A method for quickly checking the tripping matrix of main transformer protection based on hardware parallel acquisition and software analysis

Abstract

The application provides a kind of fast checking main transformer protection tripping matrix method based on hardware parallel acquisition and software analysis combination, comprising: extracting fault waveform transient characteristic quantity using time-frequency analysis method;Construct a dynamic weight allocation model to fuse multi-source data for fault scenario discrimination;Adaptive notch filter is used to suppress power electronic disturbance;Based on digital twin verification platform, the main transformer dynamic model processed by model reduction is run on hardware-in-the-loop system;Construct a multi-port network equivalent model containing balanced winding electromagnetic coupling and introduce new energy output fluctuation coefficient correction parameter;Establish the differential protection action equation containing balanced winding current compensation term and adaptive braking coefficient;Generate protection tripping logic encoding matrix and realize logic correlation through rule base;Compare the encoding matrix with the actual contact state, and output the checking result after comparison, to complete the closed loop checking. The application realizes the automatic fast checking of main transformer protection tripping matrix, and significantly improves the checking efficiency and accuracy.

Description

Technical Field

[0001] This invention relates to the field of power system relay protection technology, and in particular to a method for rapidly verifying the main transformer protection trip matrix based on a combination of hardware parallel acquisition and software analysis. Background Technology

[0002] Currently, the verification of the main transformer protection trip matrix mainly relies on manual, line-by-line testing. Testers need to verify the operating logic of the protection devices one by one for different fault types. In existing technologies, some solutions attempt to achieve automated testing through fault recording playback, but these still have the following shortcomings: Firstly, in terms of model construction, existing transformer equivalent models are mostly static multi-port networks, which fail to consider the time-varying characteristics of transformer impedance parameters caused by fluctuations in new energy output, resulting in large deviations in protection setting.

[0003] Secondly, regarding the criterion design, the traditional differential protection operating equations do not include a balancing winding current compensation term. The zero-sequence current and third harmonic current in the balancing winding can easily interfere with the correct operation of the differential protection. Furthermore, the braking coefficient of the differential protection uses a fixed value, which cannot adaptively track transient ride-through current fluctuations caused by changes in the ramp rate of new energy vehicles.

[0004] Thirdly, in terms of logic verification, existing SCADA systems mainly support binary monitoring of the status of protection devices, lacking an automated generation and dynamic comparison mechanism for trip matrices.

[0005] Fourth, regarding the testing process, the existing verification methods involve independent steps such as fault waveform generation, trip output acquisition, logic comparison, and report generation, which do not form an automated closed-loop process, resulting in a long time consumption for a single complete verification.

[0006] To address the aforementioned problems, this invention proposes a method for rapidly verifying the main transformer protection trip matrix based on a combination of hardware parallel acquisition and software analysis. Summary of the Invention

[0007] Therefore, the purpose of this invention is to provide a method for rapidly verifying the main transformer protection trip matrix based on a combination of hardware parallel acquisition and software analysis, so as to solve the above problems.

[0008] To achieve the above-mentioned objectives, this invention provides a method for rapidly verifying the tripping matrix of a main transformer protection system based on a combination of hardware parallel acquisition and software analysis. The method includes the following steps: S1. The improved S-transform time-frequency analysis method is used to perform time-frequency analysis on the fault waveform and extract transient features. The high-frequency time resolution and low-frequency frequency resolution of the time-varying window function are set according to preset requirements. S2. Based on multi-source data fusion verification, a dynamic weight allocation model is constructed, which integrates fault recording waveform features, historical fault statistical features, and real-time equipment status to generate feature vectors, and uses neural networks to identify fault scenarios. S3. An adaptive notch filter is used to suppress power electronic disturbances and track the fundamental frequency shift of the power grid. S4. Based on the digital twin verification platform, the main transformer dynamic model is run on the hardware-in-the-loop system. The main transformer dynamic model is constructed based on the energy macroscopic expression method and the model reduction technique is used to compress the calculation scale. S5. Construct a multi-port network equivalent model that includes the electromagnetic coupling relationship of the balanced winding, determine the equivalent circuit parameters by measuring the DC resistance and short-circuit impedance of the winding, and introduce the new energy power output fluctuation coefficient to correct the model parameters. S6. Establish differential protection action equations that take into account the fluctuation characteristics of new energy sources, add a balance winding current compensation term to the differential criterion, and set an adaptive braking coefficient that changes with the output of new energy sources. S7. Generate the encoding matrix of the protection tripping logic, map the various action conditions of the protection device into a matrix, and realize the logical association between fault type and tripping command through the rule base; S8. Compare the generated encoding matrix with the actual collected contact status, output verification pass information or difference information based on the comparison result, and connect to the SCADA system through the standardized data interface to complete the verification closed loop.

