Coal mine leakage protection method, device, equipment and program based on leakage fault recognition model

Abstract

The invention provides a coal mine electric leakage protection method based on an electric leakage fault identification model. The method comprises the following steps: collecting electric quantity signals in each line of a power grid; when it is detected that the zero-sequence voltage amplitude of any line is larger than a preset value, sampling values of the electric quantity signals in N cycles before and after the fault occurrence moment are intercepted, and data to be recognized are obtained after preprocessing; inputting the to-be-identified data into a pre-trained electric leakage identification model to obtain an electric leakage fault identification result; and issuing a fault removal instruction to the protection unit of the corresponding line according to the electric leakage fault identification result. According to the scheme, the AI model is combined with the conditional attention mechanism of the power grid topology to dynamically adjust the feature weight, the dynamic routing controller is introduced in the feature learning stage, the optimized feature extraction branches are switched in real time according to the fault severity, and compared with a traditional method, the method has the advantages that the efficiency is high; according to the scheme, the electric leakage detection accuracy under the underground complex working condition is obviously improved, and the fault response time is obviously shortened.

Description

Technical Field

[0001] The present disclosure relates to the technical field of power supply, and particularly to a coal mine leakage protection method, device, equipment and program based on a leakage fault identification model. Background Art

[0002] In coal mine power grids with voltages of 10 kV, 6 kV, 1140 V, etc., a neutral non-effective grounding system is mostly adopted, including a neutral non-grounding system and an arc suppression coil grounded system. A single-phase leakage fault (or single-phase grounding fault) is the most frequently occurring electrical fault in coal mine power grids. When a single-phase leakage fault occurs, it will cause the neutral point voltage and the non-grounded phase voltage to rise, which is likely to develop into a short-circuit fault, resulting in a large-scale power outage in the coal mine underground. It is also likely to generate electric sparks and trigger gas explosions, causing major safety accidents.

[0003] The detection and protection technology for single-phase grounding in neutral non-effective grounding systems has always been a research hotspot. In recent years, it has also become a reality to detect single-phase leakage in coal mine power grids using artificial intelligence methods. In view of the particularity of the coal mine production environment, a highly reliable leakage detection technology is required. Summary of the Invention

[0004] The present invention provides a coal mine leakage protection solution based on a leakage fault identification model to solve the problem of insufficient reliability of the single-phase grounding detection and protection technology for coal mine neutral non-effective grounding systems.

[0005] The present invention solves the above technical problems through the following aspects:

[0006] In a first aspect, the present invention provides a coal mine leakage protection method based on a leakage fault identification model, including:

[0007] Collecting power quantity signals in each line of the power grid, where the power quantity signals include three-phase voltages, three-phase currents, zero-sequence voltages, and zero-sequence currents;

[0008] When it is detected that the zero-sequence voltage amplitude of any line is greater than a preset value, intercepting the sampling values of the power quantity signals for N cycles before and after the fault occurrence moment, and obtaining the data to be identified after preprocessing;

[0009] Inputting the data to be identified into a pre-trained leakage identification model to obtain a leakage fault identification result;

[0010] Sending a fault removal instruction to the protection unit of the corresponding line according to the leakage fault identification result.

[0011] In a second aspect, the present invention provides a coal mine leakage protection device based on a leakage fault identification model, including:

[0012] A collection module, configured to collect power quantity signals in each line of the power grid, where the power quantity signals include three-phase voltage, three-phase current, zero-sequence voltage, and zero-sequence current;

[0013] A data acquisition module, configured to intercept the sampling values of the power quantity signals for N cycles before and after the fault occurrence moment after detecting that the zero-sequence voltage amplitude of any line is greater than a preset value, and obtain the data to be identified after preprocessing;

[0014] A detection module, configured to input the data to be identified into a pre-trained leakage fault identification model to obtain a leakage fault identification result;

[0015] A control module, configured to send a fault removal instruction to the protection unit of the corresponding line according to the leakage fault identification result.

[0016] In a third aspect, the present invention provides a coal mine leakage protection device based on a leakage fault identification model, which is arranged in a power grid line. The protection device includes a power grid interface, a data interface, a processor, and a memory. Among them,

[0017] The power grid interface is connected to the processor and is configured to connect to the power grid line;

[0018] The data interface is connected to the processor and is configured to communicate with a management platform;

[0019] The memory stores instructions executable by the processor, and the instructions are executed by the processor to implement the foregoing protection method.

[0020] In a fourth aspect, the present invention provides a computer program, which, when executed by a processor, causes the processor to implement the foregoing protection method.

