Relay protection device action time evaluation method and system based on machine learning algorithm

Abstract

The application discloses a relay protection device action time evaluation method and system based on a machine learning algorithm, and the method comprises the following steps: simulating different operation conditions of a power system, traversing various types of internal area faults according to a preset number of times, establishing a relay protection action time sample set based on the action time of the relay protection device after the internal area fault occurs; calculating the relay protection action time center through a preset machine learning algorithm, stopping the calculation when the machine learning algorithm converges, and obtaining the relay protection action time center element and the corresponding action time center subset; setting the representative probability density of the action time center element based on the action time center subset; and calculating the evaluation index of the relay protection device action time based on the action time center element and the representative probability density.

Description

Technical Field

[0001] This invention relates to the field of relay protection technology, and more specifically, to a method and system for evaluating the operating time of relay protection devices based on machine learning algorithms. Background Technology

[0002] Relay protection devices are one of the important means of ensuring the safe and stable operation of power systems. Power system dynamic simulation systems can realistically simulate various complex faults and other abnormal operating conditions in power systems by flexibly establishing test models of different voltage levels, scales, and types. This enables targeted testing and verification of various relay protection devices, such as line protection, transformer protection, bus protection, reactor protection, and generator protection. It allows for systematic and comprehensive testing of the device's functions and overall performance. At the same time, the dynamic simulation laboratory can be used to conduct performance comparison tests between relay protection devices from different manufacturers and different models from the same manufacturer.

[0003] Currently, the operation time testing of protection devices mainly involves passability testing based on relevant standards to verify whether the operation time meets the technical specifications. Because the operation time of relay protection devices is affected by operating conditions, fault types, protection principles, operating logic, and the performance of the device's hardware and software, it is usually a non-continuous discrete time variable with a certain degree of time region clustering. Dynamic model testing methods can obtain a large number of relay protection device operation times under various operating conditions. However, how to comprehensively evaluate the quality of operation time while ensuring that the standard requirements are met has always been a major challenge in device performance evaluation. Summary of the Invention

[0004] The present invention provides a method and system for evaluating the operating time of relay protection devices based on machine learning algorithms, in order to solve the problem of how to evaluate the operating time of relay protection devices based on machine learning algorithms.

[0005] To address the above problems, this invention provides a method for evaluating the operating time of relay protection devices based on machine learning algorithms, the method comprising:

[0006] Simulate different operating conditions of the power system, traverse various types of faults within the zone according to a preset number of times, and establish a sample set of relay protection action time based on the action time of the relay protection device after the occurrence of the fault within the zone.

[0007] The relay protection action time center is calculated by a preset machine learning algorithm. When the machine learning algorithm converges, the calculation stops and the action time center element and the corresponding action time center subset are obtained.

[0008] The representative probability density of the action time center element is set based on the subset of action time centers;

[0009] The evaluation index for the operation time of the relay protection device is calculated based on the central element of the operation time and the representative probability density.

[0010] Preferably, before simulating different operating conditions of the power system and traversing various types of intra-regional faults, the method further includes:

[0011] A relay protection device operation simulation test system is established, which includes a primary power system simulation unit, a fault simulation unit, a control system unit, a fault recording unit, and an external time synchronization unit.

[0012] The wiring configuration of the primary power system simulation unit is consistent with the typical wiring scheme of the prototype line and equipment components protected by the relay protection device.

[0013] The fault simulation unit is used to simulate various metallic and transition resistance fault types, including single-phase ground fault, two-phase short-circuit ground fault, two-phase short-circuit fault, three-phase short-circuit fault, and three-phase short-circuit ground fault.

[0014] The control system unit is used to change the operating mode of the primary power system simulation unit and the fault types simulated by the fault simulation unit;

[0015] The fault recording unit is used to record the electrical waveforms of the primary power system simulation unit and the operating time of the relay protection device.

[0016] The external time synchronization unit is used to provide a unified external time reference for the relay protection device operation simulation test system.

[0017] Preferably, it further includes:

[0018] Connect the relay protection device to the relay protection device operation simulation test system;

[0019] The relay protection device is configured, including setting configuration and model file configuration;

[0020] The relay protection device operation simulation test system is debugged, including: analog quantity comparison, switch quantity verification, transmission test and communication check.

[0021] Preferably, the different operating conditions of the power system include: operating conditions with different system equivalent capacity and different loads, system frequency deviation, and operating conditions with manually operated faulty components;

[0022] The faults within the zone include: single-phase ground fault, two-phase short-circuit ground fault, two-phase short-circuit fault, three-phase short-circuit fault, and three-phase short-circuit ground fault.

[0023] Preferably, the representative probability density of the action time center element is:

[0024] f i =crad(C i ) / crad(T)

[0025] Among them, crad(C i ) represents the number of samples in the central subset of the operating time, and crad(T) represents the number of samples in the sample set of the relay protection operating time.

[0026] Preferably, the preset number of times to traverse various types of faults within the area is increased.

[0027] Preferably, the step of calculating the relay protection action time center using a preset machine learning algorithm includes:

[0028] Calculate the similarity of the relay protection operation time sample set:

[0029]

[0030] Where s(i,j)=-||t i -t j || 2 ;t i Let t be the action time recorded in the i-th trial. j Let s(i,j) be the action time recorded in the j-th trial, and let s(i,j) be the similarity between action time samples i and j.

