Low-voltage power grid fault diagnosis method, system, equipment and medium

Abstract

The invention relates to the technical field of power system fault diagnosis, in particular to a low-voltage power grid fault diagnosis method, system, device and medium, and the method comprises the steps: obtaining multi-source heterogeneous data, and constructing a device-environment space-time matrix through the first preset processing of the multi-source heterogeneous data, the device-environment space-time matrix comprising a voltage sequence; performing second preset processing on the voltage sequence in the equipment-environment space-time matrix, calculating to obtain a corrosion disturbance factor, and positioning an abnormal node coordinate according to a preset threshold value; third preset processing is carried out on the corrosion disturbance factors and the abnormal node coordinates, and the contact resistance deviation rate and the corresponding fault level are determined; and adopting a multi-modal knowledge graph to match the fault level and the user repair report text, and generating a fault root cause label and a maintenance case library. The method has the beneficial effects that the problems of low precision, high false alarm rate, inaccurate fault positioning and the like when a traditional method is used for detecting the hidden salt spray corrosion fault are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of power system fault diagnosis, and in particular to a low-voltage power grid fault diagnosis method, system, equipment and medium. Background Art

[0002] As the terminal link of the power system, the stability and power supply quality of the low-voltage power grid directly determine the safety and reliability of power users. Traditional fault diagnosis methods often rely on SCADA system electrical quantity threshold judgments (such as voltage and current over-limit alarms) or regular manual inspections. These methods are inadequate for hidden corrosion faults.

[0003] In actual operation scenarios, diagnostic methods based solely on single electrical quantity characteristics, such as common voltage sag detection, can identify some obvious faults, but have limited ability to distinguish between faults caused by salt spray corrosion and normal load fluctuations, and lack quantitative assessment of environmental factors (such as salt spray concentration, temperature and humidity, etc.), resulting in a high false alarm rate and inaccurate fault location. In addition, traditional diagnostic methods fail to fully utilize user repair information and fail to effectively combine text semantics with electrical characteristics, which greatly restricts the accurate diagnosis of the root cause of the fault. The above problems greatly affect the diagnostic efficiency of salt spray corrosion faults, delay the speed of emergency repair response, and pose severe challenges to the reliability of power supply in coastal areas. Summary of the Invention

[0004] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0005] In a first aspect, the present invention provides a low-voltage power grid fault diagnosis method, comprising acquiring multi-source heterogeneous data and constructing a device-environment spatiotemporal matrix by performing a first preset processing on the multi-source heterogeneous data, wherein the device-environment spatiotemporal matrix includes a voltage sequence;

[0006] By performing a second preset processing on the voltage sequence in the equipment-environment space-time matrix, the corrosion disturbance factor is calculated and the coordinates of the abnormal node are located according to the preset threshold;

[0007] By performing a third preset processing on the corrosion disturbance factor and the abnormal node coordinates, the contact resistance deviation rate and the corresponding fault level are determined;

[0008] A multimodal knowledge graph is used to match fault levels and user repair report texts to generate fault root cause labels and a maintenance case library.

[0009] As a preferred solution of the low-voltage power grid fault diagnosis method of the present invention, the method further includes generating emergency repair work orders and prevention strategies through a multi-objective dynamic optimization algorithm based on fault levels, maintenance case database and real-time meteorological data.

[0010] As a preferred solution of the low-voltage power grid fault diagnosis method of the present invention, wherein: constructing a device-environment spatiotemporal matrix by performing a first preset processing on multi-source heterogeneous data includes:

[0011] Downsampling of electrical quantity data;

[0012] The salt spray concentration data is interpolated, and the equipment temperature data and corrosion thickness data are dynamically weighted and fused to form a multi-dimensional matrix containing spatiotemporal correlations.

[0013] As a preferred solution of the low-voltage power grid fault diagnosis method of the present invention, wherein: the voltage sequence in the equipment-environment time-space matrix is subjected to a second preset processing, and the corrosion disturbance factor is calculated to include:

[0014] By performing wavelet packet decomposition of no less than two layers on the voltage sequence, high frequency sub-band coefficients are extracted;

[0015] The corrosion disturbance factor is calculated by combining the salt spray concentration weight and the rust thickness.

