Underground coal mine substation fault early warning system

Abstract

The invention relates to a fault early warning system for an underground coal mine substation. The system comprises a data acquisition module for acquiring electrical parameter data and environmental parameter data of an underground substation of a coal mine; the feature extraction module carries out preprocessing and feature extraction on the collected data to obtain signal feature parameters, then inputs the parameters into a neural network model to construct a parameter association graph, and extracts effect feature values among the parameters by using a node embedding technology. And the fault early warning module is used for aggregating information of neighborhood nodes by using a graph convolution layer based on the obtained effect characteristic value to obtain an association influence degree of each parameter, and inputting the association influence degree into a trained fault early warning model to generate a fault diagnosis report. According to the invention, the early warning and accurate diagnosis of the fault of the underground coal mine substation are realized, the fault occurrence probability is effectively reduced, the power failure and production halt loss caused by the fault is reduced, and the safety production of the underground coal mine and the life safety of workers are guaranteed.

Description

Technical Field

[0001] The present invention belongs to the field of information technology, and particularly relates to a fault warning system for underground coal mine substations. Background Art

[0002] With the development of information technology, fault warning technology for underground coal mine substations has emerged. In coal mine production, the safe and stable operation of underground substations is crucial for ensuring the smooth progress of production operations. However, the underground coal mine environment is complex and harsh. Electrical equipment not only faces threats from environmental factors such as high humidity and high gas concentration, but also is prone to failures due to abnormal electrical parameters such as voltage and current fluctuations and transformer overheating during its own operation. Once a substation fails, it will not only cause power outages and production stoppages, affecting coal mine production efficiency, but may also trigger serious safety accidents such as gas explosions and water inrushes, endangering the lives of underground workers. Traditional fault detection methods often deal with faults after they occur, and it is difficult to meet the requirements of coal mine safety production for early warning. Summary of the Invention

[0003] Based on this, it is necessary to provide a fault warning system for underground coal mine substations that can meet the requirements of coal mine safety production for early warning in view of the above technical problems.

[0004] In a first aspect, the present application provides a fault warning system for underground coal mine substations, including:

[0005] A data acquisition module for acquiring electrical parameter data and environmental parameter data of an underground coal mine substation; the electrical parameter data includes at least one of voltage, current, power factor, transformer temperature, core vibration, partial discharge amount, and oil gas parameters; the environmental parameter data includes at least one of temperature, humidity, and gas concentration.

[0006] A feature extraction module for preprocessing and feature extraction of electrical parameter data and environmental parameter data to obtain corresponding signal feature parameters; the signal feature parameters include electrical characteristics, thermodynamic characteristics, and chemical characteristic parameters; it is also used to input the obtained signal feature parameters into a neural network model to construct a parameter correlation graph, and use node embedding technology to extract the effect eigenvalue between parameters.

[0007] A fault warning module for aggregating the information of neighboring nodes based on the effect eigenvalue using a graph convolutional layer to obtain the correlation influence degree of each parameter; it is also used to input the correlation influence degree into a trained fault warning model to generate a fault diagnosis report; the fault diagnosis report includes a fault location code, a type identifier, a time prediction interval, and a probability confidence level.

[0008] In one embodiment, preprocessing and feature extraction of electrical parameter data and environmental parameter data to obtain corresponding signal feature parameters includes:

[0009] Perform preliminary timestamp alignment and data association operations on the electrical parameter data and environmental parameter data to obtain a preliminary integrated parameter data set.

[0010] Use the moving average filtering algorithm to clean the electrical parameter data in the parameter data set and remove noise interference to obtain smoothed electrical parameters.

[0011] Process the missing values in the environmental parameter data in the parameter data set, and use the linear regression method to estimate and fill in the missing data to obtain complete environmental parameter data.

[0012] According to the complete environmental parameter data, use the density-based spatial clustering algorithm to identify and mark the abnormal environmental parameter values, and use the locally weighted regression method to correct the abnormal environmental parameter values to obtain the corrected environmental parameters.

[0013] Use the fast Fourier transform method to extract features from the smoothed electrical parameters to obtain electrical characteristic parameters; the electrical characteristic parameters include voltage amplitude, current effective value, power factor, harmonic content, voltage fluctuation coefficient, current change rate, and zero-sequence current value.

[0014] Use the machine learning algorithm to extract features from the electrical parameters and environmental parameters to obtain thermodynamic characteristic parameters; the thermodynamic characteristic parameters include temperature, temperature change rate, temperature gradient, heat flux density, and heat capacity.

[0015] Use the association rule mining algorithm to extract features from the electrical parameters and environmental parameters to obtain chemical characteristic parameters; the chemical characteristic parameters include gas concentration, oxygen content, carbon monoxide concentration, and the influence coefficient of humidity on chemical substance reactions.

[0016] Based on the electrical characteristic parameters, thermodynamic characteristic parameters, and chemical characteristic parameters, perform fusion to obtain corresponding signal characteristic parameters.

[0017] In one embodiment, input the obtained signal characteristic parameters into the neural network model to construct a parameter association graph, and use the node embedding technology to extract the effect characteristic values between the parameters, including:

[0018] Input the signal characteristic parameters into the neural network model to obtain characteristic weight values.

[0019] Construct according to the obtained characteristic weight values to generate a parameter association graph of the signal characteristic parameters.

[0020] Use the graph structure analysis algorithm to extract vectors from the parameter association graph to obtain node embedding vectors.

[0021] Calculate using a formula based on the node embedding vectors to obtain the effect eigenvalue between parameters.

[0022] Judge the effect eigenvalue based on a preset threshold. If the effect eigenvalue is greater than the preset threshold, it is determined that there is a significant correlation between the parameters, and the feature weight value of the neural network model is updated according to the significant correlation.

[0023] Generate updated effect eigenvalues corresponding to the signal feature parameters according to the updated neural network model.

