A fault detection method and system for primary and secondary fusion equipment

Abstract

The present invention discloses a fault detection method and system for primary and secondary fusion equipment, which simulates fault scenarios on the circuit simulation model of the primary and secondary fusion equipment, obtains fault data under the corresponding fault scenarios, and then uses the fault data of different scenarios to train different convolution models to obtain different target convolution models. After assigning corresponding weights to the real fault data after a fault occurs during the use of the primary and secondary fusion equipment, the real fault data is input into different target convolution models for fault detection, and the detection result with the highest weighted accuracy is taken as the final primary and secondary fusion equipment fault detection result, thereby realizing automatic detection of faults in the primary and secondary fusion equipment. This solves the technical problem that the existing primary and secondary fusion equipment needs maintenance personnel to arrive at the site to perform fault maintenance after a fault occurs during use, which is inefficient and cannot meet the requirements of rapid detection, thus affecting the efficiency of line maintenance.

Description

Technical Field

[0001] The present invention relates to the technical field of power system fault detection, and in particular to a fault detection method and system for primary and secondary fusion equipment. Background Art

[0002] With the rapid development of new energy technologies, power distribution technologies, power efficiency technologies, and information and communication technologies, the distribution network has evolved from a simple power network to an intelligent energy and information integration network.

[0003] Primary-secondary convergence refers to the integrated design concept of primary and secondary equipment. This includes standardized primary equipment, miniaturized, standardized, and plug-and-play terminal product designs, and standardized primary and secondary device interfaces. This allows for a high degree of integration to meet requirements for segmented line loss management, local feeder automation, single-phase ground fault detection, and automated testing. Primary-secondary convergence equipment refers to the integration of intelligent secondary units within the primary equipment of a power system, making the primary equipment more intelligent and incorporating built-in measurement, metering, relay protection, detection, and control functions to better meet power demand.

[0004] During the use of current primary and secondary fusion equipment, maintenance personnel arrive at the site to carry out fault repairs only after a fault occurs. This is inefficient and cannot meet the requirements of rapid detection, thus affecting the efficiency of line maintenance. Summary of the Invention

[0005] The present invention provides a fault detection method and system for primary and secondary fusion equipment, which is used to solve the technical problem that during the use of existing primary and secondary fusion equipment, after a fault occurs, maintenance personnel need to arrive at the site to carry out fault repair, which is inefficient and cannot meet the requirements of rapid detection, thus affecting the efficiency of line maintenance.

[0006] In view of this, a first aspect of the present invention provides a fault detection method for primary and secondary fusion equipment, comprising:

[0007] Conduct simulations of different fault scenarios on the established circuit simulation model of primary and secondary fusion equipment to obtain fault data under each fault scenario, where the fault data includes fault factor data and fault results;

[0008] Divide the fault data under each fault scenario into training data sets, where one training data set corresponds to one fault scenario;

[0009] After assigning different weights to the fault factor data in the training dataset, the data is input into the corresponding initial convolution model for training to obtain the trained target convolution model.

[0010] Obtain real fault data after a failure of the primary and secondary fusion equipment, assign different weights to each real fault data, and input it into all target convolution models to obtain the detection results output by each target convolution model, as well as the accuracy and weight value corresponding to the detection results;

[0011] The weighted accuracy of each detection result is calculated based on the accuracy and weight value corresponding to each detection result, and the detection result with the highest weighted accuracy is used as the final primary and secondary fusion equipment fault detection result.

[0012] Optionally, the fault scenarios include a neutral point grounding scenario and a neutral point ungrounded scenario, the training data set includes a first training data set and a second training data set, the initial convolution model includes a first initial convolution model and a second initial convolution model, and the target convolution model includes a first target convolution model and a second target convolution model.

[0013] Optionally, the circuit simulation model includes at least five feeders in both the neutral point grounding scenario and the neutral point ungrounded scenario, and the five feeders are an overhead feeder, a first hybrid feeder, a second hybrid feeder, a cable feeder and a branch feeder.

[0014] Optionally, the fault factor data corresponding to the neutral point ungrounded scenario include: fault initial angle, fault duration, fault transition resistance value and system capacity; the fault factor data corresponding to the neutral point grounded scenario include: arc suppression coil compensation degree, fault initial angle, fault duration, fault transition resistance value and system capacity.

[0015] Optionally, the fault result includes abnormal equipment working state, abnormal current and voltage values, overhead fault, mixed fault, branch fault or cable fault.