[0016] As a preferred solution of the low-voltage power grid fault diagnosis method of the present invention, wherein: by performing a third preset processing on the corrosion disturbance factor and the abnormal node coordinates, the contact resistance deviation rate and the corresponding fault level are determined, including:

[0017] Calculate the actual contact resistance based on the temperature, current, salt spray concentration, and rust thickness data of the abnormal node, combined with the initial resistance and material corrosion rate;

[0018] The fault level is divided by comparing the deviation rate between the actual contact resistance and the theoretical corrosion model.

[0019] As a preferred solution of the low-voltage power grid fault diagnosis method of the present invention, before using the multimodal knowledge graph to match the fault level and the user repair report text, it includes constructing a multimodal knowledge graph;

[0020] Constructing a multimodal knowledge graph includes:

[0021] Extract keyword vectors from user repair report texts through natural language processing models;

[0022] Generate a multimodal knowledge graph by combining fault level and corrosion disturbance factor;

[0023] And match the historical maintenance case library through graph attention network.

[0024] As a preferred solution of the low-voltage power grid fault diagnosis method of the present invention, the target dynamic optimization algorithm includes:

[0025] Taking fault level, salt spray diffusion prediction and expected maintenance time as optimization objectives, the Pareto optimal solution set is solved to generate emergency repair priority and prevention strategy.

[0026] In a second aspect, the present invention provides a low-voltage power grid fault diagnosis system, comprising: a construction module for acquiring multi-source heterogeneous data and constructing a device-environment spatiotemporal matrix by performing a first preset processing on the multi-source heterogeneous data, wherein the device-environment spatiotemporal matrix includes a voltage sequence;

[0027] a calculation module, configured to calculate a corrosion disturbance factor by performing a second preset processing on a voltage sequence in a device-environment time-space matrix, and locate coordinates of abnormal nodes according to a preset threshold;

[0028] a determination module, configured to determine a contact resistance deviation rate and a corresponding fault level by performing a third preset processing on the corrosion disturbance factor and the abnormal node coordinates;

[0029] The generation module is used to match the fault level and the user's repair report text using a multimodal knowledge graph to generate fault root cause labels and a maintenance case library.

[0030] In a third aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above method when executing the computer program.

[0031] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of the above method when the computer program is executed by a processor.

[0032] Compared with the existing technology, the beneficial effects of the present invention are as follows: first, by aligning and fusing multi-source data in time and space, the equipment-environment time-space matrix is accurately constructed, laying a solid foundation for fault diagnosis; second, through wavelet packet decomposition and salt spray feature enhancement processing, the corrosion disturbance factor is effectively extracted, and the detection and positioning accuracy of abnormal nodes is significantly improved; then, combined with thermal-electric coupling inversion and corrosion model, the contact resistance deviation rate is accurately quantified to achieve accurate division of fault levels; then, using multimodal knowledge graphs, the fault level is intelligently matched with the user's repair report text to generate accurate fault root cause labels and maintenance case libraries, thereby optimizing maintenance decisions; in addition, based on a multi-objective dynamic optimization algorithm, the fault level, maintenance cases and real-time meteorological data are comprehensively considered to generate efficient emergency repair work orders and prevention strategies, which greatly improves the diagnostic efficiency and emergency repair response capability of salt spray corrosion faults in coastal low-voltage power grids, ensuring the safe and stable operation of the power grid. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0034] Figure 1 A schematic diagram of a low-voltage power grid fault diagnosis method Figure 1 .

[0035] Figure 2 A schematic diagram of a low-voltage power grid fault diagnosis method Figure 2 . DETAILED DESCRIPTION

[0036] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, the following detailed description of the specific embodiments of the present invention is given in conjunction with the accompanying drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in this field without creative work should fall within the scope of protection of the present invention.