[0021] The present invention monitors the power signals in the power grid, intercepts the sampled values of multiple cycles before and after a fault occurs for AI recognition to determine the fault, fuses multi-domain features through spatio-temporal domain joint coding, combines the conditional attention mechanism of the power grid topology to dynamically adjust the feature weights, and innovatively introduces a dynamic routing controller in the feature learning stage to switch the optimized feature extraction branches in real time according to the fault severity. Through the parameter dynamic adaptation mechanism of the extreme learning machine, the collaborative optimization of fault classification and understanding is realized. The AI model adopted by the present invention designs a dynamic attention weighting layer, encodes the power grid structure information as attention weights, makes the feature learning process conform to the propagation law of the power system, and makes up for the defect that the traditional method ignores the topological correlation. The AI model adopted by the present invention designs multiple learning branches, which can focus on the current change rate feature in case of slight leakage, focus on the harmonic distortion feature in case of severe leakage, and capture the waveform mutation feature in case of instantaneous fault. This flexible feature adaptation mechanism realizes the leap from "perceiving faults" to "understanding faults". Compared with the traditional method, the leakage detection accuracy of this solution is improved by 15.7% under the complex working conditions in the mine, the positioning accuracy reaches the topological node level, and the response time is shortened to within 30 ms. Description of the Drawings

[0022] In order to more clearly illustrate the technical solutions in the embodiments of the present disclosure or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments recorded in the present disclosure. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0023] Figure 1 It is a flowchart of the leakage protection method based on the leakage fault recognition model provided by the embodiment of the present disclosure;

[0024] Figure 2 It is a schematic structural diagram of the leakage fault recognition model in the embodiment of the present disclosure;

[0025] Figure 3 It is a schematic network structure diagram of ELM;

[0026] Figure 4 It is a structural block diagram of the mine leakage protection device 300 based on the leakage fault recognition model provided by the embodiment of the present disclosure;

[0027] Figure 5 It is a schematic structural diagram of the protection device 400 provided by the embodiment of the present disclosure. Detailed Embodiments

[0028] To enable those skilled in the art to better understand the technical solutions in this disclosure, the following will clearly and completely describe the technical solutions in this disclosure in conjunction with the accompanying drawings in the embodiments of this disclosure. Obviously, the described embodiments are only a part of the embodiments of this disclosure, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of this disclosure without creative efforts shall fall within the scope of protection of this disclosure. In addition, for clarity, parts irrelevant to the description of the exemplary embodiments are omitted in the drawings.

[0029] In this specification, it should be understood that terms such as "including" or "having" are intended to indicate the existence of the features, numbers, steps, actions, components, parts, or combinations thereof disclosed in this disclosure, and do not intend to exclude the possibility of the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof. It should also be noted that, without conflict, the embodiments in this disclosure and the features in the embodiments can be combined with each other.

[0030] The leakage faults occurring in the coal mine power supply system include arc grounding faults, high-resistance grounding faults, metal grounding faults, instantaneous grounding faults, and intermittent grounding faults. When a leakage fault occurs, the commonly recognizable fault characteristics are: harmonic characteristics, fifth-harmonic characteristics, first half-wave characteristics, zero-sequence voltage amplitude characteristics, zero-sequence current amplitude characteristics, phase characteristics, zero-sequence active power characteristics, zero-sequence admittance characteristics, S-injection method characteristics, and medium-resistance grounding characteristics.

[0031] The manifestations and functions of these fault characteristics are described below.

[0032] Zero-sequence voltage amplitude characteristics: When a single-phase ground fault occurs in an ungrounded neutral system or a system with a neutral point grounded through an arc suppression coil, zero-sequence voltage can be detected throughout the power supply system. The magnitude of the zero-sequence voltage is related to the magnitude of the grounding resistance. When there is a high-resistance ground, the zero-sequence voltage is small; when there is a metal direct ground, the zero-sequence voltage is the largest, approaching the line voltage of the power supply line, and the fault line and non-fault line cannot be located.

[0033] Zero-sequence current amplitude characteristics: In an ungrounded neutral system, when a single-phase ground fault occurs, the zero-sequence current of the fault line is equal to the sum of the capacitive currents of all non-fault lines in the whole system to the ground; in a system with a neutral point grounded through an arc suppression coil, when a single-phase ground fault occurs, the zero-sequence current of the fault line is equal to the sum of the capacitive current of the fault line itself and the compensation current of the arc suppression coil in the power supply system minus the sum of the capacitive currents of other non-fault lines.