[0016] Optionally, obtain real fault data after a fault occurs in the primary and secondary fusion equipment, assign different weights to each real fault data, and input it into all target convolution models to obtain the detection results output by each target convolution model and the accuracy and weight values ​​corresponding to the detection results, including:

[0017] Obtaining real fault data at the first moment and the second moment after a fault occurs on the primary and secondary fusion equipment, wherein the interval between the second moment and the first moment is at least 10 seconds;

[0018] After assigning different weights to each real fault data at the first moment and each real fault data at the second moment, they are input into all target convolution models respectively, and the first detection result and the second detection result output by each target convolution model corresponding to the first moment and the second moment are obtained, as well as the accuracy and weight values ​​corresponding to the first detection result and the second detection result respectively;

[0019] Accordingly, the weighted accuracy of each detection result is calculated based on the accuracy and weight value corresponding to each detection result, and the detection result with the highest weighted accuracy is used as the final primary and secondary fusion equipment fault detection result, including:

[0020] Determine whether a first detection result corresponding to the first moment and a second detection result corresponding to the second moment are the same;

[0021] If the first detection result corresponding to the first moment and the second detection result corresponding to the second moment are the same, then the weighted accuracy of each detection result corresponding to the first moment or the second moment is calculated based on the first detection result or the second detection result and the corresponding accuracy and weight value, and the detection result with the highest weighted accuracy is used as the final primary and secondary fusion equipment fault detection result;

[0022] If the first detection result corresponding to the first moment and the second detection result corresponding to the second moment are different, the real fault data between the first moment and the second moment are obtained, and different weights are assigned to each real fault data between the first moment and the second moment. The data are input into all target convolution models to obtain the detection results output by each target convolution model and the accuracy and weight values ​​corresponding to the detection results. The weighted accuracy of each detection result is calculated according to the accuracy and weight values ​​corresponding to each detection result, and the detection result with the highest weighted accuracy is used as the final primary and secondary fusion equipment fault detection result.

[0023] Optionally, in a neutral point grounding scenario, the circuit from the neutral point to the ground includes a Z-type transformer, a resistor, and an inductor;

[0024] One end of the Z-type transformer is connected to the neutral point, the other end is connected to one end of the resistor, the other end of the resistor is connected to one end of the inductor, and the other end of the inductor is grounded;

[0025] The relationship between inductance and resistance is:

[0026] L=1 / (1+p)×1 / (3R 2 C)

[0027] Wherein, L is the inductance value, p is the compensation degree of the arc suppression coil composed of resistance and inductance, C is the distributed capacitance value of the circuit simulation model to the ground, and R is the resistance value.

[0028] Optionally, the arc suppression coil compensation degree is 5%.

[0029] A second aspect of the present invention provides a fault detection system for primary and secondary fusion equipment, comprising:

[0030] The fault simulation module is used to simulate different fault scenarios for the established circuit simulation model of the primary and secondary fusion equipment and obtain fault data under each fault scenario, where the fault data includes fault factor data and fault results;

[0031] A training data partitioning module is used to divide the fault data under each fault scenario into training data sets, where one training data set corresponds to one fault scenario;

[0032] The model training module is used to assign different weights to the fault factor data in the training data set and input them into the corresponding initial convolution model for training to obtain a trained target convolution model;

[0033] The real fault data processing module is used to obtain real fault data after a fault occurs in the primary and secondary fusion equipment. After assigning different weights to each real fault data, it is input into all target convolution models to obtain the detection results output by each target convolution model and the accuracy and weight value corresponding to the detection results;

[0034] The real fault result output module is used to calculate the weighted accuracy of each detection result based on the accuracy and weight value corresponding to each detection result, and use the detection result with the highest weighted accuracy as the final primary and secondary fusion equipment fault detection result.

[0035] Optionally, the fault scenarios include a neutral point grounding scenario and a neutral point ungrounded scenario, the training data set includes a first training data set and a second training data set, the initial convolution model includes a first initial convolution model and a second initial convolution model, and the target convolution model includes a first target convolution model and a second target convolution model.

[0036] From the above technical solutions, it can be seen that the fault detection method and system for primary and secondary fusion equipment provided by the present invention have the following advantages:

[0037] The fault detection method and system for primary and secondary fusion equipment provided by the present invention perform fault scenario simulation on the circuit simulation model of the primary and secondary fusion equipment, obtain fault data under the corresponding fault scenario, and then use the fault data of different scenarios to train different convolution models to obtain different target convolution models. After assigning corresponding weights to the real fault data after a real fault occurs in the use of the primary and secondary fusion equipment, the data is input into different target convolution models for fault detection, and the detection result with the highest weighted accuracy is taken as the final primary and secondary fusion equipment fault detection result. This realizes automatic detection of faults in the primary and secondary fusion equipment, does not require maintenance personnel to go to the site for fault detection, improves detection efficiency, and can avoid the problem of human fault detection errors. It has high accuracy, and solves the technical problem that in the use of existing primary and secondary fusion equipment, after a fault occurs, maintenance personnel need to arrive at the site for fault repair, which is inefficient and cannot meet the requirements of rapid detection, affecting the efficiency of line maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying any creative work.