[0037] Example 1, reference Figure 1 , which is the first embodiment of the present invention, provides a low-voltage power grid fault diagnosis method, comprising:

[0038] S100: Acquire multi-source heterogeneous data, and construct a device-environment spatiotemporal matrix by performing a first preset processing on the multi-source heterogeneous data, wherein the device-environment spatiotemporal matrix includes a voltage sequence;

[0039] S200: performing a second preset processing on the voltage sequence in the device-environment spatiotemporal matrix to calculate a corrosion disturbance factor, and locating the coordinates of the abnormal node according to a preset threshold;

[0040] S300: determining a contact resistance deviation rate and a corresponding fault level by performing a third preset processing on the corrosion disturbance factor and the abnormal node coordinates;

[0041] S400: Use multimodal knowledge graphs to match fault levels and user repair report texts to generate fault root cause labels and a repair case library.

[0042] It should be noted that due to the gradual and hidden nature of the increase in contact resistance caused by salt spray corrosion, its change process is relatively slow and not easily detected directly. At the same time, the change in contact resistance caused by salt spray corrosion is affected by the coupling of multiple factors such as salt spray concentration, ambient humidity, equipment surface condition, and current load. These factors are intertwined, increasing the complexity of fault diagnosis. In this case, the traditional method that relies on electrical quantity thresholds (such as voltage sag) or single temperature monitoring has obvious limitations in detecting hidden salt spray corrosion faults, with a high false alarm rate and difficulty in accurately quantifying contact resistance deviations and accurately classifying fault levels.

[0043] Therefore, to address the aforementioned issues, through steps S100-S400, a device-environment spatiotemporal matrix is constructed, integrating multi-source heterogeneous data to provide a spatiotemporal benchmark for fault feature extraction, effectively addressing the insufficient data fusion issues of traditional methods. Accurate extraction of corrosion disturbance factors can effectively distinguish corrosion faults from load fluctuations, significantly improving the detection accuracy and location efficiency of hidden corrosion faults. Furthermore, fault classification based on contact resistance deviation rate addresses the inaccurate fault severity assessment problem of traditional methods.

[0044] Example 2, reference Figure 2 , which is an embodiment of the present invention, provides a low-voltage power grid fault diagnosis method based on the above embodiment.

[0045] In an embodiment of the present application, the multi-source heterogeneous data includes electrical quantity data, salt spray concentration data, equipment temperature data, and corrosion thickness data of a low-voltage power grid node. Further, in step S100, a device-environment spatiotemporal matrix is constructed by performing a first preset processing on the multi-source heterogeneous data, including the following steps A1-A2:

[0046] A1: Downsample electrical quantity data;

[0047] A2: Interpolate the salt spray concentration data and dynamically weight the equipment temperature data and corrosion thickness data to form a multi-dimensional matrix that includes temporal and spatial correlations.

[0048] Specifically, the calculation formula of the device-environment space-time matrix is as follows:

[0049]

[0050] Where: M (x,y,t) For each time-space point (x, y, t), it contains data in five dimensions: voltage, current, salt spray concentration, temperature, and corrosion thickness; is the mean value of the voltage; is the mean value of the current; C′ salt is the interpolation data of salt spray concentration; T is the temperature of the equipment contact point; Rrust The thickness of rust on the contact points of the equipment.

[0051] For example, assume that there are three key nodes in a small low-voltage power grid system (with coordinates of (x1, y1), (x2, y2), and (x3, y3), and we need to perform fault diagnosis and monitoring on the system:

[0052] Data collection phase: A SCADA system is installed at each node to collect three-phase voltage and current data, recording it once per second. A salt spray sensor is installed on the surface of the equipment to record salt spray concentration data every 10 minutes. An infrared thermal imager is used to photograph the equipment contact points every 5 minutes to extract temperature data. Sensors are embedded inside the contacts to measure the oxide layer thickness data in real time.

[0053] Processing stage: For the first node (x1, y1):

[0054] In the window from time t1 to t2 (5 minutes interval), the multiple voltage and current data points collected are downsampled and the voltage mean in the window is calculated. and current mean

[0055] Interpolate the salt spray concentration data. Assume that the salt spray concentration data C is collected at time t1-5 minutes and t1. salt (t1-5) and C salt (t1), the interpolated salt spray concentration C′ corresponding to time t1 is obtained by linear interpolation salt (t1);

[0056] The temperature data T1 and the oxide layer thickness data R rust,1 Perform weighted fusion. For example, based on experience, the weight of temperature is set to 0.6, and the weight of oxide layer thickness is set to 0.4, and the weighted comprehensive data is obtained as 0.6T1+0.4R rust,1 .