[0034] Phase characteristics: In an ungrounded neutral system, when a single-phase grounding occurs, the zero-sequence current direction of the faulty line points from the line to the bus, while that of the non-faulty line points from the bus to the line. The zero-sequence current directions of the faulty line and the non-faulty line are exactly opposite. In a system with an arc suppression coil grounded at the neutral point, when a single-phase grounding occurs, the arc suppression coil is generally overcompensated. The zero-sequence current directions of both the faulty line and the non-faulty line point from the bus to the line, and at this time, it is impossible to distinguish between the faulty line and the non-faulty line.

[0035] Zero-sequence active current characteristics: The active current formed based on the resistive characteristics of the grounding point of the faulty line is related to the grounding resistance of the grounded line. When the grounding is a high-resistance grounding, the zero-sequence active current is small; when it is a metallic grounding, the zero-sequence active current is large.

[0036] Zero-sequence admittance characteristics: In an ungrounded neutral system, when a single-phase grounding occurs, the zero-sequence admittance of the faulty line is equal to the sum of the zero-sequence admittances of all non-faulty lines to the ground in the whole system. In a system with an arc suppression coil grounded at the neutral point, when a single-phase grounding occurs, the zero-sequence admittance of the faulty line is equal to the sum of the zero-sequence admittance of the faulty line itself and the zero-sequence admittance of the arc suppression coil in the power supply system minus the sum of the zero-sequence admittances of other non-faulty lines.

[0037] Harmonic characteristics or high-frequency component characteristics: In the case of a metallic grounding fault, mainly the fundamental wave current increases significantly, and the harmonic components are less. In the case of an arcing grounding, instantaneous grounding or intermittent grounding fault, mainly high-frequency component harmonics are generated, mainly odd harmonics, and the harmonics with large contents are mainly the third harmonic, fifth harmonic and seventh harmonic.

[0038] Fifth-harmonic characteristics: The fifth harmonics in the coal mine power supply are mainly generated by frequency conversion equipment, rectification equipment and other equipment when a leakage fault occurs. In an ungrounded neutral system, when a single-phase grounding occurs, the fifth harmonics of the faulty line are much larger than those of the non-faulty line.

[0039] First half-wave characteristics: When a single-phase grounding leakage fault occurs, an obvious mutation will appear within the first half-wave of the current or voltage waveform. This mutation is caused by the abnormal increase in current or abnormal decrease in voltage at the fault point. The first half-wave characteristics are mainly the change characteristics of the current or voltage waveform at the moment of the fault. During the first half-wave, there are obvious differences between the zero-sequence current waveforms of the faulty branch and the non-faulty branch in an ungrounded neutral system or a system with an arc suppression coil grounded at the neutral point. In a system with an arc suppression coil grounded at the neutral point, the influence of the arc suppression coil on the zero-sequence current during the first half-wave is small (after the compensation time of the arc suppression coil is greater than half a cycle).

[0040] Features of the S injection method: During single-phase grounding, a non-power-frequency signal is injected into the power supply system through the PT, mainly in the range of 80HZ to 200HZ. The protection terminal detects and tracks the injected signal to achieve line selection and tripping. The injected off-frequency signal is easily interfered by fault signals and harmonics. Especially during single-phase grounding, the non-power-frequency signal generated by the system is much larger than the injected signal, making it impossible to accurately locate the faulty line and the non-faulty line.

[0041] Features of medium-resistance grounding: During single-phase grounding, a medium-value resistor is connected between the neutral point and the ground, and a current loop is formed through the ground via the grounding point of the faulty line to limit the magnitude of the single-phase grounding fault current, prevent the arc from reigniting, and suppress the generation of system overvoltage. The resistive current magnitude at the grounding point of the faulty line is used to detect and locate the faulty line and the non-faulty line, but it has limitations for high-resistance grounding.

[0042] The present invention establishes a feature library through a large number of leakage faults, pre-trains an AI model, collects three-phase voltage and current, zero-sequence voltage and current. When the monitored value of the zero-sequence voltage exceeds the preset range, the pre-trained AI model is used for intelligent recognition, and the accurate determination of leakage faults is achieved by artificial intelligence methods. The implementation scheme of the present invention is specifically described below.

[0043] Figure 1 It is a flowchart of the leakage protection method based on the leakage fault recognition model provided by the embodiments of the present disclosure.

[0044] As Figure 1 shown, the method includes steps S110 to S140.

[0045] S110: Collect power quantity signals in each line of the power grid. The power quantity signals include three-phase voltage, three-phase current, zero-sequence voltage, and zero-sequence current.

[0046] S120: When it is detected that the zero-sequence voltage amplitude of any line is greater than the preset value, intercept the sampling values of the power quantity signals in the N cycles before and after the fault occurrence moment, and obtain the data to be recognized after preprocessing.