[0039] Figure 1 A schematic flow chart of a fault detection method for primary and secondary fusion equipment provided by the present invention;

[0040] Figure 2 This is a simulation diagram of the simulation model provided by the present invention in a scenario where the neutral point is not grounded;

[0041] Figure 3 This is a simulation diagram of the simulation model provided by the present invention in a neutral point grounding scenario;

[0042] Figure 4 A structural schematic diagram of a fault detection system for primary and secondary fusion equipment provided by the present invention. DETAILED DESCRIPTION

[0043] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0044] For easier understanding, see Figure 1 The present invention provides an embodiment of a fault detection method for primary and secondary fusion equipment, comprising:

[0045] Step 101: simulate different fault scenarios on the established circuit simulation model of the primary and secondary fusion equipment to obtain fault data under each fault scenario, wherein the fault data includes fault factor data and fault results.

[0046] It should be noted that, in the embodiment of the present invention, a circuit simulation model of the primary and secondary fusion equipment is first constructed. Considering that the fault conditions corresponding to the fault data are different in different scenarios, such as the neutral point grounding scenario and the neutral point ungrounding scenario, the circuit simulation model of the primary and secondary fusion equipment is simulated using simulation software for different fault scenarios to obtain fault data of different fault conditions in different scenarios. For example, in the neutral point ungrounding scenario, the fault factor data obtained include the fault initial angle, fault duration, fault transition resistance value and system capacity, and the corresponding fault results may include abnormal equipment working state, abnormal current and voltage values, overhead fault, mixed fault, branch fault or cable fault. In the neutral point grounding scenario, the fault factor data obtained include arc suppression coil compensation degree, fault initial angle, fault duration, fault transition resistance value and system capacity, and the corresponding fault results may include abnormal equipment working state, abnormal current and voltage values, overhead fault, mixed fault, branch fault or cable fault. Specifically, when simulating neutral grounding scenarios to obtain fault data, the simulation comprehensively considers multiple factors, including system capacity, line type, network topology, system neutral grounding method, ground fault type, number of ground fault points, fault initial phase angle, harmonic content, and fault transition resistance, to complete the fault simulation process and ensure that the fault simulation process can meet the various factors of actual faults, thereby improving the accuracy of subsequent model training. Specifically, the fault initial angle can be set to 0°, 30°, 60°, or 90°; the fault duration can be set to 10s, 1min, or 5min; the ground fault type can be set to single-phase grounding or arc grounding; the fault transition resistance can be set to 0, 100Ω, 200Ω, 400Ω, 600Ω, 800Ω, 1000Ω, 2000Ω, 3000Ω, 4000Ω, 5000Ω, or 6000Ω; the system capacity can be set to 10A or 5A; the fault phase can be set to AN, BN, or CN; and the fault location can be set to the beginning, middle, or end of the overhead line or cable line.

[0047] Step 102: Divide the fault data under each fault scenario into training data sets, wherein one training data set corresponds to one fault scenario.

[0048] It should be noted that the fault data under each fault scenario is divided into a training data set. For example, the neutral point grounding scenario and the neutral point ungrounded scenario, then there are two corresponding training data sets, that is, the neutral point ungrounded scenario corresponds to the first training data set, and the neutral point grounding scenario corresponds to the second training data set.

[0049] Step 103: After assigning different weights to the fault factor data in the training data set, the data is input into the corresponding initial convolution model for training to obtain a trained target convolution model.

[0050] It should be noted that different weights are assigned to different fault factor data based on their impact, allowing for differentiated training of the initial convolutional model based on the varying impacts of these data on the fault outcomes. After assigning different weights to different fault factor data, these data are fed into different initial convolutional models for training. The fault outcomes corresponding to the fault factor data are then compared with the output of the initial convolutional model to ensure that the initial convolutional model is more accurately trained on the fault data, thereby improving the accuracy of the model's detection results. Specifically, for the neutral point ungrounded scenario and the neutral point grounded scenario, the neutral point ungrounded scenario corresponds to the first training data set and the first initial convolution model respectively, and the neutral point grounded scenario corresponds to the second training data set and the second initial convolution model respectively. The fault factor data included in the first training data set are the fault initial angle, fault duration, fault transition resistance value and system capacity, and the fault factor data included in the second training set are the arc suppression coil compensation degree, fault initial angle, fault duration, fault transition resistance value and system capacity. Then, after assigning different weight values ​​to the fault factor data of the fault initial angle, fault duration, fault transition resistance value and system capacity, the data are input into the first initial convolution model for training. When the output accuracy of the first initial convolution model reaches the threshold, the training is stopped to obtain the trained first target convolution model. Similarly, after assigning different weight values ​​to the arc suppression coil compensation degree, fault initial angle, fault duration, fault transition resistance value and system capacity, the data are input into the second initial convolution model for training. When the output accuracy of the second initial convolution model reaches the threshold, the training is stopped to obtain the trained second target convolution model.