[0024] In one embodiment, calculating using a formula based on the node embedding vectors to obtain the effect eigenvalue between parameters includes:

[0025] Use the following formula to calculate the node embedding vectors to obtain the effect eigenvalue:

[0026]

[0027] Among them, γ represents the effect eigenvalue, d represents the vector dimension, ω i represents the weight coefficient of the i-th dimension, e i represents the embedding value of the i-th dimension, and ReLU represents the activation function.

[0028] In one embodiment, aggregating the information of neighboring nodes based on the effect eigenvalue using a graph convolutional layer to obtain the correlation influence degree of each parameter includes:

[0029] Construct a parameter correlation matrix based on the effect eigenvalue, and calculate the initial influence degree corresponding to each parameter.

[0030] Construct a graph structure according to the initial influence degree to obtain the neighborhood relationship between neighboring nodes.

[0031] Aggregate the feature information of neighboring nodes based on the neighborhood relationship using a graph convolutional algorithm to obtain the node influence value.

[0032] Judge the node influence value based on a preset threshold. If the node influence value exceeds the preset threshold, adjust the global influence distribution based on the preset threshold to obtain the correlation influence degree.

[0033] Optimize the global influence distribution using a machine learning algorithm to obtain the updated correlation influence degree.

[0034] In one embodiment, the fault warning module further includes:

[0035] Calculate based on the correlation influence degree parameter combined with a time decay factor to obtain a dynamic influence value.

[0036] Input the dynamic influence value into the trained fault warning model and use the fault feature matching rule to generate a candidate fault set.

[0037] Filter according to the candidate fault set using the probability weight threshold to obtain the target fault type.

[0038] Call the fault location mapping library according to the target fault type to generate a predicted fault diagnosis report; the fault diagnosis report includes a fault location code, a type identifier, a time prediction interval, and a probability confidence level.

[0039] When the probability confidence level in the fault diagnosis report exceeds the preset threshold or a fault may occur within the time prediction interval, an early warning message is sent through an audible and visual alarm, and the early warning message is sent to the terminal device of the management personnel.

[0040] In one embodiment, a dynamic influence value is calculated based on the correlation influence degree parameter combined with the time decay factor, including:

[0041] The following formula is used to calculate the correlation influence degree parameter to obtain the dynamic influence value:

[0042]

[0043] Among them, I(t) represents the dynamic influence value, γ represents the influence intensity coefficient, P k represents each correlation influence degree parameter value, m represents the number of parameters, λ represents the time decay rate, and t represents the time span.

[0044] In a second aspect, the present application also provides a method for fault early warning in a coal mine underground substation, and the method includes:

[0045] Obtain the electrical parameter data and environmental parameter data of the coal mine underground substation; the electrical parameter data includes at least one of voltage, current, power factor, transformer temperature, core vibration, partial discharge amount, and oil gas parameters; the environmental parameter data includes at least one of temperature, humidity, and gas concentration.

[0046] Preprocess and extract features from the electrical parameter data and environmental parameter data to obtain corresponding signal feature parameters; the signal feature parameters include electrical characteristics, thermodynamic characteristics, and chemical characteristic parameters; input the obtained signal feature parameters into the neural network model to construct a parameter correlation graph, and use the node embedding technology to extract the effect eigenvalue between the parameters.

[0047] Based on the effect eigenvalue, use the graph convolutional layer to aggregate the information of neighboring nodes to obtain the correlation influence degree of each parameter; input the correlation influence degree into the trained fault early warning model to generate a fault diagnosis report; the fault diagnosis report includes a fault location code, a type identifier, a time prediction interval, and a probability confidence level.

[0048] In a third aspect, the present application further provides a computer device, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it is like the previous system and method.

[0049] In a fourth aspect, the present application further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it is like the previous system and method.

[0050] For the above-mentioned coal mine underground substation fault warning system, the data acquisition module obtains electrical parameter data of the coal mine underground substation, such as at least one of voltage, current, power factor, transformer temperature, core vibration, partial discharge amount, and gas parameters in oil, as well as environmental parameter data, including at least one of temperature, humidity, and gas concentration. The feature extraction module then preprocesses and extracts features from the collected electrical parameter data and environmental parameter data to obtain signal feature parameters covering electrical characteristics, thermodynamic characteristics, and chemical characteristics. Then, the signal feature parameters are input into the neural network model to construct a parameter correlation graph, and the node embedding technology is used to extract the effect eigenvalue between parameters. The fault warning module, based on the obtained effect eigenvalue, uses the graph convolutional layer to aggregate the information of neighboring nodes to obtain the correlation influence degree of each parameter, and then inputs the correlation influence degree into the trained fault warning model to generate a fault diagnosis report including fault location coding, type identifier, time prediction interval, and probability confidence. It realizes the early warning and accurate diagnosis of faults in the coal mine underground substation, effectively reduces the probability of faults, reduces the power outage and production stoppage losses caused by faults, and ensures the safe production of the coal mine underground and the life safety of the staff. Description of the Drawings

[0051] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following will briefly introduce the drawings required for use in the description of the embodiments or related technologies. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0052] Figure 1 It is a structural block diagram of a coal mine underground substation fault warning system provided by an embodiment of the present invention;

[0053] Figure 2 It is a flowchart for preprocessing and extracting features from electrical parameter data and environmental parameter data to obtain corresponding signal feature parameters provided by an embodiment of the present invention;

[0054] Figure 3A flowchart provided by an embodiment of the present invention for inputting the obtained signal feature parameters into a neural network model to construct a parameter correlation graph and using node embedding technology to extract the effect eigenvalues between parameters;

[0055] Figure 4 A flowchart provided by an embodiment of the present invention for calculating the correlation influence degree of each parameter according to the network effect feature and using a graph convolutional layer to aggregate the information of neighboring nodes to obtain the corresponding global influence weight;

[0056] Figure 5 A flowchart of a fault warning method for a coal mine underground substation provided by an embodiment of the present invention. Detailed implementation manners

[0057] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0058] First, the implementation environment of the embodiments of the present application is described. Exemplarily, the implementation environment includes a data acquisition device, a computing and processing device, and an intelligent display device.