[0009] Furthermore, the specific implementation steps of the improved S-transform time-frequency analysis method described in step S1 include: The time-domain expression of the sampled fault waveform by performing a discrete S-transform is as follows:

[0010] in, For input signal, It is a time-varying window function. For time shift factor, For frequency; The imaginary unit; The window function parameters are adaptively selected based on frequency, with different window function parameter values ​​corresponding to different frequency bands; The amplitude envelope and phase transition points of each frequency component are extracted as transient characteristic quantities.

[0011] Furthermore, the specific implementation steps of the multi-source data fusion verification in step S2 include: The system collects fault waveform files, SCADA telemetry and telecontrol data, and protection action records from the historical database in real time through standardized interfaces. Calculate the standard deviation for each type of data source. and channel capacity coefficient ; The fusion weights are calculated using a dynamic weighting model, which is as follows: in For the first The fusion weight of the data source class For the first Standard deviation of data source For the first Channel capacity coefficient for data source categories, where n is the total number of data source categories; The feature values ​​of various data sources are weighted according to the fusion weights to generate a weighted feature vector. The weighted feature vector is then input into a neural network to output a probability distribution of fault types.

[0012] Furthermore, the parameter adjustment steps for the adaptive notch filter in step S3 include: Real-time sampling of voltage signals to calculate the estimated fundamental frequency; The filter coefficients are updated based on the fundamental frequency estimate. The transfer function of the adaptive notch filter is:

[0013] Where r is the filter radius parameter. This is the normalized center angular frequency; An adaptive algorithm is used to iteratively update the filter parameters to minimize the output error; Calculate the differential current and braking current from the filtered current signal.

[0014] Furthermore, the specific construction steps of the digital twin verification platform described in step S4 include: Establish the dynamic equations of the principal variable based on the macroscopic energy expression method; A model reduction technique is used to reduce the dimensionality of the state space while retaining the state variables corresponding to the principal singular values. Implement parallel computation of the reduced-order model on an FPGA; It receives fault excitations from the protection action simulator, outputs the electrical quantity response of the digital twin main transformer, and compares it with the action output of the actual protection device.

[0015] Furthermore, the construction of the multi-port network equivalent model in step S5 includes the following sub-steps: Measure the DC resistance and short-circuit impedance of each winding of the main transformer; Establish the multiport network equations that include the mutual inductance coupling of the balanced windings. The multiport network equations are as follows:

[0016] The subscripts 1, 2, 3, and 4 correspond to the high-voltage winding, medium-voltage winding, low-voltage winding, and balance winding, respectively. , , , These represent the voltages at the ports of each winding. , , , These represent the currents at the ports of each winding. These are the mutual impedance parameters between windings; Introduce a new energy power output fluctuation coefficient and dynamically adjust the mutual impedance parameters.

[0017] Furthermore, the differential protection action equation in step S6 is:

[0018]

[0019] when The fault was initially identified as being within the designated area; among which... These are the current vectors for the high-voltage winding, medium-voltage winding, low-voltage winding, and balance winding, respectively. To balance the winding current compensation coefficient; the braking coefficient used in the braking criterion. With the ramp-up rate of new energy The braking coefficient changes monotonically, and its relationship with the new energy vehicle ramp rate is as follows:

[0020] in Based on the braking coefficient, For the ramp-up rate of new energy sources, Providing real-time power for new energy sources This is the preset scaling factor.

[0021] Furthermore, the generation process of the encoding matrix in step S7 includes: The fault type is encoded in binary. Binary encoding is used to encode the protection action type; The status of the output contact is encoded in binary format. Constructing a trip matrix The row index of the matrix corresponds to the combination of fault type and protection action, the column index corresponds to the status of the output contact, and the matrix elements... Indicates the first The first combination Should each exit contact point be activated? The tripping matrix is ​​automatically generated from the SCD file using a rule base, and during verification, the tripping matrix is ​​compared with the actual collected contact status matrix. Perform an XOR comparison, the logic of which is as follows:

[0022] in Let be the difference matrix, if If there are non-zero elements in the table, output the difference item at the corresponding position.