[0057] The processed data are integrated into the device-environment spatiotemporal matrix. For the first node (x1, y1) at time t1, its data is

[0058] Similarly, by processing the data of other nodes and time windows, a complete device-environment spatiotemporal matrix is finally obtained for subsequent fault diagnosis and analysis.

[0059] It should be noted that this step constructs a device-environment spatiotemporal matrix, which allows intuitive observation of changes in various monitoring indicators at different locations and times, providing strong support for accurate positioning and fault diagnosis.

[0060] In an optional embodiment, the device-environment spatiotemporal matrix can also be constructed by using a Bayesian network to fuse multi-source heterogeneous data in a probabilistic form to construct the device-environment spatiotemporal matrix;

[0061] In another optional embodiment, the device-environment spatiotemporal matrix may be constructed by using a multi-layer perceptron (MLP) to extract and fuse features of multi-source heterogeneous data of the low-voltage power grid to construct the device-environment spatiotemporal matrix.

[0062] In the embodiment of the present application, step S200 performs a second preset processing on the voltage sequence in the device-environment spatiotemporal matrix to calculate the corrosion disturbance factor, including the following steps B1-B2:

[0063] B1: Extract high-frequency sub-band coefficients by performing wavelet packet decomposition of the voltage sequence at least two layers;

[0064] Preferably, the high frequency sub-band coefficients are extracted by performing six-layer wavelet packet decomposition on the voltage sequence.

[0065] It should be noted that through the analysis of 100 sets of salt spray corrosion fault signals, the amplitude of the wavelet coefficient in the 16-20kHz frequency band is significantly higher than that in normal working conditions (p<0.01). Further, when the sampling rate is 1kHz, the six-layer decomposition can cover the 0-500Hz frequency band, and the minimum sub-band bandwidth is 15.6Hz (meeting the 16-20kHz feature extraction requirements of salt spray micro-arc).

[0066] B2: Calculate the corrosion disturbance factor by combining the salt spray concentration weight and the rust thickness.

[0067] Specifically, the calculation formula of the corrosion disturbance factor is as follows:

[0068]

[0069] Where: F corr (x, y, t) is the corrosion disturbance factor; W k (t) is the wavelet coefficient of the kth subband after wavelet packet decomposition; N is the kth frequency band after wavelet packet decomposition; δ k (t) is the salt spray concentration weight; R rust (t) is the corrosion thickness of the equipment contact point; ε is a constant.

[0070] For example, taking the voltage sequence of the first node (x1, y1) as an example, multiple high-frequency sub-band coefficients are obtained after decomposition to form a wavelet coefficient matrix;

[0071] Assume that at time t1, a subband coefficient in the wavelet coefficient matrix of the first node is W1(t1) = 0.8, the salt spray concentration weight is δ1(t1) = 0.6, and the rust thickness is R rust (t1)=0.2,ε=0.1.

[0072] The corrosion disturbance factor is calculated according to the formula:

[0073]

[0074] The preset threshold is that the corrosion disturbance factor is greater than 0.35 and lasts for 3 windows (through the analysis of 100 sets of coastal pressure power grid fault data, under normal working conditions, F corr The average value is 0.12±0.05, and the salt spray corrosion failure F corr The mean value is 0.52±0.15. Furthermore, the ROC curve shows that when F corr >0.35, the true positive rate of fault detection reaches 95%);

[0075] Because F corr If (x1, y1, t1) = 1.6 is greater than 0.35 and lasts for three time windows, the node is marked as an abnormal node (indicating that the equipment is severely affected by salt spray corrosion at this time point and location) and its coordinates (x1, y1) are recorded.

[0076] It should be noted that in this step, through wavelet packet decomposition and salt spray feature enhancement, the electrical disturbance characteristics caused by salt spray corrosion can be accurately extracted, corrosion faults and load fluctuations can be effectively distinguished, and the detection accuracy and positioning efficiency of hidden corrosion faults can be significantly improved, providing a reliable basis for subsequent fault diagnosis and emergency repairs.