[0031] Based on the similarity of the relay protection action time sample set, the machine learning algorithm is initialized:

[0032] S1: Initialize the attraction matrix R and the belonging matrix A to zero matrices;

[0033] S2: Update the attraction matrix R:

[0034]

[0035] Among them, a i (i, j) represents the attraction between action time samples i and j, r i (i, j) represents the degree of belonging between action time samples i and j, and k represents action time sample k;

[0036] S3: Update the affiliation matrix A:

[0037]

[0038] m is the number of iterations;

[0039] S4: Attenuate the attraction matrix R and the membership matrix A based on the attenuation coefficient λ:

[0040] r m+1 (i, j) = λ × r m (i,j)+(1-λ)×r m+1 (i, j)

[0041] a m+1 (i, j) = λ × a m (i,j)+(1-λ)×a m+1 (i, j)

[0042] Repeat steps S2, S3 and S4 until the attraction matrix R and the membership matrix A no longer change in a preset number of iterations or the maximum number of iterations is reached, then end the calculation.

[0043] The element with the largest a+r is selected as the center element of the action time, and the selection rule is: max{a(i,j)+r(i,j)}

[0044] Where a+r is the sum of attraction and belonging.

[0045] Based on another aspect of the present invention, the present invention provides a relay protection device operation time evaluation system based on machine learning algorithm, the system comprising:

[0046] The initial unit is used to simulate different operating conditions of the power system, traverse various types of faults in the zone a preset number of times, and establish a sample set of relay protection action times based on the action time of the relay protection device after the occurrence of the fault in the zone.

[0047] The acquisition unit is used to calculate the relay protection action time center through a preset machine learning algorithm. When the machine learning algorithm converges, the calculation is stopped, and the action time center element and the corresponding action time center subset of the action time are acquired.

[0048] The setting unit is used to set the representative probability density of the action time center elements based on the subset of action time centers;

[0049] The result unit is used to calculate the evaluation index of the relay protection device's operating time based on the central element of the operating time and the representative probability density.

[0050] Preferably, the system further includes a setup unit for:

[0051] A relay protection device operation simulation test system is established, which includes a primary power system simulation unit, a fault simulation unit, a control system unit, a fault recording unit, and an external time synchronization unit.

[0052] The wiring configuration of the primary power system simulation unit is consistent with the typical wiring scheme of the prototype line and equipment components protected by the relay protection device.

[0053] The fault simulation unit is used to simulate various metallic and transition resistance fault types, including single-phase ground fault, two-phase short-circuit ground fault, two-phase short-circuit fault, three-phase short-circuit fault, and three-phase short-circuit ground fault.

[0054] The control system unit is used to change the operating mode of the primary power system simulation unit and the fault types simulated by the fault simulation unit;

[0055] The fault recording unit is used to record the electrical waveforms of the primary power system simulation unit and the operating time of the relay protection device.

[0056] The external time synchronization unit is used to provide a unified external time reference for the relay protection device operation simulation test system.

[0057] Preferably, the establishing unit is further configured to:

[0058] Connect the relay protection device to the relay protection device operation simulation test system;

[0059] The relay protection device is configured, including setting configuration and model file configuration;

[0060] The relay protection device operation simulation test system is debugged, including: analog quantity comparison, switch quantity verification, transmission test and communication check.

[0061] Preferably, the different operating conditions of the power system include: operating conditions with different system equivalent capacity and different loads, system frequency deviation, and operating conditions with manually operated faulty components;

[0062] The faults within the zone include: single-phase ground fault, two-phase short-circuit ground fault, two-phase short-circuit fault, three-phase short-circuit fault, and three-phase short-circuit ground fault.

[0063] Preferably, the representative probability density of the action time center element is:

[0064] f i =crad(C i ) / crad(T)

[0065] Among them, crad(C i ) represents the number of samples in the central subset of the operating time, and crad(T) represents the number of samples in the sample set of the relay protection operating time.

[0066] Preferably, the initial unit is further configured to: increase the preset number of times to traverse various types of faults within the area.

[0067] Preferably, the acquisition unit is used to calculate the relay protection action time center through a preset machine learning algorithm, including:

[0068] Calculate the similarity of the relay protection operation time sample set:

[0069]

[0070] Where s(i,j)=-||t i -t j || 2 ;t i Let t be the action time recorded in the i-th trial. j Let s(i,j) be the action time recorded in the j-th trial, and let s(i,j) be the similarity between action time samples i and j.

[0071] Based on the similarity of the relay protection action time sample set, the machine learning algorithm is initialized:

[0072] S1: Initialize the attraction matrix R and the belonging matrix A to zero matrices;

[0073] S2: Update the attraction matrix R:

[0074]

[0075] Among them, a i (i,j) represents the attraction between action time samples i and j, r i (i,j) represents the degree of belonging between action time samples i and j, and k represents action time sample k;

[0076] S3: Update the affiliation matrix A:

[0077]

[0078] m is the number of iterations;

[0079] S4: Attenuate the attraction matrix R and the membership matrix A based on the attenuation coefficient λ:

[0080] r m+1 (i, j) = λ × r m (i,j)+(1-λ)×r m+1 (i, j)

[0081] a m+1 (i, j) = λ × a m (i,j)+(1-λ)×a m+1 (i, j)

[0082] Repeat steps S2, S3 and S4 until the attraction matrix R and the membership matrix A no longer change in a preset number of iterations or the maximum number of iterations is reached, then end the calculation.

