High-voltage GIS switch cabinet fault early warning method and device based on data mining

Abstract

The present application is applicable to the technical field of high-voltage switchgear, and provides a method and device for early warning of high-voltage GIS switchgear faults based on data mining. The method includes: inputting the environmental information of the air chamber into the time series data generated into the LSTM prediction model, obtaining the predicted values ​​of the gas pressure and gas density of the air chamber, and generating the fault prediction results of the air chamber; inputting the working parameters of the opening and closing coils into the neural network model, and obtaining the fault prediction results of the opening coils and the closing coils; inputting the working parameters of the energy storage motor into the LSTM prediction model, and obtaining the predicted values ​​of the waveform similarity of the energy storage motor, and generating the fault prediction results of the energy storage motor; based on the fault prediction results of the air chamber, the fault prediction results of the opening coils and the closing coils, and the fault prediction results of the energy storage motor, the high-voltage GIS switchgear is given a fault early warning. The present application can comprehensively and more accurately give early warnings of faults in high-voltage GIS switchgear.

Description

Technical Field

[0001] The present application relates to the technical field of high-voltage switch cabinets, and in particular to a method and device for early warning of high-voltage GIS switch cabinet faults based on data mining. Background Art

[0002] The 35kVGIS switchgear of the urban rail transit power supply system is an important equipment for transforming and transmitting power to urban rail trains and stations. It plays the role of on-off, control or protection in the power generation, transmission, distribution, power conversion and consumption of the power system. Its stable operation is crucial to ensure the normal operation of the urban rail transit power supply system. However, due to the particularity of the high-voltage switchgear operating environment and the increase in workload, leakage, overload, overvoltage and other problems often occur, which poses a certain threat to the normal operation of the power system and the stability of the equipment. Due to the complexity and particularity of the high-voltage switchgear operating environment, once a fault or abnormal situation occurs, it often causes the interruption or even failure of the power system. Therefore, it is of great significance to monitor the high-voltage switchgear in real time and to promptly discover and deal with potential problems.

[0003] The traditional early warning method for the operating status of high-voltage switchgear is mainly based on early warning analysis based on factors such as temperature, fault maintenance records, and usage time, and it focuses more on the internal circuits of the switchgear. As a result, the early warning analysis of the high-voltage switchgear is not accurate and comprehensive, and it is easy to miss important possible faults. Summary of the invention

[0004] In view of this, an embodiment of the present application provides a high-voltage GIS switch cabinet fault warning method and device based on data mining, which can comprehensively and accurately warn of high-voltage GIS switch cabinet faults.

[0005] In order to achieve the above objectives, this application adopts the following technical solutions:

[0006] In a first aspect, an embodiment of the present application provides a high-voltage GIS switch cabinet fault warning method based on data mining, comprising:

[0007] Obtain environmental information of the busbar gas chamber and the circuit breaker gas chamber of the high-voltage GIS switch cabinet, the operating parameters of the opening and closing coils of the secondary components of the circuit breaker, and the operating parameters of the energy storage motor; wherein the environmental information includes the gas temperature, gas density and gas pressure in the gas chamber, the opening and closing coil operating parameters include the opening coil current, the closing coil current, the opening time and the closing time, and the operating parameters include the voltage, current and operating time;

[0008] The time series data generated by the environmental information is input into the LSTM prediction model to obtain the predicted values ​​of the gas pressure and gas density of the busbar gas chamber and the circuit breaker gas chamber within a preset time after the current moment, and the fault prediction results of the busbar gas chamber and the circuit breaker gas chamber are generated according to the predicted values ​​of the gas pressure and gas density of the busbar gas chamber and the circuit breaker gas chamber;

[0009] Inputting the opening coil current, the closing coil current, the opening time and the closing time into a neural network model to obtain fault prediction results of the opening coil and the closing coil;

[0010] The time series data of the energy storage motor waveform similarity generated by the working parameters is input into the LSTM prediction model to obtain the predicted value of the energy storage motor waveform similarity within a preset time after the current moment, and the fault prediction result of the energy storage motor is generated according to the predicted value of the energy storage motor waveform similarity; wherein the energy storage motor waveform similarity represents the similarity between the energy storage motor waveform obtained according to the working parameters and the theoretical energy storage motor waveform when the energy storage motor works normally;

[0011] Based on the fault prediction results of the busbar gas chamber and the circuit breaker gas chamber, the fault prediction results of the opening coil and the closing coil, and the fault prediction results of the energy storage motor, fault warning is given to the high-voltage GIS switchgear.

[0012] Based on the first aspect, in some embodiments, the step of inputting the environmental information generated time series data into the LSTM prediction model to obtain predicted values ​​of gas pressure and gas density of the busbar gas chamber and the circuit breaker gas chamber within a preset time after the current moment includes:

[0013] Generate a gas temperature time series of busbar gas chamber phase A, a gas density time series of busbar gas chamber phase A, and a gas pressure time series of busbar gas chamber phase A;

[0014] The gas temperature time series of the line gas chamber phase A, the gas density time series of the busbar gas chamber phase A and the gas pressure time series of the busbar gas chamber phase A are input into the first LSTM prediction model to obtain multiple predicted values ​​of the gas pressure of the busbar gas chamber phase A within a preset time after the current moment, and are input into the second LSTM prediction model to obtain multiple predicted values ​​of the gas density of the busbar gas chamber phase A within a preset time after the current moment;

[0015] Generate a gas temperature time series of the busbar gas chamber B phase, a gas density time series of the busbar gas chamber B phase, and a gas pressure time series of the busbar gas chamber B phase;

[0016] The gas temperature time series of the line gas chamber B phase, the gas density time series of the busbar gas chamber B phase and the gas pressure time series of the busbar gas chamber B phase are input into the third LSTM prediction model to obtain multiple predicted values ​​of the gas pressure of the busbar gas chamber B phase within a preset time after the current moment, and are input into the fourth LSTM prediction model to obtain multiple predicted values ​​of the gas density of the busbar gas chamber B phase within a preset time after the current moment;

[0017] Generate a gas temperature time series of the busbar gas chamber C phase, a gas density time series of the busbar gas chamber C phase, and a gas pressure time series of the busbar gas chamber C phase;

[0018] Input the gas temperature time series of the line gas chamber C phase, the gas density time series of the busbar gas chamber C phase and the gas pressure time series of the busbar gas chamber C phase into the fifth LSTM prediction model to obtain multiple predicted values ​​of the gas pressure of the busbar gas chamber C phase within a preset time after the current moment, and input them into the sixth LSTM prediction model to obtain multiple predicted values ​​of the gas density of the busbar gas chamber C phase within a preset time after the current moment;

[0019] Generate a gas temperature time series, a gas density time series and a gas pressure time series of the circuit breaker gas chamber;

[0020] The gas temperature time series, gas density time series and gas pressure time series of the circuit breaker gas chamber are input into the seventh LSTM prediction model to obtain multiple predicted values ​​of the gas density of the circuit breaker gas chamber within a preset time after the current moment, and are input into the eighth LSTM prediction model to obtain multiple predicted values ​​of the gas pressure of the circuit breaker gas chamber within a preset time after the current moment; wherein the gas is sulfur hexafluoride gas.