[0077] In an optional embodiment, the corrosion disturbance factor can also be calculated by using a deep learning model to automatically learn complex features and nonlinear relationships in the data to achieve predictive calculation of the predicted corrosion disturbance factor.

[0078] In another optional embodiment, the corrosion disturbance factor may be calculated by grouping similar equipment states and environmental conditions through cluster analysis, and then using a Markov chain model to describe the transition probability of the equipment state, thereby calculating the corrosion disturbance factor.

[0079] In the embodiment of the present application, step S300 determines the contact resistance deviation rate and the corresponding fault level by performing a third preset process on the corrosion disturbance factor and the abnormal node coordinates, including the following steps C1-C2:

[0080] C1: Calculates the actual contact resistance based on the temperature, current, salt spray concentration, and rust thickness data of the abnormal node, combined with the initial resistance and material corrosion rate;

[0081] C2: Classify the fault level by comparing the deviation rate between the actual contact resistance and the theoretical corrosion model.

[0082] Specifically, in step C1, the calculation formula for calculating the actual contact resistance is as follows:

[0083]

[0084] Where: R real is the actual contact resistance; ΔT is the infrared temperature change; η is the heat dissipation coefficient; I is the current; C′ salt is the salt spray concentration.

[0085] Specifically, in step C2, the theoretical corrosion model R contact The specific manifestations are:

[0086] R contact =R0·(1+α·C′ salt ·t 0.5 ),

[0087] Where R0 is the initial resistance value;

[0088] The calculation formula for the deviation rate between actual contact resistance and theoretical corrosion model is as follows:

[0089]

[0090] In an optional implementation, the fault level is divided into three levels according to the deviation rate value, including mild (15%≤β<30%), moderate (30%≤β<50%) and severe (β≥50%), and a fault level table is generated.

[0091] For example, assume that at time t1, the relevant data of the first node (x1, y1) is as follows:

[0092] Abnormal node data: temperature T(x1,y1,t1)=40°C, current I(t1)=20A, salt spray concentration C′ salt (t1) = 0.05 mg / m 3 and rust thickness R rust (t1) = 0.2 mm. Initial resistance R0 = 0.1 Ω, material corrosion rate α = 0.02 mm -1 The exposure time is t = 1000 h, the infrared temperature change is ΔT = 5°C, and the heat dissipation coefficient is η = 0.8.

[0093] Calculate the actual contact resistance:

[0094]

[0095] Calculate the theoretical corrosion model resistance:

[0096] R contact =0.1×(1+0.02×0.05×1000 0.5 )≈0.10326Ω

[0097] Calculate the contact resistance deviation rate:

[0098]

[0099] According to the calculation results, the contact resistance deviation rate β≈932.53% is much larger than 50%, so the fault level of this node is severe, that is, the equipment has been severely affected by salt spray corrosion and needs to be repaired or replaced immediately.

[0100] In an optional implementation, the fault levels may be divided according to Bayesian theorem, calculating the conditional probability of the device being at different fault levels, and selecting the level with the highest probability as the current fault level;

[0101] In another optional embodiment, the fault levels may be divided by using fuzzy logic to process the fuzzy relationship between the device status characteristics and the fault levels, and by defining fuzzy rules and membership functions.

[0102] In the embodiment of the present application, before using the multimodal knowledge graph to match the fault level and the user repair report text in step S400, it includes constructing the multimodal knowledge graph, wherein constructing the multimodal knowledge graph includes the following steps D1-D3:

[0103] D1: Extract keyword vectors from the user's repair report text using a natural language processing model, where the natural language processing model can be a BERT model;

[0104] D2: Generate a multimodal knowledge graph by combining the fault level and corrosion disturbance factor;

[0105] D3: Match the historical maintenance case library through graph attention network.

[0106] It should be noted that the nodes of the multimodal knowledge graph include fault types, characteristics and maintenance measures, and the edge weights are the fault characteristics and the success rate of the wrong repair measures in historical cases. When used, if the correlation degree γ>0.8, the matching maintenance case library entry C is output, including information such as fault location, maintenance measures and spare part model.