[0059] In a coal mine underground substation fault warning system, a data acquisition device, such as various sensors, is responsible for collecting key data such as current, voltage, and temperature of the equipment operation in the underground substation, and transmitting the data to the computing and processing device through a communication bus. The computing and processing device receives the raw data from the data acquisition device, performs in-depth analysis and processing using complex algorithms and models, and extracts the equipment operation status information and potential fault risk data. Subsequently, the processed results are transmitted to the intelligent display device through a high-speed communication network such as an industrial Ethernet. The intelligent display device, mainly composed of a high-resolution liquid crystal display screen or a touch operation screen, is responsible for intuitively presenting the real-time data collected by the data acquisition device, the equipment operation status analyzed by the computing and processing device, and the fault warning information, etc., providing a visual operation interface for the staff to facilitate their real-time monitoring and management of the operation status of the underground substation equipment.

[0060] The data acquisition device obtains key operation data, which covers various types of sensors, such as current sensors, voltage sensors, temperature sensors, and humidity sensors, etc., and can real-time sense the changes of electrical parameters and environmental status information during the equipment operation. The data acquisition device can quickly convert the collected analog signals into digital signals, and transmit the data to the subsequent computing and processing device efficiently according to a specific communication protocol through a communication interface, providing original and accurate data support for the entire fault warning system.

[0061] The computing and processing device conducts in-depth analysis and processing on the massive collected data, and has excellent computing power and data storage capabilities. After receiving the raw data from the data collection device, the computing and processing device uses complex algorithm models, including but not limited to data analysis algorithms, fault prediction models, etc., to conduct all-round mining and analysis of the data. Through real-time computing and logical judgment of the data, it can accurately identify the operating status of the device, quickly capture potential fault hazards, and generate corresponding processing results and warning information, providing key data that can be intuitively displayed for the intelligent display device to help the staff timely grasp the operating status of the substation equipment.

[0062] The intelligent display device generally uses a high-resolution liquid crystal display screen or an advanced touch operation screen, and has clear display effects and convenient operation performance. It is connected to the computing and processing device through a high-speed communication link such as an industrial Ethernet, and can receive and present various data information transmitted by the computing and processing device in real time. It can not only intuitively display the real-time operating parameters of the underground substation equipment, such as current, voltage, power, etc., but also present the operating status trend and fault warning information of the equipment in intuitive forms such as charts and graphs. With the help of the intelligent display device, the staff can conveniently obtain system feedback, quickly understand the overall operating situation of the substation, and then make efficient decisions to achieve precise monitoring and management of the underground substation.

[0063] Combined with the above implementation environment, the application scenarios of the embodiments of this application are described.

[0064] The coal mine underground substation fault warning system provided by the embodiments of this application realizes the safe operation of the coal mine underground substation through all-round data collection, in-depth computing and processing, and intuitive intelligent display. The data collection device uses various sensors to accurately sense information such as current, voltage, temperature, and ambient humidity of key equipment such as high- and low-voltage switch cabinets and transformers, and transmits it to the computing and processing device through a communication interface. The computing and processing device uses data analysis algorithms and fault prediction models to identify the operating status of the device and potential fault hazards. The analysis results are transmitted to the intelligent display device in real time to intuitively present real-time operating parameters, fault hazards, and emergency fault information, helping the staff timely grasp the situation of the substation, effectively ensuring the safe and stable operation of the underground substation, reducing the fault risk, and minimizing the impact on coal mine production. Exemplarily, the coal mine underground substation fault warning system provided by the embodiments of this application can be applied to at least one of the following scenarios including but not limited to the following.

[0065] First, the fault warning system for the underground coal mine substation is applied to the daily operation monitoring scenario. During the daily operation of the underground coal mine substation, various sensors are evenly distributed at key equipment such as high- and low-voltage switchgears and transformers in the substation, continuously sensing information such as current fluctuations, voltage changes, temperature rises and falls, and environmental humidity fluctuations during equipment operation. The real-time collected data is quickly transmitted to the computing and processing device via the communication interface following the established communication protocol. The computing and processing device then deeply analyzes the massive amount of raw data, and uses data analysis algorithms to accurately sort out the normal parameter range and real-time status data of equipment operation. The analysis results are transmitted to the intelligent display device in real time, and the real-time operation parameters of the substation equipment are presented in the form of intuitive numbers and charts. Through the intelligent display device, the staff can comprehensively and clearly master the daily operation status of the substation, and realize the normalized monitoring of the equipment operation status.

[0066] Second, the fault warning system for the underground coal mine substation is applied to the fault and hidden danger investigation scenario. When there may be potential fault and hidden dangers in the underground coal mine substation, the highly sensitive sensors of the data acquisition device play a key role. Once there are slight abnormalities in the equipment operation parameters, such as the current instantaneously fluctuating beyond the normal range or the temperature showing an abnormal rising trend, the sensors quickly capture the changes and efficiently transmit the abnormal data to the computing and processing device. The computing and processing device starts a complex fault prediction model, conducts a comprehensive analysis of the abnormal data, and combines the equipment historical operation data with the preset fault feature library to accurately locate the potential fault points and fault types. The obtained fault and hidden danger information is immediately transmitted to the intelligent display device, and the potential fault location, possible reasons, etc. are presented in eye-catching colors, flashing icons and detailed text descriptions. Based on the information presented by the intelligent display device, the staff can quickly and accurately carry out the work of investigating faults and hidden dangers, and take effective measures in time to eliminate the hidden dangers to ensure the safe and stable operation of the substation.