[0021] Based on the first aspect, in some embodiments, generating the fault prediction results of the busbar gas chamber and the circuit breaker gas chamber according to the predicted values ​​of the gas pressure and gas density of the busbar gas chamber and the circuit breaker gas chamber includes:

[0022] If there is a predicted value among the multiple predicted values ​​of the gas pressure of the busbar gas chamber phase A, the multiple predicted values ​​of the gas pressure of the busbar gas chamber phase B, and the multiple predicted values ​​of the gas pressure of the busbar gas chamber phase C that exceeds the first preset gas pressure range, a busbar gas chamber gas pressure warning signal is generated; wherein the busbar gas chamber gas pressure warning signal includes the busbar gas chamber number, the warning position is phase A, phase B or phase C, the predicted value of the gas pressure and the alarm time;

[0023] If a predicted value among the multiple predicted values ​​of the gas pressure of the circuit breaker gas chamber exceeds the second preset gas pressure range, a circuit breaker gas chamber gas pressure warning signal is generated; wherein the circuit breaker gas chamber gas pressure warning signal includes the circuit breaker gas chamber number, the warning location is the circuit breaker, the predicted value of the gas pressure and the alarm time;

[0024] If there is a predicted value among the multiple predicted values ​​of the gas density of the busbar gas chamber phase A, the multiple predicted values ​​of the gas density of the busbar gas chamber phase B and the multiple predicted values ​​of the gas density of the busbar gas chamber phase C that is greater than the first preset gas density, a busbar gas chamber gas density warning signal is generated; wherein the busbar gas chamber gas density warning signal includes the busbar gas chamber number, the warning location is phase A, phase B or phase C, the predicted value of the gas density and the alarm time; the first preset gas density is the average value of the gas density of the three phases of the busbar gas chamber corresponding to the historical maximum value of the gas density of phase B of the busbar gas chamber during normal operation;

[0025] If a predicted value among multiple predicted values ​​of the gas density of the circuit breaker gas chamber is greater than a second preset gas density, a circuit breaker gas chamber gas density warning signal is generated; wherein the circuit breaker gas chamber gas density warning signal includes the circuit breaker gas chamber number, the warning location is the circuit breaker, the predicted value of the gas density and the alarm time; the second preset gas density is the historical maximum value of the gas density of the circuit breaker gas chamber during normal operation.

[0026] Based on the first aspect, in some embodiments, inputting the opening coil current, the closing coil current, the opening time, and the closing time into a neural network model to obtain fault prediction results of the opening coil and the closing coil includes:

[0027] Sampling is performed to obtain the opening coil current, closing coil current, opening time and closing time, and generate a preliminary signal waveform;

[0028] The preliminary signal waveform is denoised through EMD decomposition and reconstruction to obtain a smoothed signal waveform;

[0029] Obtain a signal sequence in which the current is not zero from the smoothed signal waveform, and determine three extreme value points of the waveform for the signal sequence;

[0030] Determine the time characteristic values ​​and current characteristic values ​​corresponding to the three extreme points;

[0031] Calculate the mean μ, variance σ, skewness ξ and kurtosis δ of the signal waveform; where, N is the number of sampling points, X i is the current amplitude at the sampling point;

[0032] Generate a eigenvector from the time eigenvalue, the current eigenvalue, and the mean μ, variance σ, skewness ξ, and kurtosis δ of the signal waveform;

[0033] The characteristic vector is input into a neural network model to obtain fault prediction results of the opening coil and the closing coil.

[0034] Based on the first aspect, in some embodiments, inputting the feature vector into a neural network model to obtain fault prediction results of the opening coil and the closing coil includes:

[0035] Extracting principal component features from the feature vector using a PCA principal component dimensionality reduction algorithm to obtain an extracted feature vector;

[0036] Based on the historical opening and closing coil current waveform database, a neural network model for circuit breaker fault diagnosis is trained; wherein the historical opening and closing coil current waveform database includes sample data of the opening and closing coil current waveform under normal conditions, sample data of the opening and closing coil current waveform when the mechanism is stuck, and sample data of the opening and closing coil current waveform when the coil is aged;

[0037] The extracted feature vector is input into the trained neural network model to obtain the fault prediction results of the opening coil and closing coil, and generate a circuit breaker early warning signal, which includes the number of the opening coil or closing coil, the name of the secondary component of the circuit breaker, the predicted value and the alarm time.

[0038] Based on the first aspect, in some embodiments, the inputting the opening coil current, the closing coil current, the opening time and the closing time into a neural network model to obtain fault prediction results of the opening coil and the closing coil further includes:

[0039] If there are more than three extreme points in the waveform, all extreme points are divided into three groups according to the preset grouping interval, and the extreme point with the maximum value in the first group is taken as the extreme value of the first group, the extreme point with the minimum value in the second group is taken as the extreme value of the second group, and the extreme point with the maximum value in the third group is taken as the extreme value of the third group, so as to obtain three extreme points of the waveform.

[0040] Based on the first aspect, in some embodiments, the time series data of the energy storage motor waveform similarity generated by the working parameters is input into the LSTM prediction model to obtain a predicted value of the energy storage motor waveform similarity within a preset time after the current moment, including:

[0041] Calculating the energy storage motor waveform similarity between the energy storage motor waveform obtained by the working parameters and the theoretical energy storage motor waveform when the energy storage motor is working normally;

[0042] The ninth LSTM prediction model is trained using the historical time series of energy storage motor waveform similarity;

[0043] The time series of the energy storage motor waveform similarity at the current moment is input into the ninth LSTM prediction model to obtain multiple predicted values ​​of the energy storage motor waveform similarity within a preset time after the current moment.

[0044] Based on the first aspect, in some embodiments, generating a fault prediction result of the energy storage motor according to the predicted value of the energy storage motor waveform similarity includes:

[0045] If there is a predicted value among multiple predicted values ​​of the energy storage motor waveform similarity that is less than the similarity threshold, an energy storage motor warning signal is generated. The energy storage motor warning signal includes the energy storage motor number, the warning part is the energy storage motor, the predicted value of the energy storage motor waveform similarity and the alarm time.