[0051] Step 104: Obtain real fault data after a failure of the primary and secondary fusion equipment. After assigning different weights to each real fault data, input it into all target convolution models to obtain the detection results output by each target convolution model and the accuracy and weight value corresponding to the detection results.

[0052] It should be noted that the real fault data after a failure of the primary and secondary fusion equipment is the data after a failure of the primary and secondary fusion equipment to be detected occurs during the actual working process, and the data type is the same as the fault factor data type of the first training data set and the second training data set. After assigning different weights to each real fault data (the weight value can be the same as the weight value in the training process, or it can be adjusted manually or automatically according to actual needs), it is input into all target convolution models to obtain the detection results output by the target convolution model, as well as the accuracy and weight values ​​corresponding to the detection results. For example, as mentioned above, all target convolution models include the first target convolution model and the second target convolution model. Then, after obtaining the real fault data, the real fault data are assigned different weights, and then input into the first target convolution model and the second target convolution model to obtain the first detection result output by the first target convolution model and the second detection result output by the second target convolution model, as well as the accuracy and weight values ​​corresponding to the detection results. After inputting real fault data into the first and second target convolutional models, each convolutional model may output multiple detection results. For example, after inputting real fault data into the first target convolutional model, two detection results were obtained: Result 1 and Result 2. Result 1 has an accuracy of 60% and a weight of 1.2, while Result 2 has an accuracy of 40% and a weight of 1.5. The output of the second target convolutional model is similar and will not be further described here.

[0053] Step 105: Calculate the weighted accuracy of each detection result based on the accuracy and weight value corresponding to each detection result, and use the detection result with the highest weighted accuracy as the final primary and secondary fusion equipment fault detection result.

[0054] It should be noted that after obtaining the detection results output by all target convolutional models and the corresponding accuracy and weight values, the weighted accuracy of each detection result is calculated based on the accuracy and weight value of each detection result. The detection result with the highest weighted accuracy is used as the final primary and secondary fusion equipment fault detection result. For example, the first detection result output by the first target convolutional model has n (n is an integer not less than 1) results, the i-th (i∈n) result is denoted as Ai, the corresponding accuracy is denoted as Bi, and the weight value corresponding to the accuracy Bi is denoted as Di. Then the weighted accuracy corresponding to the i-th result in the first detection result is Bi×Di. The second detection result output by the second target convolutional model has m (m is an integer not less than 1) results, the j-th (j∈m) result is denoted as Rj, the corresponding accuracy is denoted as Pj, and the weight value corresponding to the accuracy Pj is denoted as Qj. Then the weighted accuracy corresponding to the j-th result in the second detection result is Pj×Qj. The one with the highest weighted accuracy is selected from all the detection results as the final primary and secondary fusion equipment fault detection result, thereby completing the primary and secondary fusion equipment fault detection.

[0055] The fault detection method for primary and secondary fusion equipment provided by the present invention performs fault scenario simulation on a circuit simulation model of the primary and secondary fusion equipment to obtain fault data under the corresponding fault scenario, and then uses the fault data of different scenarios to train different convolution models to obtain different target convolution models. After assigning corresponding weights to the real fault data after a real fault occurs in the use of the primary and secondary fusion equipment, the data is input into different target convolution models for fault detection, and the detection result with the highest weighted accuracy is taken as the final primary and secondary fusion equipment fault detection result. This realizes automatic detection of faults in the primary and secondary fusion equipment, does not require maintenance personnel to go to the site for fault detection, improves detection efficiency, and can avoid the problem of human fault detection errors. It has high accuracy, and solves the technical problem that in the use of existing primary and secondary fusion equipment, after a fault occurs, maintenance personnel need to arrive at the site for fault maintenance, which is inefficient and cannot meet the requirements of rapid detection, affecting the efficiency of line maintenance.

[0056] In one embodiment, the circuit simulation model includes at least five feeders in both a neutral point grounded scenario and a neutral point ungrounded scenario, where the five feeders are an overhead feeder, a first hybrid feeder, a second hybrid feeder, a cable feeder, and a branch feeder.

[0057] For the circuit simulation model in the scenario where the neutral point is not grounded, such as Figure 2 As shown, the system includes a main transformer and five feeders: an overhead feeder, a first hybrid feeder, a second hybrid feeder, a cable feeder, and a branch feeder. The overhead feeder includes a circuit breaker S1, inductors L1-L5, and load switches S2-S5; the first hybrid feeder includes a circuit breaker S6, inductors L6-L10, and load switches S7-S10; the second hybrid feeder includes a circuit breaker S11, inductors L11-L15, and load switches S12-S15; the cable feeder includes a circuit breaker S16, inductors L16-L20, and load switches S17-S20; and the branch feeder includes a circuit breaker S21, inductors L21-L25, and load switches S22 and S23. The system capacitive current is 10A. The model was simulated using digital simulation software. During the simulation, the main transformer capacity of the distribution network was 40MW·A, the transformation ratio was 110 / 10.5KV, and the connection mode was YN / △. The transformation ratio of the distribution transformer was 10.5 / 0.4KV, the connection mode was △ / YN, and the connected loads were all three-phase balanced loads.

