Power equipment fault identification method and identification system based on deep learning

Abstract

The present invention belongs to the technical field of fault identification, and discloses a method and system for identifying faults of electric power equipment based on deep learning, including collecting magnetic core data and oil tank data of electric power transformer equipment; collecting historical oil quality data and magnetic core noise data; training a machine learning model for predicting oil quality water content data; inputting real-time oil tank feature data into the machine learning model to obtain oil quality water content data; processing the magnetic core data to obtain a first magnetic core state coefficient A; processing the oil tank data and magnetic core noise data of the electric power transformer equipment to obtain a second magnetic core state coefficient C; judging whether the magnetic core state is normal or faulty; generating fault information and reporting it to a maintenance end; and realizing three-dimensional monitoring of the magnetic core state of the electric power transformer equipment; which is of great significance for ensuring the reliable and continuous operation of the electric power transformer equipment.

Description

Technical Field

[0001] The present invention relates to the technical field of fault identification, and more specifically, to a method and system for identifying faults of electric power equipment based on deep learning. Background Art

[0002] The Chinese patent with the authorization announcement number CN112147481B discloses an IGBP fault feedback circuit and electronic device, including: a first control module for receiving all types of fault information, and outputting complementary first pulse signals and second pulse signals, driving the micro signal isolation power transformer device in a differential pulse driving manner; the micro signal isolation power transformer device is used to output a target pulse signal according to the first pulse signal and the second pulse signal; the signal shaping module is used to shape the target pulse signal into a target digital signal; the second control module is used to receive the target digital signal and perform signal processing on the target digital signal to identify the type of fault information. The IBGP fault feedback circuit encodes all types of fault information to generate two complementary pulse signals, drives the micro signal isolation power transformer device in a differential pulse driving manner, and solves the problem that the magnetic core of the micro signal isolation power transformer device is easy to saturate.

[0003] However, the existing core status monitoring system for power transformer equipment has a single data source and simple analysis methods. Most of them only focus on single indicators such as oil quality, temperature, and noise. For example, it is difficult to fully judge whether the core status is faulty by only measuring the dissolved gas content in the insulating oil. The core is a core component of the power transformer equipment, and core fault detection has become very important. Therefore, it is necessary to conduct detection from multiple comprehensive data. If a single data detection method is used, some fault signals may often be missed or not captured, and the fault may not be discovered in time. The accumulation of hidden dangers will increase the probability of accidents in power transformer equipment.

[0004] In addition, the existing technology lacks effective fusion and comprehensive analysis of data in detecting core faults, and is unable to form a three-dimensional judgment on the core status of power transformer equipment, which increases the false alarm and missed alarm rates.

[0005] In view of this, the present invention proposes a power equipment fault identification method and identification system based on deep learning to solve the above problems. Summary of the invention

[0006] In order to overcome the above-mentioned defects of the prior art and to achieve the above-mentioned purpose, the present invention provides the following technical solutions: A method for identifying faults of electric power equipment based on deep learning, comprising: collecting magnetic core data and oil tank data of electric power transformer equipment;

[0007] Collect historical oil quality data and core noise data; historical oil quality data includes historical oil water content data and historical oil tank characteristic data;

[0008] Train a machine learning model to predict oil water content data based on historical oil quality data;

[0009] Input the real-time fuel tank characteristic data into the trained machine learning model to obtain the oil quality water content data;

[0010] Processing the magnetic core data to obtain a first magnetic core state coefficient A;

[0011] Processing the oil tank data and the core noise data of the power transformer equipment to obtain a second core state coefficient C;

[0012] The first magnetic core state coefficient A and the second magnetic core state coefficient C are processed to obtain a third magnetic core state coefficient W; the magnetic core state is judged to be normal or faulty according to the third magnetic core state coefficient; if the magnetic core state is judged to be faulty, fault information is generated and reported to the maintenance end.

[0013] Furthermore, the core data includes core temperature data, magnetic flux data and core current data; the power transformer equipment oil tank data includes oil tank hydrogen content data and oil acidity data.

[0014] Furthermore, the core temperature data is obtained by installing temperature sensors at n high temperature areas outside the core close to the winding; the readings of the n temperature sensors are averaged as the core temperature data;

[0015] The flux data is obtained by installing a flux measurement coil outside the magnetic core;

[0016] Wind the coil evenly outside the magnetic core near the magnetic core; lead out signal lines from both ends of the coil, and connect the signal lines to the data acquisition card to collect magnetic flux signals; the data acquisition card sends the magnetic flux signals to the industrial computer; the industrial computer processes and obtains real-time magnetic flux, which is magnetic flux data;

[0017] The core current data is obtained by installing a current transformer on the outer surface of the core;

[0018] The coil of the current transformer is completely enclosed on the outer surface of the magnetic core; the coil is wound around the entire magnetic core without leaving any gaps; the output end of the current transformer is connected to a data acquisition card for collecting mutual inductance readings; the core current value is calculated based on the transformation ratio of the current transformer and the mutual inductance readings, which is the core current data.