[0083] The element with the largest a+r is selected as the center element of the action time, and the selection rule is: max{a(i,j)+r(i,j)}

[0084] Where a+r is the sum of attraction and belonging.

[0085] This invention provides a method and system for evaluating the operating time of relay protection devices based on machine learning algorithms. The method includes: simulating different operating conditions of a power system, traversing various types of intra-zone faults a preset number of times, establishing a relay protection operating time sample set based on the operating time of the relay protection device after the intra-zone fault occurs; calculating the relay protection operating time center using a preset machine learning algorithm, stopping the calculation when the machine learning algorithm converges, and obtaining the operating time center element and the corresponding operating time center subset; setting the representative probability density of the operating time center element; and calculating the evaluation index of the relay protection device's operating time based on the operating time center element and the representative probability density. This invention provides a method for evaluating the operating time of relay protection devices based on machine learning algorithms. By establishing a unified standard dynamic simulation system for relay protection devices, various types of faults within the zone are simulated. The Affinity Propagation (AP) machine learning algorithm is used to obtain the time center and corresponding aggregate subset of the operating time sample set, and then the operating time evaluation index of the relay protection device is calculated. This enables a comprehensive evaluation of the operating time of relay protection devices under unified test conditions, different operating conditions, and various fault states. Furthermore, it allows for horizontal comparison of the operating time index of relay protection devices from different manufacturers, effectively improving the level of refined evaluation and management of relay protection devices. Attached Figure Description

[0086] Exemplary embodiments of the present invention can be more fully understood by referring to the following figures:

[0087] Figure 1 A flowchart of a relay protection device action time evaluation method based on machine learning algorithm according to a preferred embodiment of the present invention;

[0088] Figure 2 A flowchart of a relay protection device action time evaluation method based on machine learning algorithm according to a preferred embodiment of the present invention;

[0089] Figure 3 A schematic diagram of the primary system wiring for dynamic simulation of a transmission line according to a preferred embodiment of the present invention; and

[0090] Figure 4 This is a structural diagram of a relay protection device operation time evaluation system based on a machine learning algorithm according to a preferred embodiment of the present invention. Detailed Implementation

[0091] Exemplary embodiments of the invention will now be described with reference to the accompanying drawings. However, the invention may be embodied in many different forms and is not limited to the embodiments described herein. These embodiments are provided to fully and completely disclose the invention and to fully convey its scope to those skilled in the art. The terminology used in the exemplary embodiments illustrated in the drawings is not intended to limit the invention. In the drawings, the same units / elements are referred to by the same reference numerals.

[0092] Unless otherwise stated, the terms used herein (including technical terms) have their common meaning as understood by one of ordinary skill in the art. Furthermore, it is understood that terms defined in commonly used dictionaries should be understood to have a meaning consistent with the context of their relevant field, and not to be interpreted as having an idealized or overly formal meaning.

[0093] Figure 1 This is a flowchart illustrating a preferred embodiment of a relay protection device operation time evaluation method based on machine learning algorithms according to the present invention. The present invention provides a relay protection device operation time evaluation method based on machine learning algorithms. By establishing a unified standard relay protection device dynamic simulation system, various types of intra-zone faults are simulated. The Affinity Propagation (AP) machine learning algorithm is used to obtain the time center and corresponding aggregate subset of the operation time sample set, and then the relay protection device operation time evaluation index is calculated. This achieves a comprehensive evaluation of the relay protection device operation time under unified test conditions, different operating conditions, and various fault states. Furthermore, it allows for horizontal comparison of the operation time indexes of relay protection devices from different manufacturers, effectively improving the level of refined evaluation and management of relay protection devices.

[0094] like Figure 1 As shown, this invention provides a method for evaluating the operating time of a relay protection device based on a machine learning algorithm. The method includes:

[0095] Step 101: Simulate different operating conditions of the power system, traverse various types of intra-area faults, and establish a relay protection action time sample set based on the action time of the relay protection device after the intra-area fault occurs.

[0096] Step 102: Calculate the relay protection action time center using a preset machine learning algorithm. When the machine learning algorithm converges, stop the calculation and obtain the action time center element and the corresponding action time center subset of the action time.

[0097] Step 103: Set the representative probability density of the action time center elements based on the action time center subset;

[0098] Step 104: Calculate the evaluation index of the relay protection device's operating time based on the central element of the operating time and the representative probability density.

[0099] Preferably, before simulating different operating conditions of the power system and traversing various types of intra-regional faults, the process also includes:

[0100] Establish a relay protection device operation simulation test system. The test system includes a primary power system simulation unit, a fault simulation unit, a control system unit, a fault recording unit, and an external time synchronization unit.

[0101] The wiring configuration of the primary power system simulation unit is consistent with the typical wiring scheme of the prototype lines and equipment components protected by the relay protection device.

[0102] The fault simulation unit is used to simulate various metallic and transition resistance fault types, including single-phase ground fault, two-phase short-circuit ground fault, two-phase short-circuit fault, three-phase short-circuit fault, and three-phase short-circuit ground fault.