[0023] Furthermore, the method implements a closed-loop verification process through the following hardware modules: Protection action simulator: Generates fault waveform messages; Trip output collector: Collects the status of trip output contacts; Main control analysis unit: parses the protection logic relationship in the SCD file and performs trip matrix comparison and verification; Visual verification interface: Displays status information at different levels in a hierarchical manner and identifies abnormal items in a visual way; Matrix report generation module: Generates a trip matrix report based on the verification results.

[0024] Furthermore, the method performs a digital twin pre-verification step before the verification begins: After the digital twin verification platform is built, the fault scenario to be tested is input into the hardware-in-the-loop system, the reduced-order digital twin model of the main transformer is run, the expected results of each logic entry in the trip matrix are obtained in advance, and the expected results are compared with the actual action output of the protection device.

[0025] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention performs time-frequency analysis on fault waveforms using an improved S-transform time-frequency analysis method. Its time-varying window function can adaptively adjust the time resolution and frequency resolution according to the frequency. The high-frequency band ensures a high time resolution to accurately locate the fault initiation time, while the low-frequency band maintains a constant frequency resolution to accurately separate the fundamental and harmonic components, thereby achieving holographic analysis of transient fault characteristics.

[0026] 2. This invention constructs a dynamic weight allocation model, comprehensively considers the statistical characteristics and channel quality of various data sources, and adaptively allocates fusion weights, realizing deep coupling of fault recording data, SCADA status variables and historical statistical data, which significantly improves the accuracy of fault scenario identification.

[0027] 3. This invention uses an adaptive notch filter to suppress the interference of high-order harmonics generated by new energy inverters on protection criteria. The center frequency of the filter can track the fundamental frequency deviation of the power grid in real time, effectively ensuring the stability of differential current and braking current calculation. Attached Figure Description

[0028] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only preferred embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0029] Figure 1 This is a schematic diagram of the overall process of a method for rapidly verifying the main transformer protection trip matrix based on a combination of hardware parallel acquisition and software analysis, provided by an embodiment of the present invention.

[0030] Figure 2 This is a schematic diagram of the overall system framework architecture of a method for rapidly verifying the main transformer protection tripping matrix based on the combination of hardware parallel acquisition and software analysis provided in an embodiment of the present invention. Detailed Implementation

[0031] The principles and features of the present invention are described below with reference to the accompanying drawings. The listed embodiments are only used to explain the present invention and are not intended to limit the scope of the present invention.

[0032] Reference Figure 1 This embodiment provides a method for rapidly verifying the main transformer protection trip matrix based on a combination of hardware parallel acquisition and software analysis. The method includes the following steps: S1. The improved S-transform time-frequency analysis method is used to perform time-frequency analysis on the fault waveform and extract transient features. The high-frequency time resolution and low-frequency frequency resolution of the time-varying window function are set according to preset requirements. S2. Based on multi-source data fusion verification, a dynamic weight allocation model is constructed, which integrates fault recording waveform features, historical fault statistical features, and real-time equipment status to generate feature vectors, and uses neural networks to identify fault scenarios. S3. An adaptive notch filter is used to suppress power electronic disturbances and track the fundamental frequency shift of the power grid. S4. Based on the digital twin verification platform, the main transformer dynamic model is run on the hardware-in-the-loop system. The main transformer dynamic model is constructed based on the energy macroscopic expression method and the model reduction technique is used to compress the calculation scale. S5. Construct a multi-port network equivalent model that includes the electromagnetic coupling relationship of the balanced winding, determine the equivalent circuit parameters by measuring the DC resistance and short-circuit impedance of the winding, and introduce the new energy power output fluctuation coefficient to correct the model parameters. S6. Establish differential protection action equations that take into account the fluctuation characteristics of new energy sources, add a balance winding current compensation term to the differential criterion, and set an adaptive braking coefficient that changes with the output of new energy sources. S7. Generate the encoding matrix of the protection tripping logic, map the various action conditions of the protection device into a matrix, and realize the logical association between fault type and tripping command through the rule base; S8. Compare the generated encoding matrix with the actual collected contact status, output verification pass information or difference information based on the comparison result, and connect to the SCADA system through the standardized data interface to complete the verification closed loop.