[0047] S130: Input the data to be recognized into the pre-trained leakage recognition model to obtain the leakage fault recognition result.

[0048] S140: Send a fault removal instruction to the protection unit of the corresponding line according to the leakage fault recognition result.

[0049] According to the method of this embodiment, the power signal can be obtained by the protection unit arranged in the power grid line, or can be obtained by other sensors or acquisition devices in the power grid. Since the zero-sequence voltage is usually only related to the grounding resistance, and the zero-sequence current (leakage current) is related to the grounding resistance, the grounding method (whether the neutral point is equipped with an arc suppression coil), and the magnitude of the capacitive current of the system cable to the ground, it is possible to monitor whether leakage occurs through the amplitude of the zero-sequence voltage. For example, when the grounding resistance is 2 kΩ, the zero-sequence voltage of the open delta is slightly higher than 15 V. Therefore, the preset value of the zero-sequence voltage amplitude can be set to 15 V to be able to detect minor leakage faults. When a leakage fault occurs, the zero-sequence voltage always increases. A zero-sequence voltage of 15 V can reflect the occurrence of leakage, but it cannot accurately determine the line where the leakage occurs. It is also necessary to identify the leakage line through the leakage fault identification model.

[0050] The power supply of the coal mine power grid system is three-phase alternating current power supply. The three-phase alternating current consists of three sinusoidal alternating currents with the same frequency, equal amplitude, and a phase difference of 120°. Through the primary three-phase voltage transformer, current transformer, zero-sequence current transformer, etc. in the high-voltage switch and the peripheral signal conditioning hardware of the secondary protection terminal. When collecting the power signal, an ADC (Analog-to-Digital Converter) is used to convert the analog signal of the power grid into a digital signal. The higher the sampling frequency, the more accurate the waveform can be reflected, and at the same time, the higher the requirement for calculation. For example, when selecting a sampling frequency of 12,800 Hz, 256 points can be collected per cycle. The sampled signal of the zero-sequence voltage obtained can be used to calculate the zero-sequence voltage amplitude in real time through the discrete Fourier series algorithm. When the zero-sequence voltage amplitude is greater than the preset value, such as 15 V, start to record the sampling values of all voltage and current signals (three-phase voltage, three-phase current, zero-sequence voltage, and zero-sequence current) for N cycles before the fault and the sampling values of all voltage and current for N cycles after the fault. Then, take the sampling values of all 2N cycles as the original data, and obtain the data to be identified after preprocessing. Preferably, N can be set to 5, that is, preprocess the sampling values of 10 cycles. Collecting the signals of N cycles before and after the fault can reduce the false alarms caused by equipment startup or strong electrical disturbances in the power grid. The preprocessing of the sampled signal includes operations such as band-pass filtering, wavelet denoising, sliding window alignment, and 3σ outlier rejection on the sampling values of each electrical signal.

[0051] Figure 2 It is a schematic structural diagram of the leakage fault identification model in the embodiments of the present disclosure.

[0052] As Figure 2 shown, the leakage fault identification model includes a feature extraction module, a spatio-temporal domain joint coding module, a dynamic routing controller, and multiple feature learning branches.

[0053] The feature extraction module is used to extract multi-domain features from the data to be recognized. The multi-domain features include time-domain waveforms, frequency-domain harmonic energy distributions, and statistical-domain mutation parameters.

[0054] Perform multi-scale and multi-directional convolution operations on the leakage data to be recognized. The basic elements of the fault features extracted mainly include:

[0055] 1) Waveform change: Waveform change refers to the form of current or voltage signals changing over time. Leakage faults often cause waveform distortion, jitter, or mutation, and these changes can be captured by the model through convolution operations and used as important bases for fault judgment.

[0056] 2) Frequency features: Frequency features are the manifestations of signals in the frequency domain, reflecting the distribution and intensity of different frequency components in the signals. Leakage faults may cause changes in frequency components, such as increased harmonics and main frequency shift, and these features can also be extracted by the model through convolution operations.

[0057] 3) Current intensity and change rate: By monitoring the intensity of the current and its change rate over time, the model can capture abnormal current fluctuations, which are important indicators of leakage faults. The convolution kernel can be designed as a filter sensitive to current intensity and time change rate to extract these features.

[0058] 4) Voltage level and fluctuation: The stability of voltage is a key indicator for power grid operation. The model analyzes the voltage level and its fluctuations to identify potential voltage abnormalities. The convolution kernel can capture the smoothness and volatility of voltage signals, providing a strong basis for fault judgment.