[0046] Based on the first aspect, in some embodiments, the fault prediction results of the busbar gas chamber and the circuit breaker gas chamber, the fault prediction results of the opening coil and the closing coil, and the fault prediction results of the energy storage motor are used to provide a fault warning for the high-voltage GIS switch cabinet, including:

[0047] Based on the fault prediction results of the busbar gas chamber and the circuit breaker gas chamber, the fault prediction results of the opening coil and the closing coil, and the fault prediction results of the energy storage motor, a fault warning signal is sent to the client of the high-voltage GIS switchgear. The fault warning signal includes the number of the high-voltage GIS switchgear, the warning location of the high-voltage GIS switchgear, the predicted value of the fault parameter and the alarm time.

[0048] In a second aspect, the embodiment of the present application provides a high-voltage GIS switch cabinet fault warning device based on data mining, comprising:

[0049] An acquisition module is used to acquire environmental information of the busbar gas chamber and the circuit breaker gas chamber of the high-voltage GIS switch cabinet, operating parameters of the opening and closing coils of the secondary components of the circuit breaker, and operating parameters of the energy storage motor; wherein the environmental information includes the gas temperature, gas density and gas pressure in the gas chamber, the opening and closing coil operating parameters include the opening coil current, the closing coil current, the opening time and the closing time, and the operating parameters include the voltage, current and operating time;

[0050] A gas chamber prediction module, used to input the time series data generated by the environmental information into the LSTM prediction model, obtain the predicted values ​​of the gas pressure and gas density of the busbar gas chamber and the circuit breaker gas chamber within a preset time after the current moment, and generate the fault prediction results of the busbar gas chamber and the circuit breaker gas chamber according to the predicted values ​​of the gas pressure and gas density of the busbar gas chamber and the circuit breaker gas chamber;

[0051] A circuit breaker prediction module, used for inputting the opening coil current, the closing coil current, the opening time and the closing time into a neural network model to obtain fault prediction results of the opening coil and the closing coil;

[0052] An energy storage motor prediction module is used to input the time series data of the energy storage motor waveform similarity generated by the working parameters into the LSTM prediction model, obtain the predicted value of the energy storage motor waveform similarity within a preset time after the current moment, and generate a fault prediction result of the energy storage motor according to the predicted value of the energy storage motor waveform similarity; wherein the energy storage motor waveform similarity represents the similarity between the energy storage motor waveform obtained according to the working parameters and the theoretical energy storage motor waveform when the energy storage motor works normally;

[0053] The fault warning module is used to provide fault warning for the high-voltage GIS switchgear based on the fault prediction results of the busbar gas chamber and the circuit breaker gas chamber, the fault prediction results of the opening coil and the closing coil, and the fault prediction results of the energy storage motor.

[0054] The beneficial effects of the embodiments of the present application compared to the prior art include:

[0055] In an embodiment of the present application, environmental information of the busbar gas chamber and the circuit breaker gas chamber of the high-voltage GIS switch cabinet, the operating parameters of the opening and closing coils of the secondary components of the circuit breaker, and the operating parameters of the energy storage motor are obtained. Afterwards, the time series data generated by the environmental information is input into the LSTM prediction model to obtain the predicted values ​​of the gas pressure and gas density of the busbar gas chamber and the circuit breaker gas chamber, and then the predicted values ​​of the gas pressure and gas density are used to generate the fault prediction results of the busbar gas chamber and the circuit breaker gas chamber. In addition, the opening coil current, the closing coil current, the opening time and the closing time are input into the neural network model to obtain the fault prediction results of the opening coil and the closing coil. In addition, the time series data of the energy storage motor waveform similarity generated by the working parameters is input into the LSTM prediction model to obtain the predicted value of the energy storage motor waveform similarity, and then the fault prediction result of the energy storage motor is generated according to the predicted value of the energy storage motor waveform similarity. By predicting faults in the busbar gas chamber and circuit breaker gas chamber, circuit breaker and energy storage motor of the high-voltage GIS switchgear, including the pressure, temperature density of the gas chamber, circuit breaker opening and closing time, motor running time, motor waveform and other angles, it is possible to comprehensively and more accurately warn of faults in the high-voltage GIS switchgear. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0057] Figure 1 It is a flow chart of a high-voltage GIS switch cabinet fault warning method based on data mining provided in an embodiment of the present application;

[0058] Figure 2 It is a waveform diagram of the opening and closing current of components of the circuit breaker provided in the embodiment of the present application;

[0059] Figure 3 It is a structural schematic diagram of a high-voltage GIS switch cabinet fault warning device based on data mining provided in an embodiment of the present application. DETAILED DESCRIPTION

[0060] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application may also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present application.

[0061] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0062] It should also be understood that the term “and / or” used in the specification and appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0063] As used in the specification and appended claims of this application, the term "if" can be interpreted as "when" or "uponce" or "in response to determining" or "in response to detecting", depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" can be interpreted as meaning "uponce it is determined" or "in response to determining" or "uponce [described condition or event] is detected" or "in response to detecting [described condition or event]", depending on the context.

[0064] In addition, in the description of the present application specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0065] References to "one embodiment" or "some embodiments" etc. described in the specification of this application mean that one or more embodiments of the present application include specific features, structures or characteristics described in conjunction with the embodiment. Therefore, the statements "in one embodiment", "in some embodiments", "in some other embodiments", "in some other embodiments", etc. that appear in different places in this specification do not necessarily refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways. The terms "including", "comprising", "having" and their variations all mean "including but not limited to", unless otherwise specifically emphasized in other ways.

[0066] The 35kVGIS switchgear of the urban rail transit power supply system is an important equipment for transforming and transmitting power to urban rail trains and stations. The main components of the 35kVGIS switchgear include the switchgear air chamber, switchgear circuit breaker, and energy storage motor.

[0067] The switch cabinet gas chamber is mainly used to store SF6 (sulfur hexafluoride) gas, which is used for insulation between electrical components. If the SF6 gas parameters do not meet the standards, it may lead to poor insulation performance, affect the service life of the equipment, and even cause electric shock. Therefore, normal gas chamber parameters are the basis for ensuring the normal operation of the 35kV system.

[0068] Switch cabinet circuit breakers are mainly used to control the conduction and interruption of circuits. If the detection parameters are abnormal, it may cause malfunction, inadequate opening and closing, and cause electric shock accidents.

[0069] The energy storage motor is a component used to remotely control the opening and closing of the circuit breaker. If the state parameters of the energy storage motor are poor, it may cause the remote control function of the circuit breaker to fail and cause an electric shock accident.

[0070] Therefore, ensuring the normal parameters of the SF6 gas chamber of the switch cabinet can ensure that the electrical components will not have breakdown and leakage accidents; ensuring the normal parameters of the switch cabinet circuit breaker can ensure that the electrical opening and closing are in place and the opening and closing failure accidents will not occur; ensuring the normal parameters of the energy storage motor can ensure that the remote opening and closing are in place and the remote action failure accidents will not occur. The three components complement each other and are indispensable.

