequency entropy to obtain an energy distribution entropy; acquire a distribution weight parameter corresponding to the energy distribution entropy; determine a fault severity index according to the abnormal modal energy ratio, the time ratio, the amplitude ratio, the energy distribution entropy, the energy weight parameter, the time weight parameter, the amplitude weight parameter and the distribution weight parameter; determine the inverter fault degree corresponding to the fault severity index.

6. The method of claim 5, wherein, determine the inverter fault degree corresponding to the fault severity index, including: when the fault severity index is less than a first fault threshold value, the inverter fault degree is a mild fault; when the fault severity index is not less than the first fault threshold value and less than a second fault threshold value, the inverter fault degree is a moderate fault; when the fault severity index is not less than the second fault threshold value, the inverter fault degree is a severe fault.

7. An electronic device, comprising: a memory for storing a computer program; a processor for executing the computer program to implement the steps of the inverter fault fault-tolerant control method according to any one of claims 1 to 6.

8. A computer-readable storage medium, characterized in that, The computer program is stored on the computer readable storage medium and is executed by the processor to implement the steps of the inverter fault fault-tolerant control method according to any one of claims 1 to 6.

Images