[0059] 5) Power factor and reactive power: The power factor reflects the ratio of useful power to apparent power in the power grid, and the flow of reactive power may lead to a decrease in power grid efficiency. By monitoring these parameters, the model helps to detect changes in the power factor caused by leakage and abnormal flow of reactive power. The convolution kernel can be designed as a filter sensitive to changes in power factor and reactive power.

[0060] 6) Harmonic content and distribution: Harmonics in the power grid are generated by nonlinear loads, and they may cause damage to power grid equipment. The model analyzes the harmonic content and its distribution to evaluate the impact of leakage faults on the harmonic characteristics of the power grid. The convolution kernel can capture the frequency components and amplitudes of harmonic signals, providing harmonic features for fault judgment.

[0061] 7) Phase angle and phase difference: The phase angle is the angle between current and voltage, and the phase difference is the angular difference between different phases. By monitoring these parameters, the model can identify phase abnormalities, which may be related to leakage faults. The convolution kernel can be designed as a filter sensitive to changes in phase angle and phase difference.

[0062] In this embodiment, during feature extraction, the current rise rate and the voltage drop depth are mainly extracted from the time-domain waveform, the energy ratios of the 3rd to 11th harmonics are calculated for the frequency-domain harmonics, the mutation points are detected by the CUSUM (Cumulative Sum) algorithm within a sliding window, and the peak-to-peak variance is calculated to obtain the mutation parameters in the statistical domain. The power frequency interference is eliminated from the time-domain waveform by using adaptive notch filtering; for frequency-domain analysis, windowed FFT and sub-band energy normalization are adopted.

[0063] The spatio-temporal domain joint encoding module is used to jointly encode the multi-domain features to obtain the encoded features. Multi-dimensional encoded features can be generated through time-domain kurtosis analysis, frequency-domain harmonic analysis, and statistical domain mutation parameters.

[0064] The dynamic routing controller is used to activate the corresponding feature learning branches according to real-time evaluation metrics, and the feature learning branches are used to learn the encoded features. Multiple feature learning branches are implemented based on the SAE (Stacked Auto Encoder) model. Multiple branches respectively perform feature learning for slight leakage, severe leakage, and instantaneous faults. Among them, the first branch focuses on the current change rate feature, and a convolutional neural network is used to extract the current change rate feature, which is activated when the current change rate exceeds the first threshold; the second branch focuses on the harmonic distortion feature, and a frequency-domain attention network is used to extract the harmonic distortion feature, which is activated when the total harmonic distortion rate exceeds the second threshold; the third branch captures the waveform mutation feature, and a long short-term memory network is used to extract the waveform mutation feature, which is activated when the mutation point density exceeds the third threshold.

[0065] The activation thresholds corresponding to the dynamic routing controller can be set as follows: for example, the current change rate threshold is 5 A / ms, which is determined as slight leakage and the first branch is activated; the total harmonic distortion rate (THD) is greater than the threshold of 5%, which is determined as relatively severe leakage and the second branch is activated; the mutation point density threshold is 3 per cycle, which is determined as an instantaneous fault and the third branch is activated. Alternatively, the activation thresholds of the dynamic routing controller can be set according to the range of the zero-sequence voltage amplitude. For example, when the zero-sequence voltage is 15 v to 30 v, it is determined as slight leakage and the first branch is activated; when the zero-sequence voltage is 30 v to 90 v, it is determined as relatively severe leakage and the second branch is activated; when the zero-sequence voltage is greater than 90 v (severe leakage, the fault cannot persist for a long time), it is determined as an instantaneous fault and the third branch is activated.

[0066] In the leakage fault identification model of this embodiment, the spatio-temporal domain joint encoding module is also used to generate conditional attention weights based on the adjacency matrix of the power grid topology and dynamically weight the encoded features. The adjacency matrix is a 0-1 matrix generated based on the coal mine power supply network topology (adjacent nodes are 1, non-adjacent nodes are 0). Using the power grid topology for dynamic weighting can effectively adapt to the power grid structure, calculate the distance of the fault node according to the propagation delay of the fault characteristics, focus on the lines on the fault propagation path, improve the accuracy of fault location, and reduce the calculation time-consuming for irrelevant lines. In the conditional attention mechanism, the key is the basic element of the fault feature, and the value is the corresponding leakage fault. By calculating the similarity between the key and the query, the model can dynamically adjust the attention to different features, so as to more accurately judge the leakage fault. The calculation of the conditional attention weight is obtained according to the following formula:

[0067]

[0068] where α ij is the attention weight of node i to node j, W q is the query matrix, W k is the key matrix, A ij is the element of the power grid topology adjacency matrix, λ is the topology influence coefficient, and σ is a non-linear activation function, preferably GELU.