[0071] The embodiment of the present application is a technology based on industry standards and data mining algorithms, which detects and analyzes the switch cabinet busbar gas chamber, switch cabinet circuit breaker gas chamber, switch cabinet circuit breaker, and energy storage motor of the 35kVGIS switch cabinet, thereby achieving comprehensive and accurate early warning of the status and faults of the 35kVGIS switch cabinet.

[0072] The embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0073] Figure 1 is a flow chart of a high-voltage GIS switch cabinet fault warning method based on data mining provided by an embodiment of the present application, with reference to Figure 1 , the high-voltage GIS switchgear fault early warning method based on data mining is described in detail as follows:

[0074] Step 101, obtaining environmental information of the busbar gas chamber and the circuit breaker gas chamber of the high-voltage GIS switch cabinet, operating parameters of the opening and closing coils of the secondary components of the circuit breaker, and operating parameters of the energy storage motor.

[0075] Among them, the environmental information may include the gas temperature, gas density and gas pressure in the gas chamber, the opening and closing coil working parameters may include the opening coil current, the closing coil current, the opening time and the closing time, and the working parameters may include voltage, current and working time.

[0076] Step 102, input the time series data generated by the environmental information into the LSTM prediction model, obtain the predicted values ​​of the gas pressure and gas density of the busbar gas chamber and the circuit breaker gas chamber within a preset time after the current moment, and generate fault prediction results of the busbar gas chamber and the circuit breaker gas chamber according to the predicted values ​​of the gas pressure and gas density of the busbar gas chamber and the circuit breaker gas chamber.

[0077] In some embodiments, the above-mentioned inputting the environmental information generated time series data into the LSTM prediction model to obtain the predicted values ​​of the gas pressure and gas density of the busbar gas chamber and the circuit breaker gas chamber within a preset time after the current moment includes:

[0078] Generate a gas temperature time series of busbar gas chamber phase A, a gas density time series of busbar gas chamber phase A, and a gas pressure time series of busbar gas chamber phase A;

[0079] The gas temperature time series of the line gas chamber phase A, the gas density time series of the busbar gas chamber phase A and the gas pressure time series of the busbar gas chamber phase A are input into the first LSTM prediction model to obtain multiple predicted values ​​of the gas pressure of the busbar gas chamber phase A within a preset time after the current moment, and are input into the second LSTM prediction model to obtain multiple predicted values ​​of the gas density of the busbar gas chamber phase A within a preset time after the current moment;

[0080] Generate a gas temperature time series of the busbar gas chamber B phase, a gas density time series of the busbar gas chamber B phase, and a gas pressure time series of the busbar gas chamber B phase;

[0081] The gas temperature time series of the line gas chamber B phase, the gas density time series of the busbar gas chamber B phase and the gas pressure time series of the busbar gas chamber B phase are input into the third LSTM prediction model to obtain multiple predicted values ​​of the gas pressure of the busbar gas chamber B phase within a preset time after the current moment, and are input into the fourth LSTM prediction model to obtain multiple predicted values ​​of the gas density of the busbar gas chamber B phase within a preset time after the current moment;

[0082] Generate a gas temperature time series of the busbar gas chamber C phase, a gas density time series of the busbar gas chamber C phase, and a gas pressure time series of the busbar gas chamber C phase;

[0083] Input the gas temperature time series of the line gas chamber C phase, the gas density time series of the busbar gas chamber C phase and the gas pressure time series of the busbar gas chamber C phase into the fifth LSTM prediction model to obtain multiple predicted values ​​of the gas pressure of the busbar gas chamber C phase within a preset time after the current moment, and input them into the sixth LSTM prediction model to obtain multiple predicted values ​​of the gas density of the busbar gas chamber C phase within a preset time after the current moment;

[0084] Generate a gas temperature time series, a gas density time series and a gas pressure time series of the circuit breaker gas chamber;

[0085] The gas temperature time series, gas density time series and gas pressure time series of the circuit breaker gas chamber are input into the seventh LSTM prediction model to obtain multiple predicted values ​​of the gas density of the circuit breaker gas chamber within a preset time after the current moment, and are input into the eighth LSTM prediction model to obtain multiple predicted values ​​of the gas pressure of the circuit breaker gas chamber within a preset time after the current moment; wherein the gas is sulfur hexafluoride gas.

[0086] In some embodiments, the above-mentioned generating the fault prediction results of the busbar gas chamber and the circuit breaker gas chamber according to the predicted values ​​of the gas pressure and gas density of the busbar gas chamber and the circuit breaker gas chamber includes:

[0087] If there is a predicted value among the multiple predicted values ​​of the gas pressure of the busbar gas chamber phase A, the multiple predicted values ​​of the gas pressure of the busbar gas chamber phase B, and the multiple predicted values ​​of the gas pressure of the busbar gas chamber phase C that exceeds the first preset gas pressure range, a busbar gas chamber gas pressure warning signal is generated; wherein the busbar gas chamber gas pressure warning signal includes the busbar gas chamber number, the warning position is phase A, phase B or phase C, the predicted value of the gas pressure and the alarm time;

[0088] If a predicted value among the multiple predicted values ​​of the gas pressure of the circuit breaker gas chamber exceeds the second preset gas pressure range, a circuit breaker gas chamber gas pressure warning signal is generated; wherein the circuit breaker gas chamber gas pressure warning signal includes the circuit breaker gas chamber number, the warning location is the circuit breaker, the predicted value of the gas pressure and the alarm time;

[0089] If there is a predicted value among the multiple predicted values ​​of the gas density of the busbar gas chamber phase A, the multiple predicted values ​​of the gas density of the busbar gas chamber phase B and the multiple predicted values ​​of the gas density of the busbar gas chamber phase C that is greater than the first preset gas density, a busbar gas chamber gas density warning signal is generated; wherein the busbar gas chamber gas density warning signal includes the busbar gas chamber number, the warning location is phase A, phase B or phase C, the predicted value of the gas density and the alarm time; the first preset gas density is the average value of the gas density of the three phases of the busbar gas chamber corresponding to the historical maximum value of the gas density of phase B of the busbar gas chamber during normal operation;

[0090] If a predicted value among multiple predicted values ​​of the gas density of the circuit breaker gas chamber is greater than a second preset gas density, a circuit breaker gas chamber gas density warning signal is generated; wherein the circuit breaker gas chamber gas density warning signal includes the circuit breaker gas chamber number, the warning location is the circuit breaker, the predicted value of the gas density and the alarm time; the second preset gas density is the historical maximum value of the gas density of the circuit breaker gas chamber during normal operation.