[0019] Furthermore, the hydrogen content data of the fuel tank is obtained by installing a gas analysis sensor on the top of the fuel tank;

[0020] The gas analysis sensor is provided with a sensor probe; the sensor probe has a built-in thermal conductivity hydrogen analyzer; the sensor probe outputs an analog electrical signal of hydrogen concentration; the analog electrical signal is connected to a data acquisition card, analog-to-digital conversion is performed to obtain a digital electrical signal, and the digital electrical signal is sent to an industrial computer; the industrial computer processes the digital electrical signal according to the oil volume in the power transformer equipment to obtain the hydrogen content, which is the hydrogen content data of the fuel tank;

[0021] The oil acidity data is obtained by installing an online acidity sensor in the oil tank of the power transformer equipment.

[0022] Furthermore, the historical oil quality water content data is obtained by extracting the insulating oil of the power transformer equipment for testing in the historical Q intervals P time period in advance, and the insulating oil water content obtained is the historical oil quality water content data;

[0023] The historical fuel tank characteristic data includes fuel tank ambient temperature data, fuel tank ambient humidity data, load current data and fuel pump flow data.

[0024] Furthermore, the oil tank ambient temperature data is the average value of the oil tank temperature when the insulating oil of the power transformer equipment is extracted in advance during the interval P time period, and the oil tank ambient temperature data is obtained through a temperature sensor; the oil tank ambient humidity data is the average value of the oil tank humidity when the insulating oil of the power transformer equipment is extracted in advance during the interval P time period, and the oil tank ambient humidity data is obtained through a humidity sensor; the load current data is the maximum value of the load current when the insulating oil of the power transformer equipment is extracted in advance during the interval P time period, and the load current data is obtained through an electronic ammeter; the oil pump flow data is the average value of the oil pump flow when the insulating oil of the power transformer equipment is extracted in advance during the interval P time period, and the oil pump flow data is obtained through a flow meter installed on the insulating oil circulation pipeline.

[0025] Furthermore, the specific training methods of the machine learning model include:

[0026] Convert historical fuel tank feature data into a set of feature vectors;

[0027] Each group of feature vectors is used as the input of a machine learning model. The machine learning model uses the historical oil water content data corresponding to each group of feature data as output, uses the actual historical oil water content data corresponding to each group of actual fuel tank feature data as a prediction target, and uses minimizing the machine learning model loss function value as a training target; training is stopped when the machine learning model loss function value is less than or equal to a preset target loss value, and the machine learning model is one of a logistic regression model, a naive Bayes model, or a support vector machine model.

[0028] Furthermore, the core noise data is obtained as follows:

[0029] A directional microphone is used to collect environmental noise signals; a circular disc microphone is used to collect a mixed sound signal of a core sound signal of an electric transformer and an environmental noise signal; an adaptive digital filter is used to filter the environmental noise signal and the mixed sound signal respectively to obtain a filtered environmental noise signal and a filtered mixed sound signal with the same intensity; a noise cancellation method is used to perform frequency spectrum difference on the filtered environmental noise signal and the filtered mixed sound signal to obtain a core sound signal of an electric transformer; an analog-to-digital converter is used to convert the core sound signal of an electric transformer into digital form of core sound data of an electric transformer, namely, core noise data.

[0030] Furthermore, the first core state coefficient

[0031] Wherein, T represents the core temperature data, B(t) represents the flux data that changes with time, and I(t) represents the core current data that changes with time; α, β, λ and δ are adjustment parameters used to adjust the weights and influence of each item in the coefficient; t is the acquisition time; the adjustment parameters α, β, λ and δ are obtained by fitting the experimental data through regression analysis and least squares method; wherein the experimental data is the collected n sets of historical core data;

[0032] Second core state coefficient

[0033] Among them, H 2 (t) represents the fuel tank hydrogen content data changing with time, A(t) represents the oil acidity data changing with time, N(t) represents the core noise data changing with time; Tz is the total time for obtaining the fuel tank hydrogen content data, oil acidity data and core noise data; e is the base of the exponential function; y, k, j and l are the second adjustment parameters used to adjust the weight and influence of each data; the acquisition process of the second adjustment parameter is the same as the acquisition process of the adjustment parameter.

[0034] Furthermore, the third core state coefficient

[0035] The method for judging whether the core status is normal or faulty is as follows:

[0036] Set the core maximum state threshold U 1 and the lowest core state threshold U 2 ;

[0037] If the third core state coefficient W>U 1 Or W<U 2 , then the core status is judged to be faulty;

[0038] If the third core state coefficient W≤U 1 And W ≥ U2 , then the core state is judged to be normal;

[0039] Core normal state threshold U 1 and the lowest core state threshold U 2 The settings are as follows:

[0040] Under the normal operation state of the magnetic core, m groups of magnetic core data are collected and m first magnetic core state coefficients are calculated. Under the normal operation state of the magnetic core, m groups of oil tank data of the power transformer equipment and magnetic core noise data are collected and m second magnetic core state coefficients are calculated. According to the m first magnetic core state coefficients and the m second magnetic core state coefficients, m third magnetic core state coefficients are calculated. The m third magnetic core state coefficients are traversed to find the third magnetic core state coefficient with the smallest value as the lowest state threshold value U of the magnetic core. 2 , traverse the m third core state coefficients to find the third core state coefficient with the largest value as the core highest state threshold U 1 ;

[0041] The fault information includes the core data of the electric power transformer equipment when the core state is determined to be faulty, the oil tank data of the electric power transformer equipment, the water content data of the oil and the core noise data.