[0069] During the calculation, is used to stabilize the gradient (d k is the key vector dimension), W q and W k are both trainable parameters, and λA ij enables the model to autonomously learn the correlation between electrical parameters and topology, and the GELU activation can enhance the non-linear expression ability. The value of A ij is 0 or 1, indicating the connectivity of the node. The topology influence coefficient λ is a hyperparameter and is an explicit coupling term of the power grid adjacency matrix. The value range of λ can be between [0.1, 0.5]. Through A ij embeds the physical connection relationship of the power grid into the attention calculation (such as busbars, feeder connection points), avoiding over-response to irrelevant nodes. For example, if node A and node B are physically connected in the power grid (A AB = 1), then when the model processes the features of node A, it will strengthen the attention calculation of the features of node B. The λ coefficient controls the topology influence intensity and can adapt to the wiring density differences of different mines. When the power grid topology is complex, such as multi-branch power supply, increasing λ can strengthen the topology constraint; when the feature correlation is significant, such as high-frequency harmonics are prominent, reducing λ to focus on data-driven.

[0070] The weight coefficient α ijIt changes in real time with the input features. Through the combination of explicit topological coding and implicit feature learning, a weight distribution that conforms more to the physical laws of the power grid than traditional attention mechanisms is achieved, which is beneficial to improving the accuracy of leakage location. Specifically, it is manifested in the following aspects:

[0071] (1) Enhanced spatial correlation: In the analysis of the leakage fault propagation path, priority is given to paying attention to the characteristics of adjacent lines through which the fault current flows (such as bus → feeder → load node). For example, when an arc grounding occurs in a certain branch, the model automatically strengthens the harmonic feature weights of this branch and adjacent nodes, while suppressing the noise signals of irrelevant branches.

[0072] (2) Topological filtering: For abnormal features of non - adjacent nodes, such as the instantaneous fluctuations of remote loads, low weights are assigned to reduce the misjudgment rate.

[0073] (3) Hierarchical weight distribution: In the deep network, a large λ value is adopted for the high - voltage side (sparse topological connection) to strengthen the structural constraints; a small λ value is adopted for the low - voltage side (dense connection) to focus on feature correlation.

[0074] The leakage fault identification model in this embodiment may further include a feature fusion module and an ELM (extreme learning machine) classifier.

[0075] The feature fusion module is used to perform weighted fusion on the feature vectors output by multiple feature learning branches through a dynamic gating mechanism, and dynamically adjust the hidden - layer activation function and regularization parameters of the ELM classifier based on the activated feature learning branches. If the first branch is activated, the ReLU activation function is adopted for the hidden layer of the ELM and the L1 regularization constraint is imposed; if the second branch is activated, the Sigmoid activation function is adopted for the hidden layer and the L2 regularization constraint is imposed; if it is the third branch, the LeakyReLU activation function is adopted for the hidden layer and the Dropout mechanism is enabled.

[0076] ELM is a feed - forward neural network algorithm, which consists of an input layer, a hidden layer, and an output layer. ELM has strong generalization ability and high learning efficiency, and is widely used in various fields. Its network structure is as Figure 3 shown. Usually, when solving classification problems, deep - learning methods adopt the Softmax classifier, which classifies based on the probability principle, strengthens the maximum eigenvalue and weakens the influence of other features, and has a poor recognition effect on fault types with small feature differences. To overcome the deficiencies of Softmax, this embodiment adopts the ELM classifier, which has good non - linear fitting ability and strong self - learning ability for fault data, and is suitable for identifying leakage fault types with small feature differences.

[0077] After the ELM classifier processes the steady-state information fault criterion and the transient information fault criterion through the fault measure function, information fusion is performed on them, and the faulty line is selected jointly through multiple fault feature information, improving the accuracy of leakage fault line selection. The number of hidden layer nodes of the ELM can be determined through experiments. Preferably, the number of hidden layer nodes is 9, which can obtain a high line selection accuracy with a relatively small number of nodes and take into account the detection efficiency.

[0078] During implementation, the data to be recognized obtained after preprocessing is input Figure 2 into the leakage fault recognition model shown in the figure, obtaining a fault recognition result, and sending a fault removal instruction to the corresponding protection device. The monitoring platform can execute the corresponding maintenance plan according to the recognition result.

[0079] The method of this embodiment can identify different types of leakage faults. It focuses on the current change rate feature during slight leakage, the harmonic distortion feature during severe leakage, and the waveform mutation feature during instantaneous faults. This flexible feature adaptation mechanism realizes the leap from "perceiving faults" to "understanding faults". Compared with traditional methods, the leakage detection accuracy of this solution is increased by 15.7% under the complex underground working conditions, the positioning accuracy reaches the topological node level, and the response time is shortened to within 30 ms.