[0103] The control system unit is used to change the operating mode of the primary power system simulation unit and the fault types simulated by the fault simulation unit;

[0104] The fault recording unit is used to record the electrical waveforms of the primary power system simulation unit and the operating time of the relay protection device;

[0105] The external time synchronization unit is used to provide a unified external time reference for the relay protection device operation simulation test system.

[0106] Preferably, it further includes:

[0107] Connect the relay protection device to the relay protection device operation simulation test system;

[0108] Configure the relay protection device, including setting configuration and model file configuration;

[0109] The relay protection device operation simulation test system is debugged, including: analog quantity comparison, switch quantity verification, transmission test and communication check.

[0110] Preferably, the different operating conditions of the power system include: operating conditions with different system equivalent capacity and different loads, system frequency deviation, and operating conditions with manually operated faulty components;

[0111] The faults within the zone include: single-phase ground fault, two-phase short-circuit ground fault, two-phase short-circuit fault, three-phase short-circuit fault, and three-phase short-circuit ground fault.

[0112] Preferably, the representative probability density of the central element of the action time is:

[0113] f i =crad(C i ) / crad(T)

[0114] Among them, crad(C i ) represents the number of samples in the central subset of the operating time, and crad(T) represents the number of samples in the sample set of the relay protection operating time.

[0115] Preferably, the preset number of times to traverse various types of faults within the area is increased.

[0116] Preferably, the relay protection action time center is calculated using a preset machine learning algorithm, including:

[0117] Calculate the similarity of the relay protection action time sample sets:

[0118]

[0119] Where s(i,j)=-||t i -t j || 2 ;t i Let t be the action time recorded in the i-th trial. j Let s(i,j) be the action time recorded in the j-th trial, and let s(i,j) be the similarity between action time samples i and j.

[0120] The machine learning algorithm is initialized based on the similarity of the relay protection action time sample set:

[0121] S1: Initialize the attraction matrix R and the belonging matrix A to zero matrices;

[0122] S2: Update the attraction matrix R:

[0123]

[0124] Among them, a i (i, j) represents the attraction between action time samples i and j, r i (i, j) represents the degree of belonging between action time samples i and j, and k represents action time sample k;

[0125] S3: Update the affiliation matrix A:

[0126]

[0127] m is the number of iterations;

[0128] S4: Attenuate the attraction matrix R and the membership matrix A based on the attenuation coefficient λ:

[0129] r m+1 (i, j) = λ × r m (i,j)+(1-λ)×r m+1 (i, j)

[0130] a m+1 (i, j) = λ × a m (i,j)+(1-λ)×a m+1 (i, j)

[0131] Repeat steps S2, S3 and S4 until the attraction matrix R and the membership matrix A no longer change in the preset number of iterations or the maximum number of iterations is reached, then end the calculation.

[0132] The element with the largest a+r is selected as the center element of the action time, and the selection rule is: max{a(i,j)+r(i,j)}

[0133] Where a+r is the sum of attraction and belonging.

[0134] like Figure 2 As shown, this invention proposes a method for evaluating the operating time of relay protection devices based on the Affinity Propagation machine learning algorithm. The method includes:

[0135] S1: Construct a dynamic model test system for relay protection devices. The test system includes a primary power system simulation unit, a fault simulation unit, a control system unit, a fault recording unit, and an external time synchronization unit. In step S1 of this invention, the wiring configuration of the primary power system simulation unit is consistent with the typical wiring scheme of the prototype line and equipment components protected by the relay protection device.

[0136] The fault simulation unit can simulate various metallic and transition resistance fault types, including single-phase ground fault, two-phase short-circuit ground fault, two-phase short-circuit fault, three-phase short-circuit fault, and three-phase short-circuit ground fault.

[0137] The control system unit can change the operating mode of the primary power system simulation unit and the type of fault simulated by the fault simulation unit;

[0138] The fault recording unit can record the electrical waveforms of a primary power system simulation unit and the operating time of relay protection devices;

[0139] The external time synchronization unit provides a unified external time reference for the dynamic model testing system.

[0140] S2: Connect the relay protection device to the dynamic model test system to complete the device configuration and test system debugging; In step S2 of this invention, the device configuration includes setting value adjustment and model file configuration; the test system debugging includes analog quantity comparison, switch quantity verification, transmission test and communication check.

[0141] S3: Simulate various types of faults within the power system under different operating conditions, record the operating time of the relay protection device after each fault, and obtain the protection operating time sample set T = {t1, t2, ... t}. n};

[0142] In step S3 of this invention, different operating conditions of the power system include different system equivalent capacity and different load operating conditions. Optionally, different operating conditions of the power system may also include system frequency deviation and manual fault component operating conditions.

[0143] Various types of faults within the zone include single-phase ground faults, two-phase short-circuit ground faults, two-phase short-circuit faults, three-phase short-circuit faults, and three-phase short-circuit ground faults.

[0144] The protection action time t is calculated as the relative time between the fault occurrence and the receipt of the protection action signal by the fault recording unit. To improve the accuracy of the action time evaluation index, this invention can repeat the fault simulation traversal in step S3 to increase the number of protection action time samples T.

[0145] S4: The Affinity Propagation (AP) machine learning algorithm is used to analyze the action time sample set T, calculate the similarity S between samples, set the algorithm parameters, and complete the convergence of the AP algorithm to obtain the time center subset E = {e1, e2, ..., e} of the action time T. k}, and the corresponding time-aggregated sample set C = {C1, C2, ..., C} k In step S4 of this invention, the similarity between samples is calculated. Where s(i,j)=-||t i -t j || 2 .