[0067] In one of the embodiments, as Figure 1 shown, the present application provides a fault warning system for an underground coal mine substation, which may include:

[0068] A data acquisition module 101, configured to obtain electrical parameter data and environmental parameter data of the underground coal mine substation; the electrical parameter data includes at least one of voltage, current, power factor, transformer temperature, core vibration, partial discharge amount, and oil gas parameters; the environmental parameter data includes at least one of temperature, humidity, and gas concentration.

[0069] Specifically, in terms of electrical parameter data, voltage and current, as key indicators reflecting the basic operating state of the power system, are of great significance for judging whether there are faults such as overload and short circuit in equipment; the power factor is related to the efficient utilization efficiency of electric energy, and abnormal values may imply poor equipment operating conditions or reactive power compensation problems. Monitoring the transformer temperature can timely detect the overheating hidden danger inside the transformer caused by reasons such as too high load or poor heat dissipation. The detection of core vibration and partial discharge amount is an important basis for evaluating the insulation performance and internal structure integrity of the transformer. Tiny changes may indicate potential fault risks. For the gas parameters in oil, such as the content changes of gases like hydrogen, methane, and acetylene, they can be monitored through technical means such as gas chromatography analysis. Abnormal concentrations of different gases are often closely related to different types of faults inside the transformer. In the field of environmental parameter data, temperature and humidity have a significant impact on the insulation performance of underground substation equipment. Too high temperature or humidity may cause problems such as equipment short circuit and corrosion; the monitoring of gas concentration is a key link to ensure safe production. Once the gas accumulates to a certain concentration, it is extremely easy to trigger serious safety accidents such as explosions.

[0070] The feature extraction module 102 is used to preprocess and extract features from the electrical parameter data and environmental parameter data to obtain corresponding signal feature parameters; the signal feature parameters include electrical characteristics, thermodynamic characteristics, and chemical characteristic parameters; it is also used to input the obtained signal feature parameters into the neural network model to construct a parameter correlation graph, and use node embedding technology to extract the effect eigenvalue between parameters.

[0071] First, through preprocessing algorithms, such as filtering for noise reduction, normalization, etc., the original data is cleaned and regularized to eliminate noise interference and dimensional differences in the data, ensuring the accuracy and consistency of the data. Subsequently, professional feature extraction techniques are used to mine corresponding signal feature parameters for different types of data. In terms of electrical characteristics, the fluctuation laws, harmonic contents, etc. of electrical parameters such as voltage and current are deeply analyzed. These features can intuitively reflect the operating stability of the power system and whether there are abnormal operating conditions of the equipment. For the extraction of thermodynamic characteristic parameters, around data such as transformer temperature, the change trend over time, temperature gradient distribution, etc. are explored, providing key bases for judging the heat dissipation status and thermal stability of the equipment. At the level of chemical characteristic parameters, for gas parameters in oil, through complex analysis means, features such as the proportion change of different gas components and gas generation rate are refined. After the signal feature parameters are extracted, the parameters are further input into the constructed neural network model. With the powerful learning and mapping capabilities of the neural network, a comprehensive and detailed parameter correlation graph is constructed, which can clearly present the complex non-linear relationships between various parameters. On this basis, using cutting-edge node embedding technology, the nodes in the parameter correlation graph are deeply analyzed to accurately extract the effect eigenvalue between parameters. The effect eigenvalue contains rich information, which can not only reveal the comprehensive impact of the interaction between different parameters on the equipment operating state, but also help the system more efficiently and accurately identify potential fault risks, ensuring the safe and stable operation of the underground coal mine substation.

[0072] The fault warning module 103 is used to aggregate the information of neighboring nodes based on the effect eigenvalue using a graph convolutional layer to obtain the correlation influence degree of each parameter; it is also used to input the correlation influence degree into the trained fault warning model to generate a fault diagnosis report; the fault diagnosis report includes a fault location code, a type identifier, a time prediction interval, and a probability confidence level.

[0073] Specifically, based on the effect eigenvalue, the information of neighboring nodes is effectively aggregated through a graph convolutional layer to calculate the correlation influence degree of each parameter, clearly revealing the tightness of the interaction between parameters. Subsequently, the correlation influence degree data is input into the trained fault warning model to generate a comprehensive and accurate fault diagnosis report. The report content includes a fault location code to clarify the specific location of the fault in the substation system; a type identifier to accurately define the type of the fault; a time prediction interval to estimate the time period when the fault may occur; and a probability confidence level to quantify the reliability of the fault occurrence prediction, providing key information for maintenance personnel to handle faults in a timely manner and ensuring the safe and stable operation of the substation.

[0074] The above-mentioned fault early warning system for coal mine underground substations. The data acquisition module obtains electrical parameter data of the coal mine underground substation, such as at least one of voltage, current, power factor, transformer temperature, core vibration, partial discharge quantity, and oil gas parameters, as well as environmental parameter data, including at least one of temperature, humidity, and gas concentration. The feature extraction module then preprocesses and extracts features from the collected electrical parameter data and environmental parameter data to obtain signal feature parameters covering electrical characteristics, thermodynamic characteristics, and chemical characteristics. Then, the signal feature parameters are input into the neural network model to construct a parameter correlation graph, and the node embedding technology is used to extract the effect eigenvalue between parameters. The fault early warning module, based on the obtained effect eigenvalue, uses the graph convolutional layer to aggregate the information of neighboring nodes to obtain the correlation influence degree of each parameter, and then inputs the correlation influence degree into the trained fault early warning model to generate a fault diagnosis report containing fault location coding, type identifier, time prediction interval, and probability confidence level. It realizes the early warning and accurate diagnosis of faults in coal mine underground substations, effectively reduces the probability of faults, reduces the losses caused by power outages and production suspensions due to faults, and ensures the safe production of coal mine underground and the lives of workers.