[0080] The embodiments of the coal mine leakage protection method based on the leakage fault recognition model have been described above. Correspondingly, the present disclosure also provides an embodiment of a coal mine leakage protection device based on the leakage fault recognition model.

[0081] Figure 4 It is a structural block diagram of a coal mine leakage protection device 300 based on the leakage fault recognition model provided by an embodiment of the present disclosure. The device 300 can be implemented as part or all of an electronic device through software, hardware, or a combination of both.

[0082] As Figure 4 shown, the device 300 includes a collection module 310, a data acquisition module 320, a detection module 330, and a control module 340. The device 300 can execute the protection method as described above.

[0083] The collection module 310 is used to collect the power quantity signals in each line of the power grid, and the power quantity signals include three-phase voltage, three-phase current, zero-sequence voltage, and zero-sequence current;

[0084] The data acquisition module 320 is used to intercept the sampling values of the power quantity signals for N cycles before and after the fault occurrence moment when the zero-sequence voltage amplitude of any line is detected to be greater than a preset value, and obtain the data to be recognized after preprocessing;

[0085] The detection module 330 is configured to input the data to be recognized into a pre-trained leakage recognition model to obtain a leakage fault recognition result;

[0086] The control module 340 is configured to issue a fault removal instruction to the protection unit of the corresponding line according to the leakage fault recognition result.

[0087] Based on the same inventive concept, the present disclosure also provides a coal mine leakage protection device 400 based on a leakage fault recognition model, which is arranged in a power grid line. This protection device 400 can be implemented as part or all of an electronic device through software, hardware, or a combination of both.

[0088] Figure 5 It is a schematic structural diagram of the protection device 400 provided by the embodiments of the present disclosure. As Figure 5 shown, the protection device includes a power grid interface 410, a data interface 420, a processor 430, and a memory 440.

[0089] The power grid interface 410 is connected to the processor 430 and is used to connect to the power grid line.

[0090] The data interface 420 is connected to the processor 430 and is used to communicate with the management platform.

[0091] The memory 440 stores instructions executable by the processor 430, and these instructions are executed by the processor 430 to implement the various methods described above.

[0092] This protection device 400 can exist in a variety of physical entities, including intelligent IoT protection terminals, mine explosion-proof and intrinsically safe high-voltage vacuum distribution devices, or low-voltage intelligent boxes. These coal mine leakage protection devices are communicatively connected to a pre-trained AI model. In some embodiments, the processor 430 can be a processor with AI computing power, and the device has a pre-trained AI model built-in for real-time calculation. In other embodiments, the pre-trained AI model is configured on an edge computing platform, and this protection device 400 is connected to it through a transmission line.

[0093] The embodiments of the present disclosure also provide a computer program, which, when executed by a processor, enables the processor to implement the various protection methods described above. This computer program can be built into a power protection device, a distribution device, or other electrical equipment to become part of an intelligent device, or can be placed in an edge server or a central server and start working after receiving the data to be recognized.

[0094] Embodiments of the present disclosure monitor the power quantity signals in the power grid, intercept the sampling values of multiple cycles before and after a leakage fault occurs for AI recognition to determine the fault, fuse multi-domain features through spatio-temporal domain joint coding, combine the conditional attention mechanism of the power grid topology to dynamically adjust the feature weights, and innovatively introduce a dynamic routing controller in the feature learning stage to switch the optimized feature extraction branch in real time according to the fault severity. Through the parameter dynamic adaptation mechanism of the extreme learning machine, the collaborative optimization of fault classification and understanding is realized. The AI model adopted in the embodiments of the present disclosure designs a dynamic attention weighting layer, encodes the power grid structure information as attention weights, makes the feature learning process conform to the propagation law of the power system, and solves the defect that traditional methods ignore topological associations. The AI model adopted in the embodiments of the present disclosure designs multiple learning branches, which can focus on the current change rate feature in case of slight leakage, focus on the harmonic distortion feature in case of severe leakage, and capture the waveform mutation feature in case of instantaneous faults. This flexible feature adaptation mechanism realizes the leap from "perceiving faults" to "understanding faults". Compared with traditional methods, the accuracy of leakage detection in complex underground working conditions of this solution is increased by 15.7%, the positioning accuracy reaches the topological node level, and the response time is shortened to within 30 ms.

[0095] The various embodiments in the present disclosure are described in a progressive manner. For the same or similar parts among the various embodiments, reference can be made to each other. Each embodiment focuses on the differences from other embodiments. In particular, for the embodiments of devices, equipment, and computer programs, since they are basically similar to the method embodiments, the description is relatively simple, and reference can be made to the description of the method embodiments for the relevant parts.