[0058] For the circuit simulation model in the neutral point grounding scenario, such as Figure 3As shown, the system includes a main transformer, five feeders, a Z-type grounding transformer 31, and a simulated coil 32. The five feeders are an overhead feeder, a first hybrid feeder, a second hybrid feeder, a cable feeder, and a branch feeder. The overhead feeder includes a circuit breaker S1, inductors L1-L5, and load switches S2-S5. The first hybrid feeder includes a circuit breaker S6, inductors L6-L10, and load switches S7-S10; the second hybrid feeder includes a circuit breaker S11, inductors L11-L15, and load switches S12-S15; the cable feeder includes a circuit breaker S16, inductors L16-L20, and load switches S17-S20; and the branch feeder includes a circuit breaker S21, inductors L21-L25, and load switches S22 and S23. Since the connection mode of the 10kV distribution main transformer is YN / △, a neutral point cannot be formed naturally, so a Z-type grounding transformer is required to form a neutral point.

[0059] In the neutral point grounding scenario, the circuit simulation model includes a Z-type grounding transformer 31, an inductor 321, and a resistor 322. One end of the Z-type grounding transformer 301 is connected to the parallel node of the five feeders. The other end of the Z-type transformer 31 is connected to one end of the resistor 322. The other end of the resistor 322 is connected to one end of the inductor 321. The other end of the inductor 321 is grounded. The inductor 321 and the resistor 322 form a simulated coil 32 to simulate an arc suppression coil. The inductor 321 and the resistor 322 satisfy the following formula:

[0060] L=1 / (1+p)×1 / (3R 2 C)

[0061] Where L is the inductance, p is the compensation degree of the arc suppression coil composed of resistance and inductance, C is the distributed capacitance of the circuit simulation model to ground, and R is the resistance value. Because the zero-sequence current during a fault is closely related to the compensation degree of the arc suppression coil, the simulation compensation degree can be set to 5%.

[0062] The resistance and inductance limited by the above formula can ensure the accuracy of the simulated arc suppression coil, thereby improving the simulation authenticity of the neutral point grounding scenario and also improving the accuracy of the subsequent second target convolution model output detection results.

[0063] In one embodiment, considering that only the fault data at a certain moment is detected, the detection result may have a certain deviation, which will affect the accuracy of the fault detection result. In an embodiment of the present invention, obtaining the real fault data after the primary and secondary fusion equipment fails is specifically obtaining the real fault data at the first moment and the real fault data at the second moment after the primary and secondary fusion equipment fails, wherein the second moment and the first moment are separated by at least 10s.

[0064] After assigning different weights to each real fault data at the first moment and each real fault data at the second moment, they are respectively input into all target convolution models to obtain the first detection result and the second detection result output by each target convolution model corresponding to the first moment and the second moment, as well as the accuracy and weight values ​​corresponding to the first detection result and the second detection result respectively.

[0065] Determine whether a first detection result corresponding to the first moment and a second detection result corresponding to the second moment are the same;

[0066] If the first detection result corresponding to the first moment and the second detection result corresponding to the second moment are the same, then the weighted accuracy of each detection result corresponding to the first moment or the second moment is calculated based on the first detection result or the second detection result and the corresponding accuracy and weight value, and the detection result with the highest weighted accuracy is used as the final primary and secondary fusion equipment fault detection result;

[0067] If the first detection result corresponding to the first moment and the second detection result corresponding to the second moment are different, the real fault data between the first moment and the second moment are obtained, and different weights are assigned to each real fault data between the first moment and the second moment. The data are input into all target convolution models to obtain the detection results output by each target convolution model and the accuracy and weight values ​​corresponding to the detection results. The weighted accuracy of each detection result is calculated according to the accuracy and weight values ​​corresponding to each detection result, and the detection result with the highest weighted accuracy is used as the final primary and secondary fusion equipment fault detection result.

[0068] In an embodiment of the present invention, when a fault occurs in a primary and secondary fusion device, by respectively detecting the real fault data at the first moment and the second moment with a certain interval, the real fault data at two different moments can be obtained respectively, so as to facilitate a more accurate judgment. When the first target convolution model and the second target convolution model respectively detect the real fault data at the first moment and finally obtain the first final detection result that meets the requirements, and the first target convolution model and the second target convolution model respectively detect the actual fault data at the second moment and finally obtain the second final detection result that meets the requirements, by detecting the first final detection result and the second final detection result, if the first final detection result and the second final detection result are different, it indicates that there is a difference in the faults detected at the two moments. Therefore, the real fault data at the middle moment of the first moment and the second moment is used as the real fault data, and this data is detected by the first target convolution model and the second target convolution model to obtain the final target detection result. If the first final detection result and the second final detection result are the same, the first final detection result or the second final detection result is used as the target detection result to complete the detection. Through the above method, it is possible to effectively avoid the situation where only a single fault data at the moment of the fault occurrence or a long time later is used for detection, which may result in inaccurate detection results, thereby effectively improving the accuracy of the first target convolution model and the second target convolution model in detecting fault data.