[0033] The specific implementation steps of the improved S-transform time-frequency analysis method described in step S1 include: The time-domain expression of the sampled fault waveform by performing a discrete S-transform is as follows:

[0034] in, For input signal, It is a time-varying window function. For time shift factor, For frequency; The imaginary unit; The window function parameters are adaptively selected based on frequency, with different window function parameter values ​​corresponding to different frequency bands; The amplitude envelope and phase transition points of each frequency component are extracted as transient characteristic quantities.

[0035] The specific implementation steps of the multi-source data fusion verification in step S2 include: The system collects fault waveform files, SCADA telemetry and telecontrol data, and protection action records from the historical database in real time through standardized interfaces. Calculate the standard deviation for each type of data source. and channel capacity coefficient ; The fusion weights are calculated using a dynamic weighting model, which is as follows: in For the first The fusion weight of the data source class For the first Standard deviation of data source For the first Channel capacity coefficient for data source categories, where n is the total number of data source categories; The feature values ​​of various data sources are weighted according to the fusion weights to generate a weighted feature vector. The weighted feature vector is then input into a neural network to output a probability distribution of fault types.

[0036] The parameter adjustment steps for the adaptive notch filter in step S3 include: Real-time sampling of voltage signals to calculate the estimated fundamental frequency; The filter coefficients are updated based on the fundamental frequency estimate. The transfer function of the adaptive notch filter is:

[0037] Where r is the filter radius parameter. This is the normalized center angular frequency; An adaptive algorithm is used to iteratively update the filter parameters to minimize the output error; Calculate the differential current and braking current from the filtered current signal.

[0038] The specific construction steps of the digital twin verification platform described in step S4 include: Establish the dynamic equations of the principal variable based on the macroscopic energy expression method; A model reduction technique is used to reduce the dimensionality of the state space while retaining the state variables corresponding to the principal singular values. Implement parallel computation of the reduced-order model on an FPGA; It receives fault excitations from the protection action simulator, outputs the electrical quantity response of the digital twin main transformer, and compares it with the action output of the actual protection device.

[0039] The construction of the multi-port network equivalent model in step S5 includes the following sub-steps: Measure the DC resistance and short-circuit impedance of each winding of the main transformer; Establish the multiport network equations that include the mutual inductance coupling of the balanced windings. The multiport network equations are as follows:

[0040] The subscripts 1, 2, 3, and 4 correspond to the high-voltage winding, medium-voltage winding, low-voltage winding, and balance winding, respectively. , , , These represent the voltages at the ports of each winding. , , , These represent the currents at the ports of each winding. This refers to the mutual impedance parameter between windings; Introduce a new energy power output fluctuation coefficient and dynamically adjust the mutual impedance parameters.

[0041] The differential protection action equation in step S6 is:

[0042]

[0043] when The fault was initially identified as being within the designated area; among which... These are the current vectors for the high-voltage winding, medium-voltage winding, low-voltage winding, and balance winding, respectively. To balance the winding current compensation coefficient; the braking coefficient used in the braking criterion. With the ramp-up rate of new energy The braking coefficient changes monotonically, and its relationship with the new energy vehicle ramp rate is as follows:

[0044] in Based on the braking coefficient, For the ramp-up rate of new energy sources, Providing real-time power for new energy sources This is the preset scaling factor.

[0045] The process of generating the encoding matrix in step S7 includes: The fault type is encoded in binary. Binary encoding is used to encode the protection action type; The status of the exit contact is encoded in binary. Constructing a trip matrix The row index of the matrix corresponds to the combination of fault type and protection action, the column index corresponds to the status of the output contact, and the matrix elements... Indicates the first The first combination Should the exit contact point be activated? The tripping matrix is ​​automatically generated from the SCD file using a rule base, and during verification, the tripping matrix is ​​compared with the actual collected contact status matrix. Perform an XOR comparison, the logic of which is as follows:

[0046] in Let be the difference matrix, if If there are non-zero elements in the table, output the difference item at the corresponding position.

[0047] The method implements a closed-loop verification process through the following hardware modules: Protection action simulator: Generates fault waveform messages; Trip output collector: Collects the status of trip output contacts; Main control analysis unit: parses the protection logic relationship in the SCD file and performs trip matrix comparison and verification; Visual verification interface: Displays status information at different levels in a hierarchical manner and identifies abnormal items in a visual way; Matrix report generation module: Generates a trip matrix report based on the verification results.