[0096] The specific embodiments of the present disclosure have been described above. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be executed in a different order than in the embodiments and still achieve the desired results. Additionally, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0097] The above are only the embodiments of the present disclosure and are not used to limit the present disclosure. For those skilled in the art, the present disclosure can have various changes and modifications. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present disclosure shall be included within the scope of the claims of the present disclosure.

Claims

1. A coal mine leakage protection method based on a leakage fault identification model, characterized in that, Including: Collect power quantity signals in each line of the power grid, where the power quantity signals include three-phase voltage, three-phase current, zero-sequence voltage, and zero-sequence current; When it is detected that the amplitude of the zero-sequence voltage of any line is greater than a preset value, intercept the sampling values of the power quantity signals for N cycles before and after the fault occurrence moment, and obtain the data to be recognized after preprocessing; Input the data to be recognized into a pre-trained leakage recognition model to obtain a leakage fault recognition result; Send a fault removal instruction to the protection unit of the corresponding line according to the leakage fault recognition result.

2. The method according to claim 1, characterized in that, The leakage fault recognition model includes: A feature extraction module for extracting multi-domain features from the data to be recognized, where the multi-domain features include time-domain waveforms, frequency-domain harmonic energy distributions, and statistical-domain mutation parameters; A spatio-temporal domain joint encoding module for fusing and encoding the multi-domain features to obtain encoded features; A dynamic routing controller and multiple feature learning branches, where the dynamic routing controller is used to activate the corresponding feature learning branch according to real-time evaluation indicators, and the feature learning branch is used to learn the encoded features.

3. The method according to claim 2, wherein The multiple feature learning branches are based on the SAE model and include: The first branch uses a convolutional neural network to extract the current change rate feature and is activated when the current change rate exceeds the first threshold; The second branch uses a frequency-domain attention network to extract the harmonic distortion feature and is activated when the total harmonic distortion rate exceeds the second threshold; The third branch uses a long short-term memory network to extract the waveform mutation feature and is activated when the mutation point density exceeds the third threshold.

4. The method according to claim 2, wherein The spatio-temporal domain joint encoding module is also used to generate conditional attention weights based on the adjacency matrix of the power grid topology and dynamically weight the encoded features.

5. The method according to claim 4, wherein The calculation of the conditional attention weights satisfies: Among them, α ij is the attention weight of node i to node j, W q is the query matrix, W k is the key matrix, A ij is an element of the power grid topology adjacency matrix, λ is the topology influence coefficient, and σ is the non-linear activation function.

6. The method according to claim 2, characterized in that The leakage recognition model further includes: A feature fusion module and an ELM classifier. The feature fusion module is used to weight and fuse the feature vectors output by the multiple feature learning branches through a dynamic gating mechanism, input them into the ELM classifier, and dynamically adjust the hidden layer activation function and regularization parameter of the ELM classifier based on the feature learning branch activated by the dynamic routing controller.

7. A coal mine leakage protection device based on a leakage fault identification model, characterized in that, Including: An acquisition module for collecting power quantity signals in each line of the power grid, where the power quantity signals include three-phase voltage, three-phase current, zero-sequence voltage, and zero-sequence current; A data acquisition module for intercepting the sampling values of the power quantity signals for N cycles before and after the fault occurrence moment when it is detected that the amplitude of the zero-sequence voltage of any line is greater than a preset value, and obtaining the data to be recognized after preprocessing; A detection module for inputting the data to be recognized into a pre-trained leakage recognition model to obtain a leakage fault recognition result; A control module for sending a fault removal instruction to the protection unit of the corresponding line according to the leakage fault recognition result.

8. A coal mine leakage protection device based on a leakage fault identification model, which is arranged in a power grid line, is characterized in that, The protection device includes a power grid interface, a data interface, a processor, and a memory, where The power grid interface is connected to the processor and is used to connect to the power grid line; The data interface is connected to the processor and is used to communicate with the management platform; The memory stores instructions executable by the processor, and the instructions are executed by the processor to implement the protection method according to any one of claims 1-6.

9. The coal mine leakage protection device according to claim 8, wherein the coal mine leakage protection device is an intelligent IoT protection terminal, a mining flameproof and intrinsically safe high-voltage vacuum distribution device or a low-voltage intelligent box, and the coal mine leakage protection device is communicatively connected to a pre-trained AI model.

10. A computer program which, when executed by a processor, causes the processor to implement the protection method according to any one of claims 1 to 6.

Images