[0146] The algorithm parameters include the preference parameter p, the damping coefficient λ, the maximum number of iterations n, and the number of iterations g that does not change in the estimated number of clusters at which convergence stops;

[0147] The initialization of the Affinity Propagation (AP) machine learning algorithm of this invention includes:

[0148] Step 4-1, Initialization: Initialize the attraction matrix R and the belonging matrix A to zero matrices;

[0149] Step 4-2, Update the attraction matrix:

[0150] Step 4-3, Update the affiliation matrix:

[0151]

[0152] Step 4-4: Attenuate the two formulas according to the attenuation coefficient λ:

[0153] r m+1 (i, j) = λ × r m (i,j)+(1-λ)×r m+1 (i, j)

[0154] a m+1 (i, j) = λ × a m (i,j)+(1-λ)×a m+1 (i, j);

[0155] Steps 4-5 are repeated, including steps 4-2, 4-3, and 4-4, until the attraction matrix R and the belonging matrix A no longer change in the g iterations or the maximum number of iterations n is reached. The calculation ends then.

[0156] Steps 4-6: Finally, select the t with the largest a+r. k As t i The cluster center is selected according to the rule: max{a(i,j)+r(i,j)}.

[0157] S5: Set the motion time center e i The representative probability density f of (i = 1, 2, ..., k) i In step S5 of this invention, the representative probability density f of the action time center is... i =crad(C i ) / crad(T).

[0158] S6: Calculate the evaluation index of relay protection device operating time

[0159] This invention provides a method for evaluating the operating time of relay protection devices based on the Affinity Propagation (AP) machine learning algorithm. The method includes: constructing a dynamic simulation test system for the relay protection device, comprising a primary power system simulation unit, a fault simulation unit, a control system unit, a fault recording unit, and an external time synchronization unit; connecting the relay protection device to the dynamic simulation test system and completing device configuration and test system debugging; simulating various types of intra-area faults under different operating conditions of the power system, recording the operating time of the relay protection device after each fault occurs, and obtaining a protection operating time sample set T; using the Affinity Propagation (AP) machine learning algorithm to obtain the time center E of the operating time T and the corresponding time aggregation sample set C; and setting the operating time center e. i The representative probability density f i The operating time evaluation index p of the relay protection device is calculated.

[0160] This invention establishes a standard dynamic simulation test system that can connect relay protection devices of the same type from different manufacturers and using different algorithms to the test system. Under the condition that the external test conditions are completely consistent, simulations of all fault types under various system operating conditions are carried out to obtain a series of protection device action times. By using the AP machine learning algorithm to aggregate and classify the action time samples, the action time center of the relay protection device and the corresponding representative probability density are obtained, and then the action time evaluation index is obtained to achieve a comprehensive evaluation of the action time of the relay protection device.

[0161] The relay protection device operation time evaluation method provided by this invention uses machine learning algorithms to automatically aggregate and classify a large number of operation times, reducing the preprocessing steps of operation time sample data, and has strong robustness and high accuracy in time center calculation.

[0162] The relay protection device operation time evaluation method provided by this invention realizes the comprehensive evaluation of the operation time performance of relay protection devices under the unified standard dynamic simulation test conditions, as well as the horizontal comparison of the performance of devices from different manufacturers, thus solving the problem of comprehensive evaluation of the performance of relay protection devices.

[0163] Taking the evaluation of line protection action time as an example, this invention builds a dynamic simulation system for a 500kV 400km double-circuit non-inductor transmission line based on the requirements of GB / T 26864-2011 "Dynamic Model Test of Power System Relay Protection Products" and the functions of complete sets of protection devices for ultra-high voltage transmission lines.

[0164] See primary system wiring diagram Figure 3As shown, plant N is connected to system L via a 500kV double-circuit non-inductor transmission line. Plant N has two generator units, 12G and 14G, with a total installed capacity of 5880MW. Substation L is connected to a regional equivalent system and has three operating modes: large, medium, and small, with corresponding short-circuit capacities of 20000MVA, 10000MVA, and 3000MVA, respectively.

[0165] The main parameters of each generator set are shown in Table 1.

[0166] Table 1 Main parameters of generator set

[0167]

[0168] The transmission line is 400 km long, and its main parameters are shown in Table 2. Plant N uses a single busbar connection, and station L uses a one-and-a-half switch connection. Each transmission line is equipped with a shunt reactor with a capacity of 150 MVar at both ends, and the reactor parameters are shown in Table 3.

[0169] Table 2. Main parameters of transmission lines (parameters per 100km)

[0170]

[0171] Table 3 Main parameters of high-voltage shunt reactors

[0172]

[0173]

[0174] Line L2 is the protected line. During the test, a total of 6 fault points were set at both ends and the middle of the protected line, both ends of the unprotected line, and the bus outside the zone. They are numbered FD1, FD2, FD3, FD6, FD9 and FD10 respectively. Each fault point can simulate various types of metallic or short-circuit faults through transition resistance.