[0091] In this step, the SF6 gas density and pressure content data of the switch cabinet are analyzed, and the SF6 gas density, pressure, and micro-water content curves of the busbar gas chamber and the circuit breaker gas chamber are analyzed and diagnosed on a daily, monthly, and yearly basis by site, so as to realize data display, abnormal state provision, over-limit alarm, and data trend warning. By analyzing and diagnosing the SF6 gas density and pressure curves of the busbar gas chamber and the circuit breaker gas chamber in real time, the abnormal state (gas leakage, abnormal pressure) can be identified and provided.

[0092] According to relevant regulations, the gas pressure of the circuit breaker chamber of the ground station is scored as 100 points in the range of 120kPa-135kPa; the gas pressure of the circuit breaker chamber of the underground station is scored as 100 points in the range of 110kPa-150kPa. Assume that the current SF6 gas pressure value of each phase of the busbar chamber is Pb_now, and the SF6 gas pressure value of the circuit breaker chamber is Pd_now. If Pb_now>150kPa, the gas pressure of each phase of the busbar chamber of the switch cabinet belongs to the abnormal gas pressure state, and the gas pressure of the circuit breaker chamber belongs to the abnormal gas pressure state; if Pb_now<110kPa, the phase of the busbar chamber of the switch cabinet belongs to the gas leakage state, and the gas pressure of the circuit breaker chamber belongs to the gas leakage state, otherwise it is normal.

[0093] In this step, the long short-term memory neural network (LSTM) is applied to the prediction of SF6 gas density and gas pressure in the busbar gas chamber and circuit breaker gas chamber. The historical time series data of equipment-related parameters are used as training samples to guide LSTM to learn the laws and connections of data in time and space, thereby completing trend prediction and obtaining the predicted values ​​corresponding to the subsequent time of the parameters.

[0094] The density, pressure and temperature of SF6 gas are closely related. Therefore, the time series data of the specified time step of the three parameters of gas density, gas pressure and gas temperature are used as LSTM input (3 feature vectors, multivariate input tasks). A total of 8 LSTM prediction models are trained (respectively predicting the SF6 gas density and gas pressure of the busbar gas chamber A, B and C, and the SF6 gas density and gas pressure of the circuit breaker gas chamber). The parameter value array (2 predicted values) of the relevant parameters 2 steps (1 step every 5 minutes) after the current moment is predicted. If there is a predicted value that exceeds the alarm threshold, an early warning signal is issued at the current moment. The specific input and output of LSTM are shown in Table 1.

[0095] Table 1 Input and output of LSTM for predicting gas pressure

[0096]

[0097]

[0098] If 110kPa≤the predicted value of the gas pressure of the busbar gas chamber≤150kPa, it means that the predicted value of the SF6 gas pressure of the busbar gas chamber is normal. If it exceeds the limit, a warning signal of the SF6 gas pressure of the busbar gas chamber is issued. If 110kPa≤the predicted value of the gas pressure of the circuit breaker gas chamber≤150kPa, it means that the predicted value of the SF6 gas pressure of the circuit breaker gas chamber is normal. If it exceeds the limit, a warning signal of the SF6 gas pressure of the circuit breaker gas chamber is issued.

[0099] The SF6 gas density warning limit of the busbar gas chamber is set to: when the busbar gas chamber B phase SF6 gas density reaches its maximum value during the peak period of normal operation, the average gas density of the three phases is Db_mar. The SF6 gas density warning limit of the circuit breaker gas chamber is set to: when the circuit breaker gas chamber SF6 gas density reaches its maximum value during the peak period of normal operation, Dd_m.

[0100] If the gas density of each phase of the busbar gas chamber is less than or equal to Db_mar, it means that the predicted value of the busbar gas chamber gas density is normal; if it is greater than Db_mar, a busbar gas chamber SF6 gas density warning signal is issued. If the gas density of the circuit breaker gas chamber is less than or equal to Dd_m, it means that the predicted value of the circuit breaker gas chamber gas density is normal; if it is greater than Dd_m, a circuit breaker gas chamber SF6 gas density warning signal is issued.

[0101] The above-mentioned warning signal may include the equipment number, equipment warning location and parameters, parameter prediction value, and the expected time of the alarm (if the prediction value has multiple alarms, the most recent time is used; the expected time of the alarm = current time + (the number of prediction steps where the warning value is located * 5min)). The input and output of the warning algorithm are shown in Table 2 below.

[0102] Table 2 Input and output of LSTM for predicting gas density

[0103]

[0104] Step 103, inputting the opening coil current, the closing coil current, the opening time and the closing time into a neural network model to obtain fault prediction results of the opening coil and the closing coil.

[0105] In some embodiments, step 103 may specifically include:

[0106] Sampling is performed to obtain the opening coil current, closing coil current, opening time and closing time, and generate a preliminary signal waveform;

[0107] The preliminary signal waveform is denoised through EMD decomposition and reconstruction to obtain a smoothed signal waveform;

[0108] Obtain a signal sequence in which the current is not zero from the smoothed signal waveform, and determine three extreme value points of the waveform for the signal sequence;

[0109] Determine the time characteristic values ​​and current characteristic values ​​corresponding to the three extreme points;

[0110] Calculate the mean μ, variance σ, skewness ξ and kurtosis δ of the signal waveform; where, N is the number of sampling points, X i is the current amplitude at the sampling point;

[0111] Generate a eigenvector from the time eigenvalue, the current eigenvalue, and the mean μ, variance σ, skewness ξ, and kurtosis δ of the signal waveform;

[0112] The characteristic vector is input into a neural network model to obtain fault prediction results of the opening coil and the closing coil.

[0113] Exemplarily, the above-mentioned inputting the feature vector into the neural network model to obtain the fault prediction results of the opening coil and the closing coil includes:

[0114] Extracting principal component features from the feature vector using a PCA principal component dimensionality reduction algorithm to obtain an extracted feature vector;

[0115] Based on the historical opening and closing coil current waveform database, a neural network model for circuit breaker fault diagnosis is trained; wherein the historical opening and closing coil current waveform database includes sample data of the opening and closing coil current waveform under normal conditions, sample data of the opening and closing coil current waveform when the mechanism is stuck, and sample data of the opening and closing coil current waveform when the coil is aged;

[0116] The extracted feature vector is input into the trained neural network model to obtain the fault prediction results of the opening coil and closing coil, and generate a circuit breaker early warning signal, which includes the number of the opening coil or closing coil, the name of the secondary component of the circuit breaker, the predicted value and the alarm time.

[0117] Exemplarily, if the number of extreme points in a waveform is greater than three, all extreme points are divided into three groups according to a preset grouping interval, and the extreme point with the maximum value in the first group is taken as the extreme value of the first group, the extreme point with the minimum value in the second group is taken as the extreme value of the second group, and the extreme point with the maximum value in the third group is taken as the extreme value of the third group, to obtain three extreme points of the waveform.