[0107] It should be noted that the BERT model is used to process the user's repair text and extract the keyword vector S text =[s1,s2,...,s n ], for example, “sparks” are mapped to arc discharges, and “flickering” is mapped to voltage sags;

[0108] The extracted keyword vector is combined with the fault level β and the corrosion disturbance factor F corr , forming multimodal input features [β,F corr ,S text ], which is input into the BERT model to construct a knowledge graph. The edge weight is the failure characteristics and the success rate of the repair measures in the historical cases, expressed as:

[0109] Edge weight = number of successful cases / total number of cases, where the number of successful cases refers to the number of cases in historical cases that were successfully repaired after a certain maintenance measure was adopted for a specific fault feature; the total number of cases refers to the number of all cases in which the maintenance measure was adopted for the fault feature.

[0110] Furthermore, we use the Graph Attention Network (GAT) to classify the multimodal features [β,F corr ,S text ] to calculate the correlation γ with the historical fault knowledge graph. The calculation formula of the correlation is as follows:

[0111] γ=Sigmoid(W·[β,F corr ,S text ] T +b)

[0112] Where W is the weight matrix, and b is the bias term. This formula multiplies the multimodal feature vector by the weight matrix, adds the bias term, and then activates it through the Sigmoid function to obtain a correlation value between 0 and 1. If γ > 0.8, the current fault condition is considered to be a good match with a maintenance case in the historical knowledge graph, and the corresponding maintenance case library entry is output.

[0113] The following is an example of generating a fault root cause label:

[0114] Assume that the repair text is "arcing discharge occurs at the equipment contacts, and there is flickering phenomenon", and the extracted keyword vector is S text = [arc discharge, flicker], further, β = 30%, F corr =1.2 and S text = [arcing, flicker], then the multimodal eigenvector is [0.3, 1.2, arcing, flicker]. Assuming that the calculated γ is greater than 0.8, and the matching maintenance case library entry is "Case 023: Replacement of rusted contacts + IP68 protective cover", the root cause label of the fault is "arcing and flickering caused by rusted contacts".

[0115] It should be noted that in this step, the fault level and user repair report text are matched through the multimodal knowledge graph, combined with the fault characteristics and historical maintenance experience, to generate the fault root cause label and maintenance case library, which can provide accurate root cause location and maintenance guidance for fault diagnosis, further improve the accuracy of fault diagnosis, shorten the maintenance time, and enhance the reliability and safety of the power grid.

[0116] In an optional implementation, the multimodal knowledge graph can also be constructed by utilizing the semantic embedding model in deep learning to map multimodal data such as text and numerical values into the same semantic space, and constructing the multimodal knowledge graph through similarity calculation;

[0117] In another optional embodiment, the construction of a multimodal knowledge graph can also utilize a graph database to store multimodal data, and construct the relationship between entities through a rule reasoning engine to form a knowledge graph.

[0118] In the embodiment of the present application, step S500 is also included to generate emergency repair work orders and prevention strategies through a multi-objective dynamic optimization algorithm based on the fault level, maintenance case library and real-time meteorological data, wherein the real-time meteorological data includes salt spray concentration, wind speed and temperature and humidity data.

[0119] In an optional embodiment, emergency repair work orders and prevention strategies are generated through a multi-objective dynamic optimization algorithm, including taking fault level, salt spray exposure attenuation coefficient and expected maintenance time as optimization objectives, solving the Pareto optimal solution set, and generating emergency repair priorities and prevention strategies.

[0120] It should be noted that the specific expression of the objective function of the multi-objective dynamic optimization algorithm is:

[0121]

[0122] Where: β i represents the fault level of the i-th device; represents the salt spray exposure attenuation coefficient of the i-th device; T i repair Indicates the estimated maintenance time of the i-th equipment. ω1, ω2, ω3 are weight coefficients, which can be adjusted according to actual needs (for example, ω1 = 0.6, ω2 = 0.3, ω3 = 0.1)

[0123] Furthermore, the constraints are:

[0124] Temperature rise risk: Equipment with temperature > 40°C must be processed within 2 hours;

[0125] Salt spray diffusion: Salt spray concentration forecast for the next 6 hours Trigger guard installed.