[0042] The power equipment fault identification system based on deep learning includes: a first data acquisition module for collecting magnetic core data and electric power transformer equipment oil tank data; sending the magnetic core data to a first data preprocessing module; sending the electric power transformer equipment oil tank data to a second preprocessing module;

[0043] The second data acquisition module is used to collect historical oil quality data and magnetic core noise data; send the historical oil quality data to the machine training module; send the magnetic core noise data to the second data preprocessing module;

[0044] The machine training module trains a machine learning model for predicting oil water content data based on historical oil quality data;

[0045] The oil quality water content acquisition module inputs the real-time oil tank characteristic data into the trained machine learning model to obtain the oil quality water content data; and sends the oil quality water content data to the second data preprocessing module;

[0046] The first data preprocessing module processes the magnetic core data to obtain a first magnetic core state coefficient A; and sends the first magnetic core state coefficient A to the magnetic core state analysis module;

[0047] The second data preprocessing module processes the oil tank data and the magnetic core noise data of the power transformer equipment to obtain a second magnetic core state coefficient C; and sends the second magnetic core state coefficient C to the magnetic core state analysis module;

[0048] The core state analysis module processes the first core state coefficient A and the second core state coefficient C to obtain a third core state coefficient W; determines whether the core state is normal or faulty according to the third core state coefficient; if the core state is determined to be faulty, generates fault information and reports it to the maintenance end; and each module is connected by wire and / or wireless means to realize data transmission between modules.

[0049] Technical effects and advantages of the power equipment fault identification method and identification system based on deep learning of the present invention:

[0050] By comprehensively analyzing multi-source data, it is possible to more comprehensively and accurately determine whether the magnetic core of the power transformer equipment is faulty than by monitoring only a single data indicator; it is impossible to determine whether the magnetic core itself has a problem based only on oil quality data, and the present invention can accurately and timely discover the fault of the magnetic core, and take remedial measures in time to avoid further expansion of the fault; by setting the upper and lower limits of the judgment threshold, the judgment process is made more accurate, avoiding the probability of false reporting and missed reporting; the present invention realizes three-dimensional monitoring of the state of the magnetic core of the power transformer equipment, and the judgment result is more comprehensive and reliable; this has important practical significance for ensuring the reliable and continuous operation of the power transformer equipment, avoiding large-scale power outages, and improving the safety and stability of power supply of the power grid. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 This is a flow chart of a method for identifying power equipment faults based on deep learning according to the present invention;

[0052] Figure 2 A schematic diagram of a power equipment fault identification system based on deep learning of the present invention;

[0053] Figure 3 is a schematic diagram of an electronic device according to Embodiment 3 of the present invention;

[0054] Figure 4 This is a schematic diagram of a storage medium according to Embodiment 4 of the present invention. DETAILED DESCRIPTION

[0055] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0056] Example 1

[0057] See also Figure 1 As shown, the power equipment fault identification method based on deep learning described in this embodiment includes collecting magnetic core data and power transformer equipment oil tank data;

[0058] The core data includes core temperature data, magnetic flux data and core current data;

[0059] The oil tank data of power transformer equipment includes the hydrogen content data and oil acidity data of the oil tank;

[0060] It should be noted that the core temperature reflects the core loss and overheating. If the temperature rises too high, it means that the core may fail. The flux waveform reflects the core flux distribution and magnetic saturation. The waveform distortion indicates that the core may fail. The core current is the core induced current. If the current is too large, it means that the core circuit may be short-circuited, indicating that the core may fail. The increase in hydrogen content in the oil tank means that the core has discharged and the insulation has been damaged, indicating that the core may fail. The increase in oil acidity indicates that the insulating oil in the oil tank has been damaged, affecting the insulation of the core.

[0061] The core temperature data is obtained by installing n temperature sensors at n high temperature areas outside the core close to the winding;

[0062] Specifically, the reading values ​​of n temperature sensors are averaged as the core temperature data;

[0063] It should be noted that the temperature distribution outside the core is different. The area close to the winding is usually hotter than other areas. This is because when the core conducts magnetic flux, the magnetic pole area close to the winding has a larger magnetic induction intensity, and the iron loss and eddy current loss are also larger, generating more heat. The main winding is the working winding of the power transformer equipment. The transmission resistance loss generated by the current carrying function is the main heat source. This part of heat is easily conducted to the core, causing the corresponding area to heat up. When installing the temperature sensor, use an insulating gasket or insulating wire to firmly attach it to the surface of the high-temperature area of ​​the core to ensure good thermal contact, and firmly lead out the signal lines of each sensor to avoid magnetic field interference with the signal line.