[0048] This method performs a digital twin pre-verification step before the verification begins: After the digital twin verification platform is built, the fault scenario to be tested is input into the hardware-in-the-loop system, the reduced-order digital twin model of the main transformer is run, the expected results of each logic entry in the trip matrix are obtained in advance, and the expected results are compared with the actual action output of the protection device.

[0049] For example, the overall system architecture of the rapid verification method for main transformer protection tripping matrix based on the combination of hardware parallel acquisition and software analysis proposed in this invention is as follows: Figure 2 As shown, it mainly consists of five components: a protection action simulator, a trip output acquisition unit, a main control analysis unit, a visual verification interface, and a matrix report generation module. The data interaction process between the modules is as follows: The protection action simulator generates a fault waveform based on the test sequence and sends it to the protection device via a GOOSE message; after responding to the fault waveform, the protection device triggers a trip output action; the trip output collector collects the trip output contact status and uploads it to the main control analysis unit; the main control analysis unit compares the actual trip output status with the preset trip matrix and generates a verification result; the visual verification interface displays the verification result and the differences; the matrix report generation module automatically generates a trip matrix report based on the verification result.

[0050] The specific implementation of the present invention is carried out according to the following steps: Step 1: Import SCD file parsing and protection logic The main control analysis unit loads the substation configuration description file (SCD file) and parses the GOOSE control block configuration, the input and output signal mapping relationship of each intelligent electronic device, and the action logic parameters of the protection device from it.

[0051] Specifically, the main control analysis unit constructs a protection logic rule base based on the Drools rule engine. This rule base covers major protection functions such as differential protection, overcurrent protection, and non-electrical quantity protection. The rule engine transforms the logical nodes and data attributes defined in the SCD file into executable rule entries according to the three-level mapping relationship of "protection action, output contact, circuit breaker number," providing a logical foundation for the subsequent automatic generation of the trip matrix.

[0052] Step 2: Configure test cases Users select the items to be tested through a visual verification interface. Available test items include scenarios such as in-zone faults, out-of-zone faults, transitional faults, and harmonic injection. The system automatically generates test sequences based on user selections. Each test sequence contains multi-dimensional parameters such as fault type, fault location, transition resistance value, and renewable energy output level.

[0053] The system supports two modes: single-point-of-failure testing and batch sequence testing.

[0054] Step 3: Start the automated test and generate a report. The protection action simulator sends GOOSE messages one by one according to the test sequence. In a preferred embodiment, the protection action simulator is implemented based on an FPGA, with a timing accuracy better than 0.1ms, and has a built-in IEEE 1588 precise time synchronization protocol.

[0055] The trip output acquisition unit collects the status of the output contacts after the protection device operates. In a preferred embodiment, the trip output acquisition unit includes a multi-channel optically isolated input board, supporting parallel acquisition of multiple hard contacts, and is equipped with a fiber optic interface card for parsing the FT3 protocol messages of the protection device. The acquired contact signals are then processed by a DSP chip to perform a signal de-jitter algorithm before being uploaded to the main control analysis unit.

[0056] After receiving the actual trip output status, the main control analysis unit performs a comparison as follows: Based on the test sequence generated in the second step, the main control analysis unit parses the expected trip matrix under the current fault scenario from the rule base. The actual collected contact state matrix and Perform an XOR comparison:

[0057] The matrix's row and column encoding rules are as follows: the row index is composed of the fault type code and the protection action code, and the column index corresponds to the physical channel number of the trip output collector. If the difference matrix... If all elements in the checklist are 0, it means the checklist has passed; if If a non-zero element is found, it is considered a discrepancy, and the system will issue a warning through the traffic light alarm mechanism on the visual verification interface.

[0058] The visual verification interface adopts a layered architecture to display the verification status: the physical layer displays the actual on / off status of hard contacts, the logic layer displays the triggering status of each protection function, and the application layer displays the rule matching results of the trip matrix.

[0059] After verification, the matrix report generation module automatically fills in the Excel template and generates a complete trip matrix verification report. The report also supports exporting the CID configuration file in IEC 61850-6 format.

[0060] Step 4: Manually review the discrepancies and update the set values. After viewing the details of the discrepancies through the visual verification interface, users manually confirm each discrepancy. For discrepancies confirmed to be caused by improper protection setting, users write new setting parameters to the protection device through the main control analysis unit.