[0118] The current waveform and opening and closing time data of the circuit breaker's opening coil, closing coil and other secondary components are analyzed, and the opening coil and closing coil action waveforms are formed to complete data display, abnormal state judgment, and over-limit alarm functions, and realize mechanical and electrical fault diagnosis of the operating mechanism, such as mechanism jamming, coil aging, etc.

[0119] Based on relevant standards and operating specifications, set the limit values ​​of various detection parameters, and issue an alarm if the real-time monitoring parameter value exceeds the limit value. Obtain standard reference waveform samples through the factory data of the circuit breaker opening and closing coil current or the historical waveform database (the opening and closing coil current waveform when the circuit breaker is not faulty).

[0120] Specifically, when the current peak is greater than 10% of the reference value (the current peak in the standard sample), it indicates that there is inter-turn breakdown in the circuit breaker opening / closing control circuit coil; when the circuit breaker opening / closing control circuit current is zero, it indicates that the switch device opening / closing control circuit is open; when the circuit breaker opening / closing control circuit current is greater than the reference value (the current peak in the standard sample) and remains unchanged, it indicates that the circuit breaker opening / closing control circuit is short-circuited.

[0121] The circuit breaker coil is an important component of controlling the circuit breaker to open. The current signal in the coil contains a large amount of circuit breaker operation information, which can intuitively reflect the working status of the electromagnet coil current signal and the free tripping device. In the process of closing and opening, the action of the iron core is basically the same, but their touching objects are different. When closing, the iron core touches the traditional system of the connecting rod, and then the contact connected to it moves; when opening, the iron core touches the locking device, and the opening spring is used for the opening operation. Therefore, the closing and opening coil current waveform can reflect the action of the circuit breaker operating mechanism, from which it can be understood whether the circuit breaker has a fault.

[0122] According to the different characteristics of the opening and closing coil current, after data analysis and comparison with the characteristic value of the opening and closing coil current of the circuit breaker during normal operation, it is possible to determine whether the circuit breaker is faulty, thereby evaluating and classifying the health status. Different circuit breaker opening and closing coils have different current sizes, but they all have similar curve shapes. The normal opening and closing current waveform is as follows: Figure 2 shown.

[0123] In summary, by extracting 7 local features (time characteristic value t 1 ,t 2 ,t 3 ,t 4 , current characteristic value i 1 、i 2 、i 3 ) can evaluate the working status of the circuit breaker, thereby monitoring and diagnosing faults.

[0124] Considering that local features cannot fully reflect the difference of fault signals, the algorithm introduces global feature parameters (mean μ, variance σ, skewness ξ and kurtosis δ) that are representative of the overall characteristics of the signal to obtain the final feature vector f used for input, which is expressed as: f = {t 1 ,t 2 ,t 3 ,t 4 ,i 1 ,i 2 ,i 3 ,μ,σ,ξ,δ}.

[0125] Troubleshooting process:

[0126] Step 1: Extract the principal component features of the feature vector f through the PCA principal component dimensionality reduction algorithm, and use the extracted feature vector as the input of the fault diagnosis algorithm.

[0127] Step 2: Based on the provided historical opening and closing coil current waveform database (including sample data of three categories: normal, mechanism stuck, and coil aging), select the appropriate network structure and number of parameter matrices, and train to obtain a neural network model for circuit breaker fault diagnosis (normal, mechanism stuck, and coil aging).

[0128] Step 104, input the time series data of the energy storage motor waveform similarity generated by the working parameters into the LSTM prediction model, obtain the predicted value of the energy storage motor waveform similarity within a preset time after the current moment, and generate the fault prediction result of the energy storage motor according to the predicted value of the energy storage motor waveform similarity.

[0129] The energy storage motor waveform similarity represents the similarity between the energy storage motor waveform obtained according to the working parameters and the theoretical energy storage motor waveform when the energy storage motor works normally.

[0130] In some embodiments, the time series data of the energy storage motor waveform similarity generated by the working parameters is input into the LSTM prediction model to obtain the predicted value of the energy storage motor waveform similarity within a preset time after the current moment, including:

[0131] Calculating the energy storage motor waveform similarity between the energy storage motor waveform obtained by the working parameters and the theoretical energy storage motor waveform when the energy storage motor is working normally;

[0132] The ninth LSTM prediction model is trained using the historical time series of energy storage motor waveform similarity;

[0133] The time series of the energy storage motor waveform similarity at the current moment is input into the ninth LSTM prediction model to obtain multiple predicted values ​​of the energy storage motor waveform similarity within a preset time after the current moment.

[0134] In some embodiments, the above-mentioned generating the fault prediction result of the energy storage motor according to the predicted value of the energy storage motor waveform similarity includes:

[0135] If there is a predicted value among multiple predicted values ​​of the energy storage motor waveform similarity that is less than the similarity threshold, an energy storage motor warning signal is generated. The energy storage motor warning signal includes the energy storage motor number, the warning part is the energy storage motor, the predicted value of the energy storage motor waveform similarity and the alarm time.

[0136] The voltage, current and working time data of the energy storage motor are analyzed, and the action waveform and alarm status are formed to complete the functions of data display, abnormal state judgment, over-limit alarm, data trend warning, etc. The data input includes the energy storage motor running time of the 35kV incoming circuit breaker, the energy storage motor running current and the energy storage motor waveform similarity.

[0137] Based on relevant standards and operating department specifications, the limit values ​​of various detection parameters are set in the system. If the real-time monitoring parameter value exceeds the limit value, an alarm is issued. Assume that the current energy storage motor waveform similarity value is Se. If Se<60%, the energy storage motor is judged to be in an abnormal waveform state, and its related abnormal state information is output.

[0138] LSTM is applied to the prediction of energy storage motor waveform similarity. By taking the historical time series data of equipment waveform similarity parameters as training samples, LSTM is guided to learn the rules and connections of data in time and space, thereby completing trend prediction and obtaining the predicted values ​​corresponding to the subsequent time of the parameters.

[0139] The historical time series data of the energy storage motor waveform similarity parameters with a specified time step is used as the LSTM input (1 feature vector, single variable input task), and an LSTM prediction model is trained to predict the similarity value array (2 predicted values) of the energy storage motor waveform similarity 2 steps after the current moment (1 step every 5 minutes). If there is a predicted value that exceeds the alarm threshold, an early warning signal is issued at the current moment. The specific input of LSTM is the energy storage motor waveform similarity time series, and the output is the energy storage motor waveform similarity prediction value array.