[0064] The flux data is obtained by installing a flux measurement coil outside the magnetic core;

[0065] Specifically, a coil is evenly wound outside the magnetic core near the magnetic core; signal lines are led out from both ends of the coil, and the signal lines are connected to a data acquisition card for collecting magnetic flux signals; the data acquisition card sends the magnetic flux signals to an industrial computer; the industrial computer processes and obtains real-time magnetic flux, which is magnetic flux data;

[0066] It should be noted that the frequency of the magnetic flux signal collected by the data acquisition card is not less than 10 times the working frequency; for the power transformer equipment in the power system, its working frequency refers to the power frequency of the power grid, that is, the power frequency; the reason for this setting is that it can correctly collect the instantaneous magnetic flux of the magnetic core of the power transformer equipment under normal power frequency, reflecting its magnetic flux changes, without causing serious sampling errors or discrete effects;

[0067] The core current data is the induced current generated during the working process of the magnetic core, which is obtained by installing a current transformer on the outer surface of the magnetic core;

[0068] Specifically, the coil of the current transformer is completely sheathed on the outer surface of the magnetic core; the coil is wound around the entire magnetic core without leaving any gaps; the output end of the current transformer is connected to a data acquisition card for collecting mutual inductance readings; the core current value is calculated according to the transformation ratio of the current transformer and the mutual inductance readings, that is, the core current data;

[0069] It should be noted that the coil of the current transformer is completely enclosed on the outer surface of the magnetic core in order to accurately intercept all the magnetic flux on the magnetic core and reflect its magnetic field changes; the coil should be wound without gaps to avoid magnetic flux leakage and ensure that all magnetic flux passes through the coil; the output end is connected to the acquisition card to obtain the voltage / current signal containing the core current information; the current value is calculated according to the transformation ratio because the number of turns of the transformer coil and the core size have a certain proportional relationship, which determines the transformation ratio; the core current data and the rated current of the power transformer equipment are also related to each other. Under normal conditions, the maximum core current does not exceed 2% of the rated current of the power transformer equipment. The rated current value of the power transformer equipment varies according to the capacity of the power transformer equipment. Generally, the rated current of small-capacity power transformer equipment (tens of KVA level) is generally between 10A and 100A; the rated current of medium-capacity power transformer equipment is about 100A to 800A; the rated current of large-capacity power transformer equipment can reach more than 1000A; it is usually indicated on the nameplate of the power transformer equipment;

[0070] The hydrogen content data of the fuel tank is obtained by installing a gas analysis sensor on the top of the fuel tank;

[0071] Specifically, the gas analysis sensor is provided with a sensor probe; the sensor probe has a built-in thermal conductivity hydrogen analyzer; the sensor probe outputs an analog electrical signal of hydrogen concentration; the analog electrical signal is connected to a data acquisition card, analog-to-digital conversion is performed to obtain a digital electrical signal, and the digital electrical signal is sent to an industrial computer; the industrial computer processes the digital electrical signal according to the oil volume in the power transformer to obtain the hydrogen content, which is the hydrogen content data of the fuel tank;

[0072] It should be noted that the oil tank volume of the power transformer equipment can be obtained through the parameters on the oil tank. By obtaining the liquid level of the insulating oil in the oil tank of the power transformer equipment, the oil volume in the power transformer equipment can be obtained; according to the volume fraction of hydrogen in the insulating oil, that is, the concentration, the hydrogen content of the oil tank can be calculated according to the proportional relationship, that is, hydrogen content = oil volume in the oil tank x hydrogen concentration;

[0073] The oil acidity data is obtained by installing an online acidity sensor in the oil tank of the power transformer equipment;

[0074] Collect historical oil quality data and core noise data. The historical oil quality data includes historical oil water content data and historical oil tank characteristic data.

[0075] The historical oil quality water content data is obtained by the staff extracting the insulating oil of the power transformer equipment at the historical Q intervals P time period and testing the insulating oil to obtain the water content of the insulating oil, which is the historical oil quality water content data;

[0076] The historical fuel tank characteristic data includes fuel tank ambient temperature data, fuel tank ambient humidity data, load current data and fuel pump flow data;

[0077] The oil tank ambient temperature data is the average oil tank temperature when the staff extracts the insulating oil of the power transformer equipment during the interval P time period, which is obtained through the temperature sensor; the oil tank ambient humidity data is the average oil tank humidity when the staff extracts the insulating oil of the power transformer equipment during the interval P time period, which is obtained through the humidity sensor; the load current data is the maximum load current when the staff extracts the insulating oil of the power transformer equipment during the interval P time period, which is obtained through the electronic ammeter; the oil pump flow data is the average oil pump flow when the staff extracts the insulating oil of the power transformer equipment during the interval P time period, which is obtained through the flow meter installed on the insulating oil circulation pipeline;

[0078] A machine learning model for predicting oil water content data is trained based on historical oil quality data; the specific training methods of the machine learning model include:

[0079] Convert historical fuel tank feature data into a set of feature vectors;

[0080] Each set of feature vectors is used as input of a machine learning model, the machine learning model uses historical oil quality water content data corresponding to each set of feature data as output, uses actual historical oil quality water content data corresponding to each set of actual fuel tank feature data as a prediction target, and uses minimizing the machine learning model loss function value as a training target; and stops training when the machine learning model loss function value is less than or equal to a preset target loss value;