[0069] For easier understanding, see Figure 4 The present invention provides an embodiment of a fault detection system for primary and secondary fusion equipment, comprising:

[0070] The fault simulation module is used to simulate different fault scenarios for the established circuit simulation model of the primary and secondary fusion equipment and obtain fault data under each fault scenario, where the fault data includes fault factor data and fault results;

[0071] A training data partitioning module is used to divide the fault data under each fault scenario into training data sets, where one training data set corresponds to one fault scenario;

[0072] The model training module is used to assign different weights to the fault factor data in the training data set and input them into the corresponding initial convolution model for training to obtain a trained target convolution model;

[0073] The real fault data processing module is used to obtain real fault data after a fault occurs in the primary and secondary fusion equipment. After assigning different weights to each real fault data, it is input into all target convolution models to obtain the detection results output by each target convolution model and the accuracy and weight value corresponding to the detection results;

[0074] The real fault result output module is used to calculate the weighted accuracy of each detection result based on the accuracy and weight value corresponding to each detection result, and use the detection result with the highest weighted accuracy as the final primary and secondary fusion equipment fault detection result.

[0075] The fault scenarios include a neutral point grounding scenario and a neutral point ungrounded scenario, the training data set includes a first training data set and a second training data set, the initial convolution model includes a first initial convolution model and a second initial convolution model, and the target convolution model includes a first target convolution model and a second target convolution model.

[0076] The circuit simulation model includes at least five feeders in both the neutral point grounding scenario and the neutral point ungrounded scenario. The five feeders are an overhead feeder, a first hybrid feeder, a second hybrid feeder, a cable feeder, and a branch feeder.

[0077] The fault factor data corresponding to the neutral point ungrounded scenario includes: fault initial angle, fault duration, fault transition resistance value and system capacity. The fault factor data corresponding to the neutral point grounded scenario includes: arc suppression coil compensation degree, fault initial angle, fault duration, fault transition resistance value and system capacity.

[0078] Fault results include abnormal equipment operating status, abnormal current and voltage values, overhead faults, mixed faults, branch faults or cable faults.

[0079] Obtain real fault data after a failure of the primary and secondary fusion equipment. After assigning different weights to each real fault data, input it into all target convolution models to obtain the detection results output by each target convolution model and the corresponding accuracy and weight values, including:

[0080] Obtaining real fault data at the first moment and the second moment after a fault occurs on the primary and secondary fusion equipment, wherein the interval between the second moment and the first moment is at least 10 seconds;

[0081] After assigning different weights to each real fault data at the first moment and each real fault data at the second moment, they are input into all target convolution models respectively, and the first detection result and the second detection result output by each target convolution model corresponding to the first moment and the second moment are obtained, as well as the accuracy and weight values ​​corresponding to the first detection result and the second detection result respectively;

[0082] Accordingly, the weighted accuracy of each detection result is calculated based on the accuracy and weight value corresponding to each detection result, and the detection result with the highest weighted accuracy is used as the final primary and secondary fusion equipment fault detection result, including:

[0083] Determine whether a first detection result corresponding to the first moment and a second detection result corresponding to the second moment are the same;

[0084] If the first detection result corresponding to the first moment and the second detection result corresponding to the second moment are the same, then the weighted accuracy of each detection result corresponding to the first moment or the second moment is calculated based on the first detection result or the second detection result and the corresponding accuracy and weight value, and the detection result with the highest weighted accuracy is used as the final primary and secondary fusion equipment fault detection result;

[0085] If the first detection result corresponding to the first moment and the second detection result corresponding to the second moment are different, the real fault data between the first moment and the second moment are obtained, and different weights are assigned to each real fault data between the first moment and the second moment. The data are input into all target convolution models to obtain the detection results output by each target convolution model and the accuracy and weight values ​​corresponding to the detection results. The weighted accuracy of each detection result is calculated according to the accuracy and weight values ​​corresponding to each detection result, and the detection result with the highest weighted accuracy is used as the final primary and secondary fusion equipment fault detection result.