[0061] The system supports the historical review function for verification records.

[0062] The technical effects of the present invention will be illustrated below through specific application examples: A trip matrix verification was performed on a main transformer protection device (model PCS-978T2-DG) with a balancing winding at a 220kV renewable energy collection substation. The substation has a renewable energy penetration rate of 35%, a main transformer capacity of 240MVA, voltage levels of 220kV / 110kV / 35kV, and a balancing winding voltage level of 10kV.

[0063] Before verification, users can select test items through a visual verification interface, including single-phase grounding fault, two-phase short-circuit fault, two-phase grounding fault, three-phase short-circuit fault, and transitional fault, covering a total of 32 tripping logic combinations.

[0064] After the system automatically generates the test sequence, the protection action simulator sends GOOSE messages according to the sequence, and the trip output collector synchronously collects the status of multiple output contacts of the protection device. After the main control analysis unit performs an XOR comparison, the visual verification interface displays that all test items have passed. The matrix report generation module automatically outputs a verification report, and a single complete verification takes 3.2 minutes. In comparison, the traditional manual verification method takes much longer. The test of 32 tripping logics for the same protection device requires the testers to simulate different fault conditions, check the protection action reports and record the comparison results one by one. The average time for a complete verification is 4.5 hours.

[0065] The performance of this invention in distinguishing between faults inside and outside the protection zone was verified. The test sequence included 20 sets of faults inside the protection zone and 20 sets of faults outside the protection zone. The differential protection operating equation described in step S6 of this invention was adopted, wherein the balance winding current compensation coefficient... The setting principle is determined based on the ratio of the rated current of the balancing winding to the rated current of the main winding; in this embodiment, it is taken as 0.3; the basic braking coefficient... Selected according to the transformer's rated capacity and the recommended value in the protection device's instruction manual, taking 0.5; preset proportional coefficient. Take 0.05; New energy ramp rate Calculations were performed in real time based on the measured output data from the collection station. Under the above parameter configuration, all 40 test sets were correctly identified, with a false positive rate of 0 for faults outside the designated area.

[0066] In contrast, the same test sequence was checked using the traditional differential protection criterion (without a balance winding current compensation term and a braking coefficient fixed at 0.5). Among the 40 test sets, 6 sets of external faults were misjudged, with a misjudgment rate of 15%.

[0067] To further verify the timeliness of transient feature identification of the present invention, time-frequency analysis tests were conducted on 100 sets of fault waveforms of different types on the RTDS real-time digital simulation platform. The fault types covered 25 sets each of single-phase grounding, two-phase short circuit, three-phase short circuit and transitional fault.

[0068] The improved S-transform time-frequency analysis method described in step S1 of this invention has a time resolution better than 0.1 ms in the high-frequency band and a constant frequency resolution in the low-frequency band for its time-varying window function. In RTDS experiments, the average transient feature identification time for 100 fault groups was 1.2 ms, as shown in Table 1. Table 1

[0069] In contrast, the traditional Fourier transform method was used to extract features from the same fault waveform, with an average identification time of 2.0 ms.

[0070] In a photovoltaic power station containing a balancing winding transformer, the system uses the method described in this invention to verify the trip matrix. Before the verification begins, the user selects to perform a digital twin pre-verification step.

[0071] Specifically, all 32 fault scenarios to be tested were input into the FPGA hardware-in-the-loop system, and the main transformer digital twin model (originally 1000-order, reduced to 200-order with a model accuracy error of less than 2%) was run to obtain the expected results for each logic entry in the trip matrix. After comparing the expected results with the setting logic of the actual protection device, the system pre-screened two logic entries that might have discrepancies. During field testing, the user focused on these two pre-screened entries, while the remaining 30 entries were directly verified using the digital twin. The final field test completed all verifications in just 1.9 minutes. The two pre-screened entries were confirmed on-site; one was due to a deviation in the protection setting, which passed verification after setting correction; the other was caused by a slight deviation between the digital twin model parameters and the actual equipment parameters, which was ruled out as a problem with the protection device itself after manual review.

[0072] To further verify the performance of this invention in multi-source data fusion, 12 months of historical operating data from a 220kV renewable energy collection station were used for testing. A total of 1587 fault events under different operating conditions were selected as test samples.