[0081] The loss function value of the machine learning model is the mean square error;

[0082] By transforming the loss function formula Minimization is used as the goal to train the model so that the machine learning model can better fit the data, thereby improving the performance and accuracy of the model;

[0083] In the loss function, MSE is the loss function value of the machine learning model, x is the feature vector group number; m is the number of feature vector groups; f x is the predicted historical oil water content data corresponding to the xth group of eigenvectors, is the historical oil quality water content data actually corresponding to the xth group of characteristic vectors;

[0084] Other model parameters of the machine learning model, target loss value, optimization algorithm, training set test set validation set ratio, and loss function optimization are all obtained through actual engineering implementation and continuous experimental tuning;

[0085] The core noise data is obtained as follows:

[0086] A directional microphone is used to collect environmental noise signals; a circular disc microphone is used to collect a mixed sound signal of a core sound signal of an electric power transformer and an environmental noise signal; an adaptive digital filter is used to filter the environmental noise signal and the mixed sound signal respectively to obtain a filtered environmental noise signal and a filtered mixed sound signal with the same intensity; a noise cancellation method is used to perform a frequency difference between the filtered environmental noise signal and the filtered mixed sound signal to obtain a core sound signal of the electric power transformer; an analog-to-digital converter (A / D converter) is used to convert the core sound signal of the electric power transformer into digital form of core sound data of the electric power transformer, namely core noise data;

[0087] Input the real-time fuel tank characteristic data into the trained machine learning model to obtain the oil quality water content data;

[0088] It should be noted that there are many factors that affect the water content of the oil in the oil tank of the power transformer equipment. Through the prediction of the machine learning model, the influence of various factors on the water content of the oil can be comprehensively considered to make the acquired data more accurate and true; the above-mentioned acquisition of core noise data uses two different directional pickups to distinguish the target sound source and the noise sound source; the adaptive digital filtering ensures that the frequency band range of the two signals is consistent, which is conducive to subsequent cancellation; the noise cancellation uses the principle of spectrum difference to effectively reduce the normal noise; making the acquired core sound signal clearer; the process of collecting the above data has been unified in time to ensure that the above data reflects the state of the core at the same time;

[0089] Processing the magnetic core data to obtain a first magnetic core state coefficient A; sending the first magnetic core state coefficient A to a magnetic core state analysis module;

[0090] Specifically, the first core state coefficient

[0091] Wherein, T represents the core temperature data, B(t) represents the flux data that changes with time, and I(t) represents the core current data that changes with time; α, β, λ and δ are adjustment parameters used to adjust the weights and influence of each item in the coefficient; t is the acquisition time; the adjustment parameters α, β, λ and δ are obtained by fitting the experimental data through regression analysis and least squares method; wherein the experimental data are n sets of historical core data collected in the experimental environment;

[0092] It should be noted that the least squares method is used to fit the data; the goal of the least squares method is to minimize the sum of the squares of the differences between the model predictions and the actual observations; mathematical software or programming languages ​​can be used to perform this fitting; in the above process, the values ​​of α, β, λ and δ are adjusted to minimize the error between the model predictions and the actual observations, which can be achieved through an iterative algorithm, constantly adjusting the values ​​of α, β, λ and δ until the minimum error is found; the least squares method can use statistical software or programming languages, such as PyThon's SciTy library or MATLAB, to perform fitting operations;

[0093] It should be noted that e α×T It is an exponential function of the core temperature, indicating the effect of the core temperature on the coefficient; is an infinite integral symbol, which means integration over time; here, it is used to integrate the flux data to consider the change of flux over time; -λ×t It is an exponential function about time, which is used to consider the decay or growth of magnetic flux over time; 1+δ×I represents the influence of core current on the coefficient;

[0094] Processing the oil tank data and the core noise data of the power transformer equipment to obtain a second core state coefficient C; sending the second core state coefficient C to the core state analysis module;

[0095] Specifically, the second core state coefficient

[0096] Among them, H 2 (t) represents the hydrogen content data of the fuel tank that changes with time, A(t) represents the oil acidity data that changes with time, and N(t) represents the core noise data that changes with time; Tz is the total time for obtaining the above data; y, k, j and l are the second adjustment parameters, which are used to adjust the weight and influence of each data; the acquisition process is the same as the acquisition process of the adjustment parameters;

[0097] It should be noted that It represents the integral within the total time range. The trigonometric functions cosine and sinine are added to introduce periodic influencing factors. This is because the hydrogen content in the oil tank, the acidity of the oil, and the core noise may show a periodic change during the operation of the power transformer equipment. The reason for putting the core noise and the oil tank data of the power transformer equipment together for calculation is that the core noise data can reflect the physical state of the core, while the oil tank data of the power transformer equipment can provide information about the insulating oil. Combining them can provide a more comprehensive understanding of the working state of the core.