[0086] In a neutral-grounded scenario, the circuit from the neutral point to ground includes a Z-type transformer, a resistor, and an inductor;

[0087] One end of the Z-type transformer is connected to the neutral point, the other end is connected to one end of the resistor, the other end of the resistor is connected to one end of the inductor, and the other end of the inductor is grounded;

[0088] The relationship between inductance and resistance is:

[0089] L=1 / (1+p)×1 / (3R 2 C)

[0090] Wherein, L is the inductance value, p is the compensation degree of the arc suppression coil composed of resistance and inductance, C is the distributed capacitance value of the circuit simulation model to the ground, and R is the resistance value.

[0091] The arc suppression coil compensation degree is 5%.

[0092] The fault detection method for primary and secondary fusion equipment provided by the present invention performs fault scenario simulation on a circuit simulation model of the primary and secondary fusion equipment to obtain fault data under the corresponding fault scenario, and then uses the fault data of different scenarios to train different convolution models to obtain different target convolution models. After assigning corresponding weights to the real fault data after a real fault occurs in the use of the primary and secondary fusion equipment, the data is input into different target convolution models for fault detection, and the detection result with the highest weighted accuracy is taken as the final primary and secondary fusion equipment fault detection result. This realizes automatic detection of faults in the primary and secondary fusion equipment, does not require maintenance personnel to go to the site for fault detection, improves detection efficiency, and can avoid the problem of human fault detection errors. It has high accuracy, and solves the technical problem that in the use of existing primary and secondary fusion equipment, after a fault occurs, maintenance personnel need to arrive at the site for fault maintenance, which is inefficient and cannot meet the requirements of rapid detection, affecting the efficiency of line maintenance.

[0093] The fault detection system for primary and secondary fusion equipment provided in the embodiment of the present invention is used to execute the method in the aforementioned embodiment of the fault detection method for primary and secondary fusion equipment. Its principle is the same as the method in the aforementioned embodiment of the fault detection method for primary and secondary fusion equipment, and will not be repeated here.

[0094] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments can still be modified, or some of the technical features thereof can be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A fault detection method for primary and secondary fusion equipment, characterized in that: include: Conduct simulations of different fault scenarios on the established circuit simulation model of primary and secondary fusion equipment to obtain fault data under each fault scenario, where the fault data includes fault factor data and fault results; Divide the fault data under each fault scenario into training data sets, where one training data set corresponds to one fault scenario; After assigning different weights to the fault factor data in the training dataset, the data is input into the corresponding initial convolution model for training to obtain the trained target convolution model. Obtain real fault data after a failure of the primary and secondary fusion equipment, assign different weights to each real fault data, and input it into all target convolution models to obtain the detection results output by each target convolution model, as well as the accuracy and weight value corresponding to the detection results; Calculate the weighted accuracy of each test result based on the accuracy and weight value corresponding to each test result, and use the test result with the highest weighted accuracy as the final primary and secondary fusion equipment fault detection result; Obtain real fault data after a failure of the primary and secondary fusion equipment. After assigning different weights to each real fault data, input it into all target convolution models to obtain the detection results output by each target convolution model and the corresponding accuracy and weight values, including: Obtaining real fault data at the first moment and the second moment after a fault occurs on the primary and secondary fusion equipment, wherein the interval between the second moment and the first moment is at least 10 seconds; After assigning different weights to each real fault data at the first moment and each real fault data at the second moment, they are input into all target convolution models respectively, and the first detection result and the second detection result output by each target convolution model corresponding to the first moment and the second moment are obtained, as well as the accuracy and weight values ​​corresponding to the first detection result and the second detection result respectively; Accordingly, the weighted accuracy of each detection result is calculated based on the accuracy and weight value corresponding to each detection result, and the detection result with the highest weighted accuracy is used as the final primary and secondary fusion equipment fault detection result, including: Determine whether a first detection result corresponding to the first moment and a second detection result corresponding to the second moment are the same; If the first detection result corresponding to the first moment and the second detection result corresponding to the second moment are the same, then the weighted accuracy of each detection result corresponding to the first moment or the second moment is calculated based on the first detection result or the second detection result and the corresponding accuracy and weight value, and the detection result with the highest weighted accuracy is used as the final primary and secondary fusion equipment fault detection result; If the first detection result corresponding to the first moment and the second detection result corresponding to the second moment are different, the real fault data between the first moment and the second moment are obtained, and different weights are assigned to each real fault data between the first moment and the second moment. The data are input into all target convolution models to obtain the detection results output by each target convolution model and the accuracy and weight values ​​corresponding to the detection results. The weighted accuracy of each detection result is calculated according to the accuracy and weight values ​​corresponding to each detection result, and the detection result with the highest weighted accuracy is used as the final primary and secondary fusion equipment fault detection result.

2. The fault detection method for primary and secondary fusion equipment according to claim 1, characterized in that: The fault scenarios include a neutral point grounding scenario and a neutral point ungrounded scenario, the training data set includes a first training data set and a second training data set, the initial convolution model includes a first initial convolution model and a second initial convolution model, and the target convolution model includes a first target convolution model and a second target convolution model.