[0175] The AC voltage for the N station line protection device is connected to a line-side capacitive voltage transformer (CVT) PT73 with a transformation ratio of 500kV / 0.1kV, and the analog AC current is connected to a current transformer CT78 with a transformation ratio of 1250A / 1. The AC voltage for the L station line protection device is connected to a line-side capacitive voltage transformer (CVT) PT15 with a transformation ratio of 500kV / 0.1kV, and the AC current is connected to current transformers CT16 / CT9 with a transformation ratio of 1250A / 1. The protection device is connected to circuit breaker position and failure input signals.

[0176] The line protection devices on both sides adopt a communication method of direct fiber optic connection + 2M multiplexed channel.

[0177] The test used a fault recorder to record the voltage and current signals of the input protection device, as well as the protection's phase trip, reclosing, and signal transmission signals. A GPS or Beidou time synchronization server was used to provide a unified clock synchronization signal for the relay protection device, fault recorder, and control system.

[0178] The test control system can perform generator start-up and shutdown, grid connection control and power regulation, switch opening and closing, and fault sequence simulation.

[0179] After connecting the line protection device to the dynamic simulation system, the device setting and configuration file download are completed according to the primary system equipment parameters and the communication connection relationship of the protection device. After verification, the test system is debugged. The debugging content mainly includes analog quantity comparison, switch quantity verification, transmission test and communication status check.

[0180] Under different short-circuit capacities, the system simulates heavy-load and light-load operation modes of the line. Instantaneous metallic faults and faults via transition resistance (single-phase grounding, two-phase short-circuit grounding, two-phase short-circuit, three-phase short-circuit, and three-phase short-circuit grounding) are tested within the line protection zone. The protection action time for each fault is recorded using a fault recorder, resulting in a sample set T of protection action times for similar lines from manufacturers A and B. A and T B In this embodiment, the number of action time samples for each protection is 106, and the action behavior and action time of the protection under various fault conditions meet the standard requirements.

[0181] In the Affinity Propagation (AP) machine learning algorithm, the preference parameter p = 50, the damping coefficient λ = 0.5, the maximum number of iterations n = 10000, and the number of iterations g = 15 at which the estimated number of clusters stops converging are unchanged.

[0182] The operating time T of the line protection of manufacturer A was obtained through iterative calculation. A Time Center E A = {11.25, 23.25, 30.25, 15, 12.75, 35.5} (unit: ms), and the corresponding time-aggregated sample set C. A ={C A1 C A2 C A3 C A4 C A5 C A6 Manufacturer B's line protection action time T B Time Center E B = {16, 20.5, 13.75, 78.5, 27.5, 56.75, 99.75} (unit: ms), and the corresponding time-aggregated sample set C. B ={CB1 C B2 C B3 C B4 C B5 C B6 C B7};

[0183] Based on the representative probability density f of the action time center i =crad(C i ) / crad(T), to get

[0184] f A1 =0.3208, f A2 =0.0189, f A3 =0.0472, f A4 =0.1415, f A5 =0.3679, f A6 =0.1038;

[0185] f B1 =0.4434, f B2 =0.0755, f B3 =0.3396, f B4 =0.0283, f B5 =0.0566, f B6 =0.0283, f B7 =0.0283;

[0186] The operating time evaluation index of the line protection device of manufacturer A was calculated. Evaluation index of the operating time of the line protection device of manufacturer B

[0187] The results show that, under the unified constraints of current standards and specifications, and under standardized dynamic model test conditions, the line protection device of Manufacturer A has a better operating time than the same type of line protection device of Manufacturer B.

[0188] Figure 4 This is a structural diagram of a relay protection device operation time evaluation system based on a machine learning algorithm according to a preferred embodiment of the present invention. Figure 4 As shown, this invention provides a relay protection device operation time evaluation system based on machine learning algorithms. The system includes:

[0189] Initial unit 401 is used to simulate different operating conditions of the power system, traverse various types of faults in the zone a preset number of times, and establish a sample set of relay protection action time based on the action time of the relay protection device after the occurrence of the fault in the zone.

[0190] The acquisition unit 402 is used to calculate the relay protection action time center through a preset machine learning algorithm. When the machine learning algorithm converges, the calculation stops and the action time center element and the corresponding action time center subset are acquired.

[0191] Setting unit 403 is used to set the representative probability density of action time center elements based on the action time center subset;

[0192] Result unit 404 is used to calculate the evaluation index of the operating time of the relay protection device based on the central element of the operating time and the representative probability density.

[0193] Preferably, the system further includes a setup unit for:

[0194] Establish a relay protection device operation simulation test system. The test system includes a primary power system simulation unit, a fault simulation unit, a control system unit, a fault recording unit, and an external time synchronization unit.

[0195] The wiring configuration of the primary power system simulation unit is consistent with the typical wiring scheme of the prototype lines and equipment components protected by the relay protection device.

[0196] The fault simulation unit is used to simulate various metallic and transition resistance fault types, including single-phase ground fault, two-phase short-circuit ground fault, two-phase short-circuit fault, three-phase short-circuit fault, and three-phase short-circuit ground fault.

[0197] The control system unit is used to change the operating mode of the primary power system simulation unit and the fault types simulated by the fault simulation unit;

[0198] The fault recording unit is used to record the electrical waveforms of the primary power system simulation unit and the operating time of the relay protection device;

[0199] The external time synchronization unit is used to provide a unified external time reference for the relay protection device operation simulation test system.