[0126] It should be further explained that finding the Pareto optimal solution involves determining whether the maximum number of iterations has been reached or a certain convergence accuracy has been met. When the termination criteria are met, the Pareto optimal solution is output. Furthermore, based on the optimization results, an emergency repair work order is generated, specifying the repair measures, spare part models, and handling priorities for each node. Preventive maintenance strategies are also implemented, such as shortening inspection cycles in high-risk areas and installing salt spray filters, to reduce the risk of future failures.

[0127] That is, in this step, by comprehensively considering factors such as fault level, real-time meteorological data, and maintenance cases, efficient scheduling and resource allocation of emergency repair tasks can be achieved, thereby improving emergency repair efficiency, reducing the risk of equipment overheating and fire, and ensuring the safe operation of the power grid.

[0128] In an optional embodiment, the generation of emergency repair work orders and prevention strategies can also utilize a deep reinforcement learning algorithm to learn the optimal emergency repair work order generation and prevention strategy decision-making strategy through interaction with the environment;

[0129] In another optional implementation, the generation of emergency repair work orders and prevention strategies can also adopt a fuzzy comprehensive evaluation method, taking into account multiple fuzzy factors (such as fault level, maintenance difficulty, meteorological conditions, equipment importance, etc.), and comprehensively evaluating different emergency repair and prevention plans through fuzzy operations to determine the optimal emergency repair work order and prevention strategy.

[0130] In summary, the present invention first accurately constructs the equipment-environment spatiotemporal matrix by aligning and fusing multi-source data in time and space, laying a solid foundation for fault diagnosis; secondly, through wavelet packet decomposition and salt spray feature enhancement processing, the corrosion disturbance factor is effectively extracted, significantly improving the detection and positioning accuracy of abnormal nodes; then, combining thermal-electric coupling inversion and corrosion model, the contact resistance deviation rate is accurately quantified to achieve accurate classification of fault levels; then, using multimodal knowledge graphs, the fault level is intelligently matched with the user's repair report text to generate accurate fault root cause labels and maintenance case libraries, thereby optimizing maintenance decisions; in addition, based on a multi-objective dynamic optimization algorithm, the fault level, maintenance cases and real-time meteorological data are comprehensively considered to generate efficient emergency repair work orders and prevention strategies, which greatly improves the diagnostic efficiency and emergency repair response capability of salt spray corrosion faults in coastal low-voltage power grids, ensuring the safe and stable operation of the power grid.

[0131] Example 3. The above is a schematic scheme of a low-voltage power grid fault diagnosis method. It should be noted that the technical scheme of this low-voltage power grid fault diagnosis system and the technical scheme of the low-voltage power grid fault diagnosis method described above are based on the same concept. For details not described in detail in the technical scheme of the low-voltage power grid fault diagnosis system in this embodiment, please refer to the description of the technical scheme of the low-voltage power grid fault diagnosis method described above.

[0132] This embodiment further provides a low-voltage power grid fault diagnosis system, comprising:

[0133] A construction module is used to obtain multi-source heterogeneous data and construct a device-environment spatiotemporal matrix by performing a first preset processing on the multi-source heterogeneous data, wherein the device-environment spatiotemporal matrix includes a voltage sequence;

[0134] a calculation module, configured to calculate a corrosion disturbance factor by performing a second preset processing on a voltage sequence in a device-environment time-space matrix, and locate coordinates of abnormal nodes according to a preset threshold;

[0135] a determination module, configured to determine a contact resistance deviation rate and a corresponding fault level by performing a third preset processing on the corrosion disturbance factor and the abnormal node coordinates;

[0136] The generation module is used to match the fault level and the user's repair report text using a multimodal knowledge graph to generate fault root cause labels and a maintenance case library.

[0137] This embodiment also provides an electronic device suitable for low-voltage power grid fault diagnosis, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the low-voltage power grid fault diagnosis method proposed in the above embodiment.

[0138] This embodiment further provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for diagnosing a low-voltage power grid fault as proposed in the above embodiment is implemented.