[0098] The first magnetic core state coefficient A and the second magnetic core state coefficient C are processed to obtain a third magnetic core state coefficient W; the magnetic core state is judged to be normal or faulty according to the third magnetic core state coefficient; if the magnetic core state is judged to be faulty, fault information is generated and reported to the maintenance end;

[0099] Specifically, the third core state coefficient

[0100] The method for judging whether the core status is normal or faulty is as follows:

[0101] Set the core maximum state threshold U 1 and the lowest core state threshold U 2 ;

[0102] If the third core state coefficient W>U 1 Or W<U 2 , then the core status is judged to be faulty;

[0103] If the third core state coefficient W≤U 1 And W ≥ U 2 , then the core state is judged to be normal;

[0104] Core normal state threshold U 1 and the lowest core state threshold U 2 The settings are as follows:

[0105] In the experimental environment, m groups of core data are collected under normal operation of the core and m first core state coefficients are calculated. m groups of power transformer equipment oil tank data and core noise data are collected under normal operation of the core and m second core state coefficients are calculated. m third core state coefficients are calculated based on the m first core state coefficients and the m second core state coefficients. The m third core state coefficients are traversed to find the third core state coefficient with the smallest value as the lowest core state threshold value U 2 , traverse the m third core state coefficients to find the third core state coefficient with the largest value as the core highest state threshold U 1 ;

[0106] The fault information includes the magnetic core data, the oil tank data of the power transformer equipment, the water content data of the oil quality and the magnetic core noise data at that time;

[0107] It should be noted that the third core state coefficient is in a fluctuating state during the normal operation of the core of the power transformer equipment. The highest and lowest thresholds are set to make the judgment process more accurate. The third core state coefficient W is a state coefficient obtained by combining many factors affecting the core state, and is used to judge the core state more accurately.

[0108] In this embodiment, by comprehensively analyzing multi-source data, it is possible to more comprehensively and accurately determine whether the magnetic core of the power transformer equipment is faulty than by monitoring only a single data indicator; it is impossible to determine whether the magnetic core itself has a problem based only on oil quality data. By using this embodiment, the fault of the magnetic core can be accurately and timely discovered, and remedial measures can be taken in time to avoid further expansion of the fault; by setting the upper and lower limits of the judgment threshold, the judgment process is made more accurate, and the probability of false reporting and missed reporting is avoided; this embodiment realizes three-dimensional monitoring of the state of the magnetic core of the power transformer equipment, and the judgment result is more comprehensive and reliable; this has important practical significance for ensuring the reliable and continuous operation of the power transformer equipment, avoiding large-scale power outages, and improving the safety and stability of power supply of the power grid;

[0109] Example 2

[0110] See also Figure 2 As shown, the part not described in detail in this embodiment is described in Example 1, and a power equipment fault identification system based on deep learning is provided, including: a first data acquisition module for collecting magnetic core data and electric power transformer equipment oil tank data; sending the magnetic core data to a first data preprocessing module; sending the electric power transformer equipment oil tank data to a second preprocessing module;

[0111] The second data acquisition module is used to collect historical oil quality data and magnetic core noise data; send the historical oil quality data to the machine training module; send the magnetic core noise data to the second data preprocessing module;

[0112] The machine training module trains a machine learning model for predicting oil water content data based on historical oil quality data;

[0113] The oil quality water content acquisition module inputs the real-time oil tank characteristic data into the trained machine learning model to obtain the oil quality water content data; and sends the oil quality water content data to the second data preprocessing module;

[0114] The first data preprocessing module processes the magnetic core data to obtain a first magnetic core state coefficient A; and sends the first magnetic core state coefficient A to the magnetic core state analysis module;

[0115] The second data preprocessing module processes the oil tank data and the magnetic core noise data of the power transformer equipment to obtain a second magnetic core state coefficient C; and sends the second magnetic core state coefficient C to the magnetic core state analysis module;

[0116] The core state analysis module processes the first core state coefficient A and the second core state coefficient C to obtain a third core state coefficient W; determines whether the core state is normal or faulty according to the third core state coefficient; if the core state is determined to be faulty, generates fault information and reports it to the maintenance end; and each module is connected by wire and / or wireless means to realize data transmission between modules.

[0117] Example 3

[0118] See also Figure 3 As shown, according to another aspect of the present application, an electronic device 500 is also provided. The electronic device 500 may include one or more processors and one or more memories. The memories store computer readable codes, and when the computer readable codes are run by one or more processors, one of the above-mentioned operations may be performed.

[0119] The method or system according to the embodiment of the present application can also be used by Figure 3 The electronic device architecture shown in FIG. Figure 3 As shown, the electronic device 500 may include a bus 501, one or more CPUs 502, a read-only memory (ROM) 503, a random access memory (RAM) 504, a communication port 505 connected to a network, an input / output 506, a hard disk 507, etc. The storage device in the electronic device 500, such as the ROM 503 or the hard disk 507, may store the power equipment fault identification method based on deep learning provided in the present application. Furthermore, the electronic device 500 may also include a user interface 508. Of course, Figure 3 The architecture shown is only exemplary and can be omitted according to actual needs when implementing different devices. Figure 3 One or more components of an electronic device are shown.