[0075] In one embodiment, as Figure 2 shown, preprocessing and feature extraction of the electrical parameter data and environmental parameter data to obtain the corresponding signal feature parameters may include the following steps:

[0076] Step S201, perform preliminary timestamp alignment and data association operations on the electrical parameter data and environmental parameter data to obtain a preliminary integrated parameter data set.

[0077] Step S202, use the moving average filtering algorithm to clean the electrical parameter data in the parameter data set and remove noise interference to obtain smoothed electrical parameters.

[0078] Step S203, perform missing value processing on the environmental parameter data in the parameter data set, and use the linear regression method to estimate and fill in the missing data to obtain complete environmental parameter data.

[0079] Step S204, identify and mark abnormal environmental parameter values according to the complete environmental parameter data using the density-based spatial clustering algorithm, and use the locally weighted regression method to correct the abnormal environmental parameter values to obtain corrected environmental parameters.

[0080] Step S205, use the fast Fourier transform method to extract features from the smoothed electrical parameters to obtain electrical characteristic parameters; the electrical characteristic parameters include voltage amplitude, current effective value, power factor, harmonic content, voltage fluctuation coefficient, current change rate, and zero-sequence current value.

[0081] Step S206, using a machine learning algorithm to extract features from electrical parameters and environmental parameters to obtain thermodynamic characteristic parameters; the thermodynamic characteristic parameters include temperature, temperature change rate, temperature gradient, heat flux density, and heat capacity.

[0082] Step S207, using an association rule mining algorithm to extract features from electrical parameters and environmental parameters to obtain chemical characteristic parameters; the chemical characteristic parameters include gas concentration, oxygen content, carbon monoxide concentration, and the influence coefficient of humidity on chemical substance reactions.

[0083] Step S208, based on the electrical characteristic parameters, thermodynamic characteristic parameters, and chemical characteristic parameters, perform fusion to obtain corresponding signal characteristic parameters.

[0084] First, perform preliminary timestamp alignment and data association operations on the electrical parameter data and environmental parameter data. Through precise matching of timestamps and sorting out the logical relationships between data, obtain a preliminary integrated parameter data set. Then, use the moving average filtering algorithm to carry out data cleaning work on the electrical parameter data in the parameter data set, effectively removing noise interference, so as to obtain smoothed electrical parameters and improve the reliability of electrical data. Subsequently, for the environmental parameter data in the parameter data set, perform missing value processing, and use the linear regression method to scientifically estimate and fill in the missing data. Based on the complete environmental parameter data, use the density-based spatial clustering algorithm to identify and mark abnormal environmental parameter values, and then use the locally weighted regression method to correct the abnormal values to obtain corrected environmental parameters to ensure the accuracy of environmental data. Use the fast Fourier transform method to extract features from the smoothed electrical parameters, and obtain electrical characteristic parameters including voltage amplitude, current effective value, power factor, etc. Use a machine learning algorithm to extract thermodynamic characteristic parameters such as temperature and temperature change rate from electrical parameters and environmental parameters. Use the association rule mining algorithm to analyze electrical parameters and environmental parameters to obtain chemical characteristic parameters such as gas concentration and oxygen content. Finally, fuse the electrical characteristic parameters, thermodynamic characteristic parameters, and chemical characteristic parameters to obtain corresponding signal characteristic parameters.

[0085] In this embodiment, after operations such as timestamp alignment, cleaning, missing value processing, and outlier correction on each parameter data, the data quality is greatly improved, providing a reliable data basis for subsequent feature extraction and fault warning analysis. Using a variety of advanced algorithms for feature extraction, deeply mining electrical, thermodynamic, and chemical characteristic parameters from different dimensions and fusing them into signal characteristic parameters can more comprehensively and accurately reflect the operating state of the coal mine underground substation, greatly improving the accuracy and reliability of the fault warning system, helping the operation and maintenance personnel to detect potential fault risks in a timely manner and take effective measures to ensure the safe and stable operation of the substation.

[0086] In one of the embodiments, such asFigure 3 As shown, input the obtained signal feature parameters into the neural network model to construct a parameter correlation graph, and use node embedding technology to extract the effect eigenvalue between parameters, which may include the following steps:

[0087] Step S301: Input the signal feature parameters into the neural network model to obtain feature weight values.

[0088] Step S302: Construct according to the obtained feature weight values to generate a parameter correlation graph of the signal feature parameters.

[0089] Step S303: Use the graph structure analysis algorithm to extract vectors from the parameter correlation graph to obtain node embedding vectors.

[0090] Step S304: Calculate based on the node embedding vectors using a formula to obtain the effect eigenvalue between parameters.

[0091] Step S305: Judge the effect eigenvalue based on a preset threshold. If the effect eigenvalue is greater than the preset threshold, it is determined that there is a significant correlation between parameters, and based on the significant correlation, update the feature weight value of the neural network model.

[0092] Step S306: Generate the corresponding updated effect eigenvalue of the signal feature parameter according to the updated neural network model.

[0093] Input the signal feature parameters obtained by preprocessing and fusion into the neural network model, and then output the feature weight values. Immediately, based on the obtained feature weight values, generate a parameter correlation graph that can clearly present the complex relationship between the signal feature parameters. This graph shows the connection of each parameter in an intuitive form. Subsequently, use the graph structure analysis algorithm to perform vector extraction operations on the parameter correlation graph to obtain node embedding vectors from it. Based on the node embedding vectors, use a specific formula to calculate the effect eigenvalue between parameters. Compare the obtained effect eigenvalue with the preset threshold. If the effect eigenvalue is greater than the preset threshold, it is determined that there is a significant correlation between parameters. At this time, based on this significant correlation, update the feature weight value of the neural network model to make the model better adapt to the data characteristics. Finally, use the updated neural network model to generate the corresponding effect eigenvalue of the signal feature parameter again to achieve the optimization of the effect eigenvalue.