[0139] The storage medium proposed in this embodiment and the method for implementing low-voltage power grid fault diagnosis proposed in the above embodiment belong to the same inventive concept. For technical details not described in detail in this embodiment, please refer to the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0140] Through the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented with the help of software and necessary general hardware, and of course can also be implemented by hardware. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory (FLASH), hard disk or optical disk, etc., including a number of instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute the methods of various embodiments of the present invention.

[0141] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A low-voltage power grid fault diagnosis method, characterized by: include, Acquire multi-source heterogeneous data, and construct a device-environment spatiotemporal matrix by performing a first preset processing on the multi-source heterogeneous data, wherein the device-environment spatiotemporal matrix includes a voltage sequence; By performing a second preset processing on the voltage sequence in the device-environment spatiotemporal matrix, a corrosion disturbance factor is calculated, and the coordinates of the abnormal node are located according to a preset threshold; Determining a contact resistance deviation rate and a corresponding fault level by performing a third preset processing on the corrosion disturbance factor and the abnormal node coordinates; A multimodal knowledge graph is used to match the fault level and the user's repair report text to generate fault root cause labels and a maintenance case library.

2. The low-voltage power grid fault diagnosis method according to claim 1, wherein: The method further includes generating an emergency repair work order and a preventive strategy through a multi-objective dynamic optimization algorithm based on the fault level, the maintenance case library and real-time meteorological data.

3. The low-voltage power grid fault diagnosis method according to claim 2, wherein: The constructing of the device-environment spatiotemporal matrix by performing a first preset processing on the multi-source heterogeneous data includes: performing downsampling processing on the electrical quantity data; The salt spray concentration data is interpolated, and the equipment temperature data and the corrosion thickness data are dynamically weighted and fused to form a multi-dimensional matrix containing spatiotemporal correlation.

4. The low-voltage power grid fault diagnosis method according to claim 3, wherein: The second preset processing is performed on the voltage sequence in the device-environment time-space matrix to calculate the corrosion disturbance factor, including: Extracting high-frequency sub-band coefficients by performing wavelet packet decomposition of no less than two layers on the voltage sequence; The corrosion disturbance factor is calculated by combining the salt spray concentration weight and the rust thickness.

5. The low-voltage power grid fault diagnosis method according to claim 4, wherein: The contact resistance deviation rate and the corresponding fault level are determined by performing a third preset process on the corrosion disturbance factor and the abnormal node coordinates, including: Based on the temperature, current, salt spray concentration and rust thickness data of the abnormal node, combined with the initial resistance and material corrosion rate, the actual contact resistance is calculated; The fault level is divided by comparing the deviation rate between the actual contact resistance and the theoretical corrosion model.

6. The low-voltage power grid fault diagnosis method according to claim 5, wherein: Before the multimodal knowledge graph is used to match the fault level with the user's repair report text, a multimodal knowledge graph is constructed; The construction of the multimodal knowledge graph includes: Extract keyword vectors from user repair report texts through natural language processing models; Generate a multimodal knowledge graph by combining fault level and corrosion disturbance factor; And match the historical maintenance case library through graph attention network.

7. The low-voltage power grid fault diagnosis method according to claim 6, wherein: The target dynamic optimization algorithm includes: Taking fault level, salt spray diffusion prediction and expected maintenance time as optimization objectives, the Pareto optimal solution set is solved to generate emergency repair priority and prevention strategy.

8. A low-voltage power grid fault diagnosis system, applying the method according to any one of claims 1 to 7, characterized in that: include: A construction module, configured to acquire multi-source heterogeneous data and construct a device-environment spatiotemporal matrix by performing a first preset processing on the multi-source heterogeneous data, wherein the device-environment spatiotemporal matrix includes a voltage sequence; a calculation module, configured to calculate a corrosion disturbance factor by performing a second preset processing on the voltage sequence in the device-environment spatiotemporal matrix, and locate the coordinates of abnormal nodes according to a preset threshold; a determination module, configured to determine a contact resistance deviation rate and a corresponding fault level by performing a third preset processing on the corrosion disturbance factor and the abnormal node coordinates; A generation module is used to match the fault level and the user's repair report text using a multimodal knowledge graph to generate a fault root cause label and a maintenance case library.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images