[0120] Example 4

[0121] See also Figure 4As shown, a computer-readable storage medium 600 according to an embodiment of the present application is shown. Computer-readable instructions are stored on the computer-readable storage medium 600. When the computer-readable instructions are executed by the processor, the method for identifying faults of power equipment based on deep learning according to the embodiment of the present application described with reference to the above figures can be executed. The storage medium 600 includes, but is not limited to, for example, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory (cache), etc. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0122] In addition, according to the implementation of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the present application provides a non-transitory machine-readable storage medium, which stores machine-readable instructions, and the machine-readable instructions can be executed by a processor to execute instructions corresponding to the method steps provided by the present application, such as: a power equipment fault identification method based on deep learning. When the computer program is executed by a central processing unit (CPU), the above functions defined in the method of the present application are executed.

[0123] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center through a wired network or a wireless network. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a tape), an optical medium (for example, a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0124] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in the present invention can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0125] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0126] In the several embodiments provided by the present invention, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the units is only one, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0127] The units 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 on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0128] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0129] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

[0130] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for identifying power equipment faults based on deep learning, characterized in that: include: Collect magnetic core data and oil tank data of power transformer equipment; magnetic core data includes magnetic core temperature data, magnetic flux data and iron core current data; oil tank data of power transformer equipment includes oil tank hydrogen content data and oil acidity data; Collect historical oil quality data and core noise data; historical oil quality data includes historical oil water content data and historical oil tank characteristic data; Train a machine learning model to predict oil water content data based on historical oil quality data; Input the real-time fuel tank characteristic data into the trained machine learning model to obtain the oil quality water content data; Processing the magnetic core data to obtain a first magnetic core state coefficient A; The first core state coefficient Wherein, T represents the core temperature data, B(t) represents the flux data that changes with time, and I(t) represents the core current data that changes with time; α, β, λ and δ are adjustment parameters used to adjust the weights and influence of each item in the coefficient; t is the acquisition time; the adjustment parameters α, β, λ and δ are obtained by fitting the experimental data through regression analysis and least squares method; wherein the experimental data is the collected n sets of historical core data; Processing the oil tank data and the core noise data of the power transformer equipment to obtain a second core state coefficient C; Second core state coefficient Wherein, H2(t) represents the fuel tank hydrogen content data varying with time, A(t) represents the oil acidity data varying with time, and N(t) represents the core noise data varying with time; Tz is the total time for obtaining the fuel tank hydrogen content data, oil acidity data, and core noise data; e is the base of the exponential function; y, k, j, and l are the second adjustment parameters, which are used to adjust the weight and influence of each data; the acquisition process of the second adjustment parameter is the same as the acquisition process of the adjustment parameter; The first magnetic core state coefficient A and the second magnetic core state coefficient C are processed to obtain a third magnetic core state coefficient W; the third magnetic core state coefficient The method for judging whether the core status is normal or faulty is as follows: Set the core maximum state threshold U1 and the core minimum state threshold U2; If the third core state coefficient W>U1 or W<U2, the core state is determined to be faulty; If the third core state coefficient W≤U1 and W≥U2, the core state is determined to be normal; The core state is determined to be normal or faulty according to the third core state coefficient; if the core state is determined to be faulty, fault information is generated and reported to the maintenance end.

2. The method for identifying power equipment faults based on deep learning according to claim 1, characterized in that: The core temperature data is obtained by installing temperature sensors at n high temperature areas outside the core close to the winding; the readings of the n temperature sensors are averaged as the core temperature data; The flux data is obtained by installing a flux measurement coil outside the magnetic core; Wind the coil evenly outside the magnetic core near the magnetic core; lead out signal lines from both ends of the coil, and connect the signal lines to the data acquisition card to collect magnetic flux signals; the data acquisition card sends the magnetic flux signals to the industrial computer; the industrial computer processes and obtains real-time magnetic flux, which is magnetic flux data; The core current data is obtained by installing a current transformer on the outer surface of the core; The coil of the current transformer is completely enclosed on the outer surface of the magnetic core; the coil is wound around the entire magnetic core without leaving any gaps; the output end of the current transformer is connected to a data acquisition card for collecting mutual inductance readings; the core current value is calculated based on the transformation ratio of the current transformer and the mutual inductance readings, which is the core current data.

3. The method for identifying power equipment faults based on deep learning according to claim 2, characterized in that: The hydrogen content data of the fuel tank is obtained by installing a gas analysis sensor on the top of the fuel tank; The gas analysis sensor is provided with a sensor probe; the sensor probe has a built-in thermal conductivity hydrogen analyzer; the sensor probe outputs an analog electrical signal of hydrogen concentration; the analog electrical signal is connected to a data acquisition card, analog-to-digital conversion is performed to obtain a digital electrical signal, and the digital electrical signal is sent to an industrial computer; the industrial computer processes the digital electrical signal according to the oil volume in the power transformer equipment to obtain the hydrogen content, which is the hydrogen content data of the fuel tank; The oil acidity data is obtained by installing an online acidity sensor in the oil tank of the power transformer equipment.