[0094] On the one hand, starting from signal characteristic parameters, gradually explore the complex correlations between parameters. The obtained effect characteristic values can more accurately reflect the interaction of parameters, providing a key basis for fault warning. On the other hand, through preset threshold judgment and the update of the characteristic weight values of the neural network model, the self-optimization and adaptation of the model are realized, which can continuously improve the analysis ability of the complex operating state of the underground coal mine substation, significantly improve the accuracy and reliability of fault warning, and help the operation and maintenance personnel to timely discover potential fault risks.

[0095] In one embodiment, based on the node embedding vector, calculate using a formula to obtain the effect characteristic values between parameters, which may include the following steps:

[0096] Use the following formula to calculate the node embedding vector to obtain the effect characteristic value:

[0097]

[0098] where Υ represents the effect characteristic value, d represents the vector dimension, ω i represents the weight coefficient of the i-th dimension, e i represents the embedding value of the i-th dimension, and ReLU represents the activation function.

[0099] Through the operation of this formula, the parameter correlation information contained in the node embedding vector can be fully explored, and the complex vector data can be converted into intuitive and quantifiable effect characteristic values. With this quantitative calculation method, the weak but key correlation changes between parameters can be more sensitively captured, which helps to timely detect potential abnormal situations during the operation of the underground coal mine substation, and greatly improves the sensitivity and accuracy of the fault warning system.

[0100] In one embodiment, based on the effect characteristic value, use the graph convolutional layer to aggregate the information of neighboring nodes to obtain the correlation influence degree of each parameter, which may include the following steps:

[0101] Step S401, construct a parameter correlation matrix based on the effect characteristic value, and calculate the initial influence degree corresponding to each parameter.

[0102] Step S402, construct a graph structure according to the initial influence degree to obtain the neighborhood relationship between neighboring nodes.

[0103] Step S403, based on the neighborhood relationship, use the graph convolutional algorithm to aggregate the characteristic information of neighboring nodes to obtain the node influence value.

[0104] Step S404, judge the node influence value based on a preset threshold. If the node influence value exceeds the preset threshold, adjust the global influence distribution based on the preset threshold to obtain the correlation influence degree.

[0105] Step S405: Optimize the global influence distribution using a machine learning algorithm to obtain an updated associated influence degree.

[0106] The system constructs a parameter association matrix based on the effect eigenvalue and calculates the initial influence degree corresponding to each parameter. Immediately afterwards, a graph structure is constructed using the obtained initial influence degree to show the neighborhood relationship between neighborhood nodes, presenting the abstract connection between parameters in an intuitive and visual form. Subsequently, based on the constructed neighborhood relationship, a graph convolutional algorithm is used to efficiently aggregate the feature information of neighborhood nodes, and finally, a node influence value that can reflect the comprehensive influence of nodes is obtained. The node influence value is compared with a preset threshold. If the node influence value exceeds the preset threshold, the global influence distribution is adjusted according to the preset threshold to obtain a more practical associated influence degree. Finally, the global influence distribution is further optimized using a machine learning algorithm to update the associated influence degree.

[0107] In one embodiment, as Figure 4 shown, the fault warning module may further include:

[0108] Step S501: Calculate based on the associated influence degree parameter combined with a time decay factor to obtain a dynamic influence value.

[0109] Step S502: Input the dynamic influence value into the trained fault warning model and use the fault feature matching rule to generate a candidate fault set.

[0110] Step S503: Screen according to the candidate fault set using a probability weight threshold to obtain the target fault type.

[0111] Step S504: Call the fault location mapping library according to the target fault type to generate a predicted fault diagnosis report; the fault diagnosis report includes a fault location code, a type identifier, a time prediction interval, and a probability confidence level.

[0112] Step S505: When the probability confidence level in the fault diagnosis report exceeds the preset threshold or a fault may occur within the time prediction interval, an early warning message is sent through an audible and visual alarm and the early warning message is sent to the terminal device of the management personnel.

[0113] First, based on the correlation influence degree parameter, combined with the time decay factor, calculations are carried out to obtain the dynamic influence value that can reflect the real-time dynamic influence of the parameter. Immediately afterwards, the dynamic influence value is input into the trained fault warning model. The model conducts in-depth analysis and comparison of the input data according to the pre-set fault feature matching rules, and generates a candidate fault set covering various possible fault situations. Subsequently, based on the candidate fault set, screening is carried out using the probability weight threshold to accurately identify the most likely target fault type from numerous candidates. After determining the target fault type, the system calls the fault location mapping library, combines the relevant information of the target fault type, and generates a detailed predicted fault diagnosis report. This report contains accurate fault location codes, type identifiers, time prediction intervals, and probability confidence levels. Finally, when the probability confidence level in the fault diagnosis report exceeds the preset threshold, or it is judged that a fault may occur within the time prediction interval, the system immediately issues a prominent warning message through an audible and visual alarm, and at the same time quickly sends this warning message to the terminal device of the management personnel.

[0114] By calculating the dynamic influence value in combination with the time decay factor, and using operations such as fault feature matching and probability weight threshold screening, accurate prediction and positioning of faults are realized, greatly improving the accuracy of fault warning. Generating a detailed fault diagnosis report and promptly issuing a warning message enables the management personnel to obtain key information in a timely manner, make decisions quickly and take effective measures, thereby minimizing the impact of faults on underground coal mine production, effectively ensuring the safe and stable operation of the substation, and enhancing the reliability and safety of the entire coal mine production system.

[0115] In one embodiment, calculating the dynamic influence value based on the correlation influence degree parameter in combination with the time decay factor may include the following steps:

[0116] The following formula is used to calculate the correlation influence degree parameter to obtain the dynamic influence value:

[0117]

[0118] where I(t) represents the dynamic influence value, γ represents the influence intensity coefficient, P k represents each correlation influence degree parameter value, m represents the number of parameters, λ represents the time decay rate, and t represents the time span.