[0140] Assume that the LSTM prediction value array is Spre. If all the prediction values ​​in Spre are ≥60%, it means that the prediction value is normal. If there is an over-limit, an energy storage motor warning signal is issued. The warning signal includes the equipment number, equipment warning location and parameters, parameter prediction value, and the expected time of the alarm (if the prediction value alarms at multiple points, the most recent time is taken; the expected alarm time = current time + (the number of prediction steps where the warning value is located * 5min)).

[0141] Step 105, based on the fault prediction results of the busbar gas chamber and the circuit breaker gas chamber, the fault prediction results of the opening coil and the closing coil, and the fault prediction result of the energy storage motor, a fault warning is provided for the high-voltage GIS switch cabinet.

[0142] Exemplarily, step 105 may specifically be: based on the fault prediction results of the busbar gas chamber and the circuit breaker gas chamber, the fault prediction results of the opening coil and the closing coil, and the fault prediction results of the energy storage motor, a fault warning signal is sent to the client of the high-voltage GIS switch cabinet, and the fault warning signal includes the number of the high-voltage GIS switch cabinet, the warning location of the high-voltage GIS switch cabinet, the predicted value of the fault parameter and the alarm time.

[0143] Specifically, if one or more of the busbar gas chamber gas pressure warning signal, circuit breaker gas chamber gas pressure warning signal, busbar gas chamber gas density warning signal, circuit breaker gas chamber gas density warning signal, circuit breaker warning signal and energy storage motor warning signal are received, various signals are sent to the corresponding client. For example, the relevant warning signal of the gas chamber can correspond to one client, the circuit breaker warning signal can correspond to one client, and the energy storage motor warning signal can correspond to one client, so as to facilitate different staff to check and repair. For another example, all the above-mentioned warning signals can correspond to a general client. When sending the warning signal to the corresponding client, it is also sent to the general client to facilitate the user to know the overall situation of the high-voltage switchgear.

[0144] The above-mentioned high-voltage GIS switchgear fault warning method based on data mining obtains the environmental information of the busbar gas chamber and the circuit breaker gas chamber of the high-voltage GIS switchgear, the working parameters of the opening and closing coils of the secondary components of the circuit breaker, and the working parameters of the energy storage motor. After that, the time series data generated by the environmental information is input into the LSTM prediction model to obtain the predicted values ​​of the gas pressure and gas density of the busbar gas chamber and the circuit breaker gas chamber, and then the predicted values ​​of the gas pressure and gas density are used to generate the fault prediction results of the busbar gas chamber and the circuit breaker gas chamber. In addition, the opening coil current, the closing coil current, the opening time and the closing time are input into the neural network model to obtain the fault prediction results of the opening coil and the closing coil. In addition, the time series data of the energy storage motor waveform similarity generated by the working parameters is input into the LSTM prediction model to obtain the predicted value of the energy storage motor waveform similarity, and then the fault prediction result of the energy storage motor is generated according to the predicted value of the energy storage motor waveform similarity. By predicting the faults of the busbar gas chamber and circuit breaker gas chamber, circuit breaker and energy storage motor of the high-voltage GIS switchgear, including the pressure, temperature density of the gas chamber, circuit breaker opening and closing time, motor running time, motor waveform, zero-sequence current and zero-sequence current imbalance, the faults of the high-voltage GIS switchgear can be warned comprehensively and more accurately.

[0145] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0146] Corresponding to the above-mentioned high-voltage GIS switch cabinet fault warning method based on data mining, the embodiment of the present application also provides a high-voltage GIS switch cabinet fault warning device based on data mining. Figure 3 The high-voltage GIS switch cabinet fault warning device based on data mining includes an acquisition module 201, a gas chamber prediction module 202, a circuit breaker prediction module 203, an energy storage motor prediction module 204 and a fault warning module 205.

[0147] The acquisition module 201 is used to obtain the environmental information of the busbar gas chamber and the circuit breaker gas chamber of the high-voltage GIS switch cabinet, the operating parameters of the opening and closing coils of the secondary components of the circuit breaker, and the operating parameters of the energy storage motor. The environmental information includes the gas temperature, gas density and gas pressure in the gas chamber, the opening and closing coil operating parameters include the opening coil current, the closing coil current, the opening time and the closing time, and the operating parameters include the voltage, current and working time.

[0148] The gas chamber prediction module 202 is used to input the time series data generated by the environmental information into the LSTM prediction model, obtain the predicted values ​​of the gas pressure and gas density of the busbar gas chamber and the circuit breaker gas chamber within a preset time after the current moment, and generate fault prediction results of the busbar gas chamber and the circuit breaker gas chamber according to the predicted values ​​of the gas pressure and gas density of the busbar gas chamber and the circuit breaker gas chamber.

[0149] The circuit breaker prediction module 203 is used to input the opening coil current, the closing coil current, the opening time and the closing time into a neural network model to obtain fault prediction results of the opening coil and the closing coil.

[0150] The energy storage motor prediction module 204 is used to input the time series data of the energy storage motor waveform similarity generated by the working parameters into the LSTM prediction model, obtain the predicted value of the energy storage motor waveform similarity within a preset time after the current moment, and generate the energy storage motor fault prediction result according to the predicted value of the energy storage motor waveform similarity. The energy storage motor waveform similarity represents the similarity between the energy storage motor waveform obtained according to the working parameters and the theoretical energy storage motor waveform when the energy storage motor works normally.

[0151] The fault warning module 205 is used to provide fault warning for the high-voltage GIS switch cabinet based on the fault prediction results of the busbar gas chamber and the circuit breaker gas chamber, the fault prediction results of the opening coil and the closing coil, and the fault prediction results of the energy storage motor.

[0152] Optionally, the air chamber prediction module 202 is specifically used for:

[0153] Generate a gas temperature time series of busbar gas chamber phase A, a gas density time series of busbar gas chamber phase A, and a gas pressure time series of busbar gas chamber phase A;

[0154] The gas temperature time series of the line gas chamber phase A, the gas density time series of the busbar gas chamber phase A and the gas pressure time series of the busbar gas chamber phase A are input into the first LSTM prediction model to obtain multiple predicted values ​​of the gas pressure of the busbar gas chamber phase A within a preset time after the current moment, and are input into the second LSTM prediction model to obtain multiple predicted values ​​of the gas density of the busbar gas chamber phase A within a preset time after the current moment;

[0155] Generate a gas temperature time series of the busbar gas chamber B phase, a gas density time series of the busbar gas chamber B phase, and a gas pressure time series of the busbar gas chamber B phase;

[0156] The gas temperature time series of the line gas chamber B phase, the gas density time series of the busbar gas chamber B phase and the gas pressure time series of the busbar gas chamber B phase are input into the third LSTM prediction model to obtain multiple predicted values ​​of the gas pressure of the busbar gas chamber B phase within a preset time after the current moment, and are input into the fourth LSTM prediction model to obtain multiple predicted values ​​of the gas density of the busbar gas chamber B phase within a preset time after the current moment;