4. The method for identifying power equipment faults based on deep learning according to claim 3 is characterized in that: The historical oil quality water content data is obtained by extracting the insulating oil of the power transformer equipment for testing in the historical Q intervals P time period in advance. The insulating oil water content obtained is the historical oil quality water content data; The historical fuel tank characteristic data includes fuel tank ambient temperature data, fuel tank ambient humidity data, load current data and fuel pump flow data.

5. The method for identifying power equipment faults based on deep learning according to claim 4, characterized in that: The oil tank ambient temperature data is the average value of the oil tank temperature when the insulating oil of the power transformer equipment is extracted in advance during the interval P time period, and the oil tank ambient temperature data is obtained through a temperature sensor; the oil tank ambient humidity data is the average value of the oil tank humidity when the insulating oil of the power transformer equipment is extracted in advance during the interval P time period, and the oil tank ambient humidity data is obtained through a humidity sensor; the load current data is the maximum value of the load current when the insulating oil of the power transformer equipment is extracted in advance during the interval P time period, and the load current data is obtained through an electronic ammeter; the oil pump flow data is the average value of the oil pump flow when the insulating oil of the power transformer equipment is extracted in advance during the interval P time period, and the oil pump flow data is obtained through a flow meter installed on the insulating oil circulation pipeline.

6. The method for identifying power equipment faults based on deep learning according to claim 5, characterized in that: The specific training methods of machine learning models include: Convert historical fuel tank feature data into a set of feature vectors; Each set of feature vectors is used as input of a machine learning model, the machine learning model uses historical oil quality water content data corresponding to each set of feature data as output, uses actual historical oil quality water content data corresponding to each set of actual oil tank feature data as a prediction target, and uses minimizing the machine learning model loss function value as a training target; When the loss function value of the machine learning model is less than or equal to a preset target loss value, training is stopped. The machine learning model is one of a logistic regression model, a naive Bayes model or a support vector machine model.

7. The method for identifying power equipment faults based on deep learning according to claim 6, characterized in that: The core noise data is obtained as follows: A directional microphone is used to collect environmental noise signals; a circular disc microphone is used to collect a mixed sound signal of a core sound signal of an electric transformer and an environmental noise signal; an adaptive digital filter is used to filter the environmental noise signal and the mixed sound signal respectively to obtain a filtered environmental noise signal and a filtered mixed sound signal with the same intensity; a noise cancellation method is used to perform frequency spectrum difference on the filtered environmental noise signal and the filtered mixed sound signal to obtain a core sound signal of an electric transformer; an analog-to-digital converter is used to convert the core sound signal of an electric transformer into digital form of core sound data of an electric transformer, namely, core noise data.

8. The method for identifying power equipment faults based on deep learning according to claim 7, characterized in that: The setting method of the core normal state threshold U1 and the core minimum state threshold U2 is as follows: Under normal operation of the magnetic core, m groups of magnetic core data are collected and m first magnetic core state coefficients are calculated. Under normal operation of the magnetic core, m groups of oil tank data of power transformer equipment and magnetic core noise data are collected and m second magnetic core state coefficients are calculated. According to the m first magnetic core state coefficients and the m second magnetic core state coefficients, m third magnetic core state coefficients are calculated. The m third magnetic core state coefficients are traversed to find the third magnetic core state coefficient with the smallest value as the lowest magnetic core state threshold value U2, and the m third magnetic core state coefficients are traversed to find the third magnetic core state coefficient with the largest value as the highest magnetic core state threshold value U1; The fault information includes the core data of the electric power transformer equipment when the core state is determined to be faulty, the oil tank data of the electric power transformer equipment, the water content data of the oil and the core noise data.

9. A power equipment fault identification system based on deep learning, which is implemented based on the power equipment fault identification method based on deep learning as described in any one of claims 1 to 8, characterized in that: include: The first data acquisition module is used to collect magnetic core data and power transformer equipment oil tank data; Sending the magnetic core data to a first data preprocessing module; Sending the oil tank data of the power transformer equipment to the second preprocessing module; The second data acquisition module is used to collect historical oil quality data and magnetic core noise data; send the historical oil quality data to the machine training module; send the magnetic core noise data to the second data preprocessing module; The machine training module trains a machine learning model for predicting oil water content data based on historical oil quality data; The oil quality water content acquisition module inputs the real-time oil tank characteristic data into the trained machine learning model to obtain the oil quality water content data; and sends the oil quality water content data to the second data preprocessing module; The first data preprocessing module processes the magnetic core data to obtain a first magnetic core state coefficient A; and sends the first magnetic core state coefficient A to the magnetic core state analysis module; The second data preprocessing module processes the oil tank data and the magnetic core noise data of the power transformer equipment to obtain a second magnetic core state coefficient C; and sends the second magnetic core state coefficient C to the magnetic core state analysis module; The core state analysis module processes the first core state coefficient A and the second core state coefficient C to obtain a third core state coefficient W; determines whether the core state is normal or faulty according to the third core state coefficient; if the core state is determined to be faulty, generates fault information and reports it to the maintenance end; and each module is connected by wire and / or wireless means to realize data transmission between modules.

Images