[0119] Calculate the dynamic influence value based on the correlation influence degree parameter combined with the time decay factor. This formula introduces the time decay factor, fully considering the actual situation that the correlation influence degree will change with time, making the calculation of the dynamic influence value more in line with the complex and changeable actual operation scenario of the underground coal mine substation, avoiding evaluation deviation caused by ignoring the time factor, and thus significantly improving the accuracy of the fault risk assessment. By obtaining a more accurate dynamic influence value, the fault warning system can identify potential faults more timely and effectively, provide more valuable reference information for the operation and maintenance personnel, help formulate countermeasures in advance, take targeted measures, and then ensure the safe and stable operation of the underground coal mine substation, reducing production losses and safety accidents caused by faults.

[0120] In the second aspect, the present application also provides a fault warning method for an underground coal mine substation, as Figure 5 shown. The method may include:

[0121] Step S601, obtain the electrical parameter data and environmental parameter data of the underground coal mine substation; the electrical parameter data includes at least one of voltage, current, power factor, transformer temperature, core vibration, partial discharge amount, and gas parameters in oil; the environmental parameter data includes at least one of temperature, humidity, and gas concentration.

[0122] Step S602, preprocess and extract features from the electrical parameter data and environmental parameter data to obtain corresponding signal feature parameters; the signal feature parameters include electrical characteristics, thermodynamic characteristics, and chemical characteristic parameters; input the obtained signal feature parameters into the neural network model to construct a parameter correlation graph, and use node embedding technology to extract the effect feature values between the parameters.

[0123] Step S603, based on the effect feature value, use the graph convolutional layer to aggregate the information of neighboring nodes to obtain the correlation influence degree of each parameter; input the correlation influence degree into the trained fault warning model to generate a fault diagnosis report; the fault diagnosis report includes a fault location code, a type identifier, a time prediction interval, and a probability confidence level.

[0124] The above-mentioned method for fault early warning of a coal mine underground substation first obtains the electrical parameter data and environmental parameter data of the coal mine underground substation. Among them, the electrical parameter data covers at least one of voltage, current, power factor, transformer temperature, core vibration, partial discharge quantity, and gas parameters in oil, and the environmental parameter data at least includes temperature, humidity, and gas concentration. Subsequently, preprocessing and feature extraction operations are performed on the collected electrical parameter data and environmental parameter data to obtain corresponding signal feature parameters, including electrical characteristics, thermodynamic characteristics, and chemical characteristic parameters. Then, the signal feature parameters are input into the neural network model, a parameter correlation graph is constructed through the model, and the effect eigenvalue between the parameters is deeply mined using node embedding technology. Finally, based on the obtained effect eigenvalue, the graph convolution layer is used to aggregate the neighborhood node information, and then the correlation influence degree of each parameter is calculated. The correlation influence degree is input into the trained fault early warning model to quickly generate a fault diagnosis report, which covers the fault location code, type identifier, time prediction interval, and probability confidence level. It realizes the early warning and accurate diagnosis of faults in the coal mine underground substation, effectively reduces the probability of faults, reduces the power outage and production stoppage losses caused by faults, and ensures the safe production of the coal mine underground and the lives of the staff.

[0125] It should be understood that although the steps in the flowcharts involved in the above-mentioned embodiments are shown in sequence according to the arrows, these steps do not necessarily need to be executed in the order indicated by the arrows. Unless there is a clear indication in this article, the execution of these steps has no strict order limit, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments may include multiple steps or multiple stages. These steps or stages do not necessarily need to be executed at the same time, but can be executed at different times. The execution order of these steps or stages does not necessarily need to be sequential, but can be executed alternately or alternately with at least a part of other steps or steps in other steps.

[0126] In one embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it realizes the steps of a fault early warning system for a coal mine underground substation as described above.

[0127] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, it realizes the steps in the above-mentioned system embodiments.

[0128] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to the partial descriptions of the method embodiments. The device embodiments described above are merely illustrative. The components described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the present disclosure solution. A person of ordinary skill in the art can understand and implement it without creative efforts.

[0129] The above embodiments only represent several implementation manners of the embodiments of the present application. The description is relatively specific and detailed, but it should not be construed as a limitation on the patent scope of the embodiments of the application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the embodiments of the present application, several deformations and improvements can still be made, and these all belong to the protection scope of the embodiments of the present application.

Claims

1. A coal mine underground substation fault early warning system, characterized in that: The system comprises: A data acquisition module is used to obtain electrical parameter data and environmental parameter data of the underground substation of the coal mine; the electrical parameter data includes at least one of voltage, current, power factor, transformer temperature, core vibration, partial discharge and gas parameters in oil; the environmental parameter data includes at least one of temperature, humidity and gas concentration; A feature extraction module is used to preprocess and extract features from the electrical parameter data and the environmental parameter data to obtain corresponding signal feature parameters; the signal feature parameters include electrical characteristics, thermodynamic characteristics and chemical characteristics parameters; and is also used to input the obtained signal feature parameters into a neural network model to construct a parameter association graph, and use node embedding technology to extract effect feature values ​​between parameters; A fault warning module is used to obtain the correlation influence of each parameter by aggregating the information of the neighboring nodes using the graph convolution layer based on the effect characteristic value; it is also used to input the correlation influence into the trained fault warning model to generate a fault diagnosis report; the fault diagnosis report includes a fault location code, a type identifier, a time prediction interval and a probability confidence level.