[0073] The multi-source data fusion verification method described in step S2 of this invention is adopted: fault waveform files, SCADA telemetry and telecontrol data, and protection action records in the historical database are collected in real time through the IEC 61850 interface, and the standard deviation is calculated for each type of data source. and channel capacity coefficient The fusion weights are calculated according to the dynamic weight allocation model, and a 128-dimensional weighted feature vector is generated and then input into a 3-layer deep neural network (128-64-32 node structure) to output the probability distribution of fault types.

[0074] Test results show that the method of this invention correctly identified 1581 fault events and misidentified 6, achieving a fault identification accuracy of 99.7%. In comparison, when using fault waveform data alone for identification, it correctly identified 1465 fault events and misidentified 122, achieving an accuracy of 92.3%; when using SCADA status variables alone, it correctly identified 1391 fault events and misidentified 196, achieving an accuracy of 87.6%. When using a fixed-weight strategy for fusion, it correctly identified 1544 fault events and misidentified 43, achieving an accuracy of 97.3%. The dynamic weight model of this invention misidentified 6 fault events.

[0075] 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 spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for rapidly verifying the tripping matrix of a main transformer protection system based on a combination of hardware parallel acquisition and software analysis, characterized in that, The method includes the following steps: S1. The improved S-transform time-frequency analysis method is used to perform time-frequency analysis on the fault waveform and extract transient features. The high-frequency time resolution and low-frequency frequency resolution of the time-varying window function are set according to preset requirements. S2. Based on multi-source data fusion verification, a dynamic weight allocation model is constructed, which integrates fault recording waveform features, historical fault statistical features, and real-time equipment status to generate feature vectors, and uses neural networks to identify fault scenarios. S3. An adaptive notch filter is used to suppress power electronic disturbances and track the fundamental frequency shift of the power grid. S4. Based on the digital twin verification platform, the main transformer dynamic model is run on the hardware-in-the-loop system. The main transformer dynamic model is constructed based on the energy macroscopic expression method and the model reduction technique is used to compress the calculation scale. S5. Construct a multi-port network equivalent model that includes the electromagnetic coupling relationship of the balanced winding, determine the equivalent circuit parameters by measuring the DC resistance and short-circuit impedance of the winding, and introduce the new energy power output fluctuation coefficient to correct the model parameters. S6. Establish differential protection action equations that take into account the fluctuation characteristics of new energy sources, add a balance winding current compensation term to the differential criterion, and set an adaptive braking coefficient that changes with the output of new energy sources. S7. Generate the encoding matrix of the protection tripping logic, map the various action conditions of the protection device into a matrix, and realize the logical association between fault type and tripping command through the rule base; S8. Compare the generated encoding matrix with the actual collected contact status, output verification pass information or difference information based on the comparison result, and connect to the SCADA system through the standardized data interface to complete the verification closed loop.

2. The method for rapidly verifying the main transformer protection trip matrix based on a combination of hardware parallel acquisition and software analysis as described in claim 1, characterized in that, The specific implementation steps of the improved S-transform time-frequency analysis method described in step S1 include: The time-domain expression of the sampled fault waveform by performing a discrete S-transform is as follows: in, For input signal, It is a time-varying window function. For time shift factor, For frequency; The imaginary unit; The window function parameters are adaptively selected based on frequency, with different window function parameter values ​​corresponding to different frequency bands; The amplitude envelope and phase transition points of each frequency component are extracted as transient characteristic quantities.

3. The method for rapidly verifying the main transformer protection trip matrix based on a combination of hardware parallel acquisition and software analysis as described in claim 1, characterized in that, The specific implementation steps of the multi-source data fusion verification in step S2 include: The system collects fault waveform files, SCADA telemetry and telecontrol data, and protection action records from the historical database in real time through standardized interfaces. Calculate the standard deviation for each type of data source. and channel capacity coefficient ; The fusion weights are calculated using a dynamic weighting model, which is as follows: in For the first The fusion weight of the data source class For the first Standard deviation of data source For the first Channel capacity coefficient for data source categories, where n is the total number of data source categories; The feature values ​​of various data sources are weighted according to the fusion weights to generate a weighted feature vector. The weighted feature vector is then input into a neural network to output a probability distribution of fault types.