3. The fault detection method for primary and secondary fusion equipment according to claim 2, characterized in that: The circuit simulation model includes at least five feeders in both the neutral point grounding scenario and the neutral point ungrounded scenario. The five feeders are an overhead feeder, a first hybrid feeder, a second hybrid feeder, a cable feeder, and a branch feeder.

4. The fault detection method for primary and secondary fusion equipment according to claim 3, characterized in that: The fault factor data corresponding to the neutral point ungrounded scenario includes: fault initial angle, fault duration, fault transition resistance value and system capacity. The fault factor data corresponding to the neutral point grounded scenario includes: arc suppression coil compensation degree, fault initial angle, fault duration, fault transition resistance value and system capacity.

5. The fault detection method for primary and secondary fusion equipment according to claim 4, characterized in that: Fault results include abnormal equipment operating status, abnormal current and voltage values, overhead faults, mixed faults, branch faults or cable faults.

6. The fault detection method for primary and secondary fusion equipment according to claim 3, characterized in that: In a neutral-grounded scenario, the circuit from the neutral point to ground includes a Z-type transformer, a resistor, and an inductor; One end of the Z-type transformer is connected to the neutral point, the other end is connected to one end of the resistor, the other end of the resistor is connected to one end of the inductor, and the other end of the inductor is grounded; The relationship between inductance and resistance is: ； Wherein, L is the inductance value, p is the compensation degree of the arc suppression coil composed of resistance and inductance, C is the distributed capacitance value of the circuit simulation model to the ground, and R is the resistance value.

7. The fault detection method for primary and secondary fusion equipment according to claim 6, characterized in that: The compensation degree of arc suppression coil is 5%.

8. A fault detection system for primary and secondary fusion equipment, characterized in that: include: The fault simulation module is used to simulate different fault scenarios for the established circuit simulation model of the primary and secondary fusion equipment and obtain fault data under each fault scenario, where the fault data includes fault factor data and fault results; A training data partitioning module is used to divide the fault data under each fault scenario into training data sets, where one training data set corresponds to one fault scenario; The model training module is used to assign different weights to the fault factor data in the training data set and input them into the corresponding initial convolution model for training to obtain a trained target convolution model; The real fault data processing module is used to obtain real fault data after a fault occurs in the primary and secondary fusion equipment. After assigning different weights to each real fault data, it is input into all target convolution models to obtain the detection results output by each target convolution model and the accuracy and weight value corresponding to the detection results; The real fault result output module is used to calculate the weighted accuracy of each test result based on the accuracy and weight value corresponding to each test result, and use the test result with the highest weighted accuracy as the final primary and secondary fusion equipment fault detection result; Obtain real fault data after a failure of the primary and secondary fusion equipment. After assigning different weights to each real fault data, input it into all target convolution models to obtain the detection results output by each target convolution model and the corresponding accuracy and weight values, including: Obtaining real fault data at the first moment and the second moment after a fault occurs on the primary and secondary fusion equipment, wherein the interval between the second moment and the first moment is at least 10 seconds; After assigning different weights to each real fault data at the first moment and each real fault data at the second moment, they are input into all target convolution models respectively, and the first detection result and the second detection result output by each target convolution model corresponding to the first moment and the second moment are obtained, as well as the accuracy and weight values ​​corresponding to the first detection result and the second detection result respectively; Accordingly, the weighted accuracy of each detection result is calculated based on the accuracy and weight value corresponding to each detection result, and the detection result with the highest weighted accuracy is used as the final primary and secondary fusion equipment fault detection result, including: Determine whether a first detection result corresponding to the first moment and a second detection result corresponding to the second moment are the same; If the first detection result corresponding to the first moment and the second detection result corresponding to the second moment are the same, then the weighted accuracy of each detection result corresponding to the first moment or the second moment is calculated based on the first detection result or the second detection result and the corresponding accuracy and weight value, and the detection result with the highest weighted accuracy is used as the final primary and secondary fusion equipment fault detection result; If the first detection result corresponding to the first moment and the second detection result corresponding to the second moment are different, the real fault data between the first moment and the second moment are obtained, and different weights are assigned to each real fault data between the first moment and the second moment. The data are input into all target convolution models to obtain the detection results output by each target convolution model and the accuracy and weight values ​​corresponding to the detection results. The weighted accuracy of each detection result is calculated according to the accuracy and weight values ​​corresponding to each detection result, and the detection result with the highest weighted accuracy is used as the final primary and secondary fusion equipment fault detection result.

9. The fault detection system for primary and secondary fusion equipment according to claim 8, characterized in that: The fault scenarios include a neutral point grounding scenario and a neutral point ungrounded scenario, the training data set includes a first training data set and a second training data set, the initial convolution model includes a first initial convolution model and a second initial convolution model, and the target convolution model includes a first target convolution model and a second target convolution model.

Images