[0200] Preferably, the establishment unit is also used for:

[0201] Connect the relay protection device to the relay protection device operation simulation test system;

[0202] Configure the relay protection device, including setting configuration and model file configuration;

[0203] The relay protection device operation simulation test system is debugged, including: analog quantity comparison, switch quantity verification, transmission test and communication check.

[0204] Preferably, the different operating conditions of the power system include: operating conditions with different system equivalent capacity and different loads, system frequency deviation, and operating conditions with manually operated faulty components;

[0205] The faults within the zone include: single-phase ground fault, two-phase short-circuit ground fault, two-phase short-circuit fault, three-phase short-circuit fault, and three-phase short-circuit ground fault.

[0206] Preferably, the representative probability density of the central element of the action time is:

[0207] f i =crad(C i ) / crad(T)

[0208] Among them, crad(C i ) represents the number of samples in the central subset of the operating time, and crad(T) represents the number of samples in the sample set of the relay protection operating time.

[0209] Preferably, the initial unit is also used to: increase the preset number of times to traverse various types of faults within the area.

[0210] Preferably, the acquisition unit is used to calculate the relay protection action time center through a preset machine learning algorithm, including:

[0211] Calculate the similarity of the relay protection action time sample sets:

[0212]

[0213] Where s(i,j)=-||t i -t j || 2 ;t i Let t be the action time recorded in the i-th trial. j Let s(i,j) be the action time recorded in the j-th trial, and let s(i,j) be the similarity between action time samples i and j.

[0214] The machine learning algorithm is initialized based on the similarity of the relay protection action time sample set:

[0215] S1: Initialize the attraction matrix R and the belonging matrix A to zero matrices;

[0216] S2: Update the attraction matrix R:

[0217]

[0218] Among them, a i (i,j) represents the attraction between action time samples i and j, r i (i,j) represents the degree of belonging between action time samples i and j, and k represents action time sample k;

[0219] S3: Update the affiliation matrix A:

[0220]

[0221] m is the number of iterations;

[0222] S4: Attenuate the attraction matrix R and the membership matrix A based on the attenuation coefficient λ:

[0223] r m+1 (i, j) = λ × r m (i,j)+(1-λ)×r m+1 (i, j)

[0224] a m+1 (i, j) = λ × a m (i,j)+(1-λ)×a m+1 (i, j)

[0225] Repeat steps S2, S3 and S4 until the attraction matrix R and the membership matrix A no longer change in the preset number of iterations or the maximum number of iterations is reached, then end the calculation.

[0226] The element with the largest a+r is selected as the center element of the action time, and the selection rule is: max{a(i,j)+r(i,j)}

[0227] Where a+r is the sum of attraction and belonging.

[0228] The preferred embodiment of the present invention provides a relay protection device operation time evaluation system based on machine learning algorithm, which corresponds to the preferred embodiment of the present invention provides a relay protection device operation time evaluation method based on machine learning algorithm, and will not be described in detail here.

[0229] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of the present invention can be implemented using various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0230] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0231] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0232] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0233] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0234] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

[0235] The invention has been described with reference to a few embodiments. However, as will be known to those skilled in the art, and as defined in the appended claims, other embodiments besides those disclosed above fall equivalently within the scope of the invention.

[0236] Generally, all terms used in the claims are to be interpreted according to their ordinary meaning in the art, unless otherwise expressly defined herein. All references to “a / the / the [device, component, etc.]” ​​are openly interpreted as at least one instance of said device, component, etc., unless otherwise expressly stated. The steps of any method disclosed herein need not be performed in the exact order disclosed unless explicitly stated otherwise.

Claims

1. A method for evaluating the operating time of a relay protection device based on a machine learning algorithm, the method comprising: Simulate different operating conditions of the power system, traverse various types of faults within the zone according to a preset number of times, and establish a sample set of relay protection action time based on the action time of the relay protection device after the occurrence of the fault within the zone. The relay protection action time center is calculated by a preset machine learning algorithm. When the machine learning algorithm converges, the calculation stops and the action time center element and the corresponding action time center subset are obtained. The calculation of the relay protection action time center using a preset machine learning algorithm includes: Calculate the similarity of the relay protection operation time sample set: in ; ti Let i be the action time recorded in the i-th trial. tj Let j be the action time recorded in the j-th experiment. s(i,j) The similarity between action time samples i and j; The machine learning algorithm is initialized based on the similarity of the relay protection action time sample set: S1: Initialize the attraction matrix R and the belonging matrix A to zero matrices; S2: Update the attraction matrix R: in, a i (i,j) The attraction between action time samples i and j. r i (i,j) The degree of belonging between action time samples i and j. k Let k be the action time sample; S3: Update the affiliation matrix A: m This represents the number of iterations. S4: Attenuate the attraction matrix R and the membership matrix A based on the attenuation coefficient λ: Repeat steps S2, S3 and S4 until the attraction matrix R and the membership matrix A no longer change in a preset number of iterations or the maximum number of iterations is reached, then end the calculation. Select a + r The largest element is selected as the center element of the action time, and the selection rule is as follows: in, a + r The sum of attraction and belonging. A representative probability density of the action time center elements is set based on the subset of action time centers; the representative probability density of the action time center elements is: f i =crad(C i ) / crad(T) in, crad(C) i ） Number of samples in the central subset of action time ,crad(T) This represents the number of samples in the relay protection operation time sample set. The evaluation index for the operation time of the relay protection device is calculated based on the central element of the operation time and the representative probability density.