4. The method for rapidly verifying the main transformer protection trip matrix based on a combination of hardware parallel acquisition and software analysis as described in claim 1, characterized in that, The parameter adjustment steps for the adaptive notch filter in step S3 include: Real-time sampling of voltage signals to calculate the estimated fundamental frequency; The filter coefficients are updated based on the fundamental frequency estimate. The transfer function of the adaptive notch filter is: Where r is the filter radius parameter. This is the normalized center angular frequency; An adaptive algorithm is used to iteratively update the filter parameters to minimize the output error; Calculate the differential current and braking current from the filtered current signal.

5. The method for rapidly verifying the main transformer protection trip matrix based on a combination of hardware parallel acquisition and software analysis as described in claim 1, characterized in that, The specific construction steps of the digital twin verification platform described in step S4 include: Establish the dynamic equations of the principal variable based on the macroscopic energy expression method; A model reduction technique is used to reduce the dimensionality of the state space while retaining the state variables corresponding to the principal singular values. Implement parallel computation of the reduced-order model on an FPGA; It receives fault excitations from the protection action simulator, outputs the electrical quantity response of the digital twin main transformer, and compares it with the action output of the actual protection device.

6. The method for rapidly verifying the main transformer protection trip matrix based on a combination of hardware parallel acquisition and software analysis as described in claim 1, characterized in that, The construction of the multi-port network equivalent model in step S5 includes the following sub-steps: Measure the DC resistance and short-circuit impedance of each winding of the main transformer; Establish the multiport network equations that include the mutual inductance coupling of the balanced windings. The multiport network equations are as follows: The subscripts 1, 2, 3, and 4 correspond to the high-voltage winding, medium-voltage winding, low-voltage winding, and balance winding, respectively. , , , These represent the voltages at the ports of each winding. , , , These represent the currents at the ports of each winding. These are the mutual impedance parameters between windings; Introduce a new energy power output fluctuation coefficient and dynamically adjust the mutual impedance parameters.

7. The method for rapidly verifying the main transformer protection trip matrix based on a combination of hardware parallel acquisition and software analysis as described in claim 1, characterized in that, The differential protection action equation in step S6 is: when The fault was initially identified as being within the designated area; among which... These are the current vectors for the high-voltage winding, medium-voltage winding, low-voltage winding, and balance winding, respectively. To balance the winding current compensation coefficient; the braking coefficient used in the braking criterion. With the ramp-up rate of new energy The braking coefficient changes monotonically, and its relationship with the new energy vehicle ramp rate is as follows: in Based on the braking coefficient, For the ramp-up rate of new energy sources, Providing real-time power for new energy sources This is the preset scaling factor.

8. The method for rapidly verifying the main transformer protection trip matrix based on a combination of hardware parallel acquisition and software analysis as described in claim 1, characterized in that, The process of generating the encoding matrix in step S7 includes: The fault type is encoded in binary. Binary encoding is used to encode the protection action type; The status of the exit contact is encoded in binary. Constructing a trip matrix The row index of the matrix corresponds to the combination of fault type and protection action, the column index corresponds to the status of the output contact, and the matrix elements... Indicates the first The first combination Should the exit contact point be activated? The tripping matrix is ​​automatically generated from the SCD file using a rule base, and during verification, the tripping matrix is ​​compared with the actual collected contact status matrix. Perform an XOR comparison, the logic of which is as follows: in Let be the difference matrix, if If there are non-zero elements in the table, output the difference item at the corresponding position.

9. The method for rapidly verifying the main transformer protection trip matrix based on a combination of hardware parallel acquisition and software analysis as described in claim 1, characterized in that, The method implements a closed-loop verification process through the following hardware modules: Protection action simulator: Generates fault waveform messages; Trip output collector: Collects the status of trip output contacts; Main control analysis unit: parses the protection logic relationship in the SCD file and performs trip matrix comparison and verification; Visual verification interface: Displays status information at different levels in a hierarchical manner and identifies abnormal items in a visual way; Matrix report generation module: Generates a trip matrix report based on the verification results.

10. The method for rapid verification of the main transformer protection trip matrix based on the combination of hardware parallel acquisition and software analysis according to any one of claims 1 to 9, characterized in that, The method performs a digital twin pre-verification step before the verification begins: After the digital twin verification platform is built, the fault scenario to be tested is input into the hardware-in-the-loop system, the reduced-order digital twin model of the main transformer is run, the expected results of each logic entry in the trip matrix are obtained in advance, and the expected results are compared with the actual action output of the protection device.

Images