2. The system according to claim 1, characterized in that The preprocessing and feature extraction of the electrical parameter data and the environmental parameter data to obtain corresponding signal feature parameters includes: Performing preliminary timestamp alignment and data association operations on the electrical parameter data and the environmental parameter data to obtain a preliminary integrated parameter data set; Using a sliding average filtering algorithm to clean the electrical parameter data in the parameter data set and remove noise interference to obtain smoothed electrical parameters; Performing missing value processing on the environmental parameter data in the parameter data set, and estimating and filling the missing data using a linear regression method to obtain complete environmental parameter data; According to the complete environmental parameter data, a density-based spatial clustering algorithm is used to identify and mark abnormal environmental parameter values, and a local weighted regression method is used to correct the abnormal environmental parameter values ​​to obtain corrected environmental parameters; The electrical parameters after smoothing are extracted by fast Fourier transform method to obtain electrical characteristic parameters; the electrical characteristic parameters include voltage amplitude, current effective value, power factor, harmonic content, voltage fluctuation coefficient, current change rate and zero-sequence current value; Using a machine learning algorithm to extract features from the electrical parameters and environmental parameters to obtain thermodynamic characteristic parameters; the thermodynamic characteristic parameters include temperature, temperature change rate, temperature gradient, heat flux density and heat capacity; Using an association rule mining algorithm to extract features from the electrical parameters and environmental parameters to obtain chemical characteristic parameters; the chemical characteristic parameters include the influence coefficient of gas concentration, oxygen content, carbon monoxide concentration and humidity on the reaction of chemical substances; Based on the electrical characteristic parameters, thermodynamic characteristic parameters and chemical characteristic parameters, corresponding signal characteristic parameters are obtained by fusion.

3. The system according to claim 1, characterized in that The obtained signal characteristic parameters are input into a neural network model to construct a parameter association graph, and the effect characteristic values ​​between the parameters are extracted using a node embedding technique, including: Inputting the signal characteristic parameters into a neural network model to obtain a characteristic weight value; Constructing according to the obtained characteristic weight values, generating a parameter association diagram of the signal characteristic parameters; Using a graph structure analysis algorithm to extract vectors from the parameter association graph to obtain a node embedding vector; Based on the node embedding vector, a formula is used to calculate and obtain the effect characteristic value between parameters; The effect characteristic value is judged based on a preset threshold value, and if the effect characteristic value is greater than the preset threshold value, it is determined that there is a significant correlation between the parameters, and the characteristic weight value of the neural network model is updated according to the significant correlation; Generate updated effect characteristic values ​​corresponding to the signal characteristic parameters according to the updated neural network model.

4. The system according to claim 3, characterized in that The calculation based on the node embedding vector using a formula to obtain the effect characteristic value between parameters includes: The node embedding vector is calculated using the following formula to obtain the effect eigenvalue: Among them, Υ represents the effect eigenvalue, d represents the vector dimension, ω i represents the weight coefficient of the i-th dimension, e i represents the embedding value of the i-th dimension, and ReLU represents the activation function.

5. The system according to claim 1, characterized in that The step of aggregating information of neighboring nodes using a graph convolution layer based on the effect feature value to obtain the correlation influence of each parameter includes: Construct a parameter association matrix based on the effect characteristic value, and calculate the initial influence degree corresponding to each parameter; Constructing a graph structure according to the initial influence to obtain a neighborhood relationship between neighborhood nodes; Based on the neighborhood relationship, a graph convolution algorithm is used to aggregate the feature information of the neighborhood nodes to obtain a node influence value; The node influence value is judged based on a preset threshold value, and if the node influence value exceeds the preset threshold value, the global influence distribution is adjusted based on the preset threshold value to obtain the associated influence degree; The global influence distribution is optimized using a machine learning algorithm to obtain an updated correlation influence degree.

6. The system according to claim 1, characterized in that The fault warning module further includes: Calculate based on the associated influence parameter combined with the time decay factor to obtain a dynamic influence value; The dynamic impact value is input into the trained fault warning model to generate a candidate fault set using fault feature matching rules; Screening the candidate fault set using a probability weight threshold to obtain a target fault type; Calling a fault location mapping library according to the target fault type to generate a predicted fault diagnosis report; the fault diagnosis report includes a fault location code, a type identifier, a time prediction interval, and a probability confidence level; When the probability confidence in the fault diagnosis report exceeds a preset threshold or a fault may occur within the time prediction interval, an early warning message is issued through an audible and visual alarm, and the early warning message is sent to a terminal device of a manager.

7. The system according to claim 6, characterized in that The calculation based on the associated influence parameter combined with the time decay factor to obtain the dynamic influence value includes: The dynamic impact value is obtained by calculating the associated impact parameter using the following formula: Among them, I(t) represents the dynamic impact value, γ represents the impact intensity coefficient, and P k Represents the value of each associated influence parameter, m represents the number of parameters, λ represents the time decay rate, and t represents the time span.

8. A coal mine underground substation fault early warning method, characterized in that: The method comprises: Acquire electrical parameter data and environmental parameter data of the underground substation of the coal mine; the electrical parameter data includes at least one of voltage, current, power factor, transformer temperature, core vibration, partial discharge and gas parameters in oil; the environmental parameter data includes at least one of temperature, humidity and gas concentration; Preprocessing and feature extraction are performed on the electrical parameter data and the environmental parameter data to obtain corresponding signal feature parameters; the signal feature parameters include electrical characteristics, thermodynamic characteristics and chemical characteristics parameters; the obtained signal feature parameters are input into a neural network model to construct a parameter association graph, and the effect feature values ​​between the parameters are extracted using node embedding technology; Based on the effect characteristic value, the information of the neighborhood nodes is aggregated using the graph convolution layer to obtain the correlation influence of each parameter; the correlation influence is input into the trained fault warning model to generate a fault diagnosis report; the fault diagnosis report includes the fault location code, type identifier, time prediction interval and probability confidence.

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 system 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 system according to any one of claims 1 to 7 are implemented.

Images