2. The method according to claim 1, further comprising, before simulating different operating conditions of the power system and traversing various types of intra-regional faults: A relay protection device operation simulation test system is established, which includes a primary power system simulation unit, a fault simulation unit, a control system unit, a fault recording unit, and an external time synchronization unit. The wiring configuration of the primary power system simulation unit is consistent with the typical wiring scheme of the prototype line and equipment components protected by the relay protection device. The fault simulation unit simulates various metallic and transition resistance fault types, including single-phase ground fault, two-phase short-circuit ground fault, two-phase short-circuit fault, three-phase short-circuit fault, and three-phase short-circuit ground fault. The control system unit changes the operating mode of the primary power system simulation unit and the fault type simulated by the fault simulation unit. The fault recording unit records the electrical waveforms of a primary power system simulation unit and the operating time of the relay protection device. The external time synchronization unit provides a unified external time reference for the relay protection device operation simulation test system.

3. The method according to claim 2, further comprising: Connect the relay protection device to the relay protection device operation simulation test system; The relay protection device is configured, including setting configuration and model file configuration; The relay protection device operation simulation test system is debugged, including: analog quantity comparison, switch quantity verification, transmission test and communication check.

4. The method according to claim 1, wherein the different operating conditions of the power system include: Operating conditions under different system equivalent capacities and different loads, system frequency deviation, and operating conditions of manually operated faulty components; The faults within the zone include: single-phase ground fault, two-phase short-circuit ground fault, two-phase short-circuit fault, three-phase short-circuit fault, and three-phase short-circuit ground fault.

5. The method according to claim 1, wherein the preset number of times to traverse various types of faults within the area is increased.

6. A relay protection device operation time evaluation system based on machine learning algorithms, the system comprising: The initial unit is used to simulate different operating conditions of the power system, traverse various types of faults in the zone a preset number of times, and establish a sample set of relay protection action times based on the action time of the relay protection device after the occurrence of the fault in the zone. The acquisition unit is used to calculate the relay protection action time center through a preset machine learning algorithm. When the machine learning algorithm converges, the calculation is stopped, and the action time center element and the corresponding action time center subset of the action time are acquired. The acquisition unit is used to calculate the relay protection action time center through a preset machine learning algorithm, including: Calculate the similarity of the relay protection operation time sample set: in ; ti Let i be the action time recorded in the i-th trial. tj Let j be the action time recorded in the j-th experiment. s(i,j) The similarity between action time samples i and j; The machine learning algorithm is initialized based on the similarity of the relay protection action time sample set: S1: Initialize the attraction matrix R and the belonging matrix A to zero matrices; S2: Update the attraction matrix R: in, a i (i,j) The attraction between action time samples i and j. r i (i,j) The degree of belonging between action time samples i and j. k Let k be the action time sample; S3: Update the affiliation matrix A: m This represents the number of iterations. S4: Attenuate the attraction matrix R and the membership matrix A based on the attenuation coefficient λ: Repeat steps S2, S3 and S4 until the attraction matrix R and the membership matrix A no longer change in a preset number of iterations or the maximum number of iterations is reached, then end the calculation. Select a + r The largest element is selected as the center element of the action time, and the selection rule is as follows: in, a + r The sum of attraction and belonging The setting unit is used to set the representative probability density of the action time center elements based on the subset of action time centers, wherein the representative probability density of the action time center elements is: f i =crad(C i ) / crad(T) in, crad(C) i ） Number of samples in the central subset of action time ,crad(T) This represents the number of samples in the relay protection action time sample set. The result unit is used to calculate the evaluation index of the relay protection device's operating time based on the central element of the operating time and the representative probability density.

7. The system according to claim 6, further comprising an establishment unit, configured to: A relay protection device operation simulation test system is established, which includes a primary power system simulation unit, a fault simulation unit, a control system unit, a fault recording unit, and an external time synchronization unit. The wiring configuration of the primary power system simulation unit is consistent with the typical wiring scheme of the prototype line and equipment components protected by the relay protection device. The fault simulation unit is used to simulate various metallic and transition resistance fault types, including single-phase ground fault, two-phase short-circuit ground fault, two-phase short-circuit fault, three-phase short-circuit fault, and three-phase short-circuit ground fault. The control system unit is used to change the operating mode of the primary power system simulation unit and the fault types simulated by the fault simulation unit; The fault recording unit is used to record the electrical waveforms of the primary power system simulation unit and the operating time of the relay protection device. The external time synchronization unit is used to provide a unified external time reference for the relay protection device operation simulation test system.

8. The system according to claim 7, wherein the establishing unit is further configured to: Connect the relay protection device to the relay protection device operation simulation test system; The relay protection device is configured, including setting configuration and model file configuration; The relay protection device operation simulation test system is debugged, including: analog quantity comparison, switch quantity verification, transmission test and communication check.

9. The system according to claim 6, wherein the different operating conditions of the power system include: Operating conditions under different system equivalent capacities and different loads, system frequency deviation, and operating conditions of manually operated faulty components; The faults within the zone include: single-phase ground fault, two-phase short-circuit ground fault, two-phase short-circuit fault, three-phase short-circuit fault, and three-phase short-circuit ground fault.

10. The system according to claim 6, wherein the initial unit is further configured to: increase the preset number of times to traverse various types of faults within the area.

Images