[0157] Generate a gas temperature time series of the busbar gas chamber C phase, a gas density time series of the busbar gas chamber C phase, and a gas pressure time series of the busbar gas chamber C phase;

[0158] Input the gas temperature time series of the line gas chamber C phase, the gas density time series of the busbar gas chamber C phase and the gas pressure time series of the busbar gas chamber C phase into the fifth LSTM prediction model to obtain multiple predicted values ​​of the gas pressure of the busbar gas chamber C phase within a preset time after the current moment, and input them into the sixth LSTM prediction model to obtain multiple predicted values ​​of the gas density of the busbar gas chamber C phase within a preset time after the current moment;

[0159] Generate a gas temperature time series, a gas density time series and a gas pressure time series of the circuit breaker gas chamber;

[0160] The gas temperature time series, gas density time series and gas pressure time series of the circuit breaker gas chamber are input into the seventh LSTM prediction model to obtain multiple predicted values ​​of the gas density of the circuit breaker gas chamber within a preset time after the current moment, and are input into the eighth LSTM prediction model to obtain multiple predicted values ​​of the gas pressure of the circuit breaker gas chamber within a preset time after the current moment; wherein the gas is sulfur hexafluoride gas.

[0161] Optionally, the air chamber prediction module 202 is specifically used for:

[0162] If there is a predicted value among the multiple predicted values ​​of the gas pressure of the busbar gas chamber phase A, the multiple predicted values ​​of the gas pressure of the busbar gas chamber phase B, and the multiple predicted values ​​of the gas pressure of the busbar gas chamber phase C that exceeds the first preset gas pressure range, a busbar gas chamber gas pressure warning signal is generated; wherein the busbar gas chamber gas pressure warning signal includes the busbar gas chamber number, the warning position is phase A, phase B or phase C, the predicted value of the gas pressure and the alarm time;

[0163] If a predicted value among the multiple predicted values ​​of the gas pressure of the circuit breaker gas chamber exceeds the second preset gas pressure range, a circuit breaker gas chamber gas pressure warning signal is generated; wherein the circuit breaker gas chamber gas pressure warning signal includes the circuit breaker gas chamber number, the warning location is the circuit breaker, the predicted value of the gas pressure and the alarm time;

[0164] If there is a predicted value among the multiple predicted values ​​of the gas density of the busbar gas chamber phase A, the multiple predicted values ​​of the gas density of the busbar gas chamber phase B and the multiple predicted values ​​of the gas density of the busbar gas chamber phase C that is greater than the first preset gas density, a busbar gas chamber gas density warning signal is generated; wherein the busbar gas chamber gas density warning signal includes the busbar gas chamber number, the warning location is phase A, phase B or phase C, the predicted value of the gas density and the alarm time; the first preset gas density is the average value of the gas density of the three phases of the busbar gas chamber corresponding to the historical maximum value of the gas density of phase B of the busbar gas chamber during normal operation;

[0165] If a predicted value among multiple predicted values ​​of the gas density of the circuit breaker gas chamber is greater than a second preset gas density, a circuit breaker gas chamber gas density warning signal is generated; wherein the circuit breaker gas chamber gas density warning signal includes the circuit breaker gas chamber number, the warning location is the circuit breaker, the predicted value of the gas density and the alarm time; the second preset gas density is the historical maximum value of the gas density of the circuit breaker gas chamber during normal operation.

[0166] Optionally, the circuit breaker prediction module 203 is specifically used for:

[0167] Sampling is performed to obtain the opening coil current, closing coil current, opening time and closing time, and generate a preliminary signal waveform;

[0168] The preliminary signal waveform is denoised through EMD decomposition and reconstruction to obtain a smoothed signal waveform;

[0169] Obtain a signal sequence in which the current is not zero from the smoothed signal waveform, and determine three extreme value points of the waveform for the signal sequence;

[0170] Determine the time characteristic values ​​and current characteristic values ​​corresponding to the three extreme points;

[0171] Calculate the mean μ, variance σ, skewness ξ and kurtosis δ of the signal waveform; where, N is the number of sampling points, X i is the current amplitude at the sampling point;

[0172] Generate a eigenvector from the time eigenvalue, the current eigenvalue, and the mean μ, variance σ, skewness ξ, and kurtosis δ of the signal waveform;

[0173] The characteristic vector is input into a neural network model to obtain fault prediction results of the opening coil and the closing coil.

[0174] Optionally, the inputting the feature vector into a neural network model to obtain fault prediction results of the opening coil and the closing coil specifically includes:

[0175] Extracting principal component features from the feature vector using a PCA principal component dimensionality reduction algorithm to obtain an extracted feature vector;

[0176] Based on the historical opening and closing coil current waveform database, a neural network model for circuit breaker fault diagnosis is trained; wherein the historical opening and closing coil current waveform database includes sample data of the opening and closing coil current waveform under normal conditions, sample data of the opening and closing coil current waveform when the mechanism is stuck, and sample data of the opening and closing coil current waveform when the coil is aged;

[0177] The extracted feature vector is input into the trained neural network model to obtain the fault prediction results of the opening coil and closing coil, and generate a circuit breaker early warning signal, which includes the number of the opening coil or closing coil, the name of the secondary component of the circuit breaker, the predicted value and the alarm time.

[0178] Optionally, the circuit breaker prediction module 203 can also be used for: if the number of extreme points in the waveform is greater than three, all extreme points are divided into three groups according to a preset grouping interval, the extreme point with the maximum value in the first group is used as the extreme value of the first group, the extreme point with the minimum value in the second group is used as the extreme value of the second group, and the extreme point with the maximum value in the third group is used as the extreme value of the third group, so as to obtain three extreme points of the waveform.

[0179] Optionally, the energy storage electrode prediction module 204 is specifically used for:

[0180] Calculating the energy storage motor waveform similarity between the energy storage motor waveform obtained by the working parameters and the theoretical energy storage motor waveform when the energy storage motor is working normally;

[0181] The ninth LSTM prediction model is trained using the historical time series of energy storage motor waveform similarity;

[0182] The time series of the energy storage motor waveform similarity at the current moment is input into the ninth LSTM prediction model to obtain multiple predicted values ​​of the energy storage motor waveform similarity within a preset time after the current moment.

[0183] Optionally, the energy storage electrode prediction module 204 is specifically used to: if there is a predicted value among multiple predicted values ​​of the energy storage motor waveform similarity that is less than the similarity threshold, then generate an energy storage motor warning signal, and the energy storage motor warning signal includes the energy storage motor number, the warning part is the energy storage motor, the predicted value of the energy storage motor waveform similarity and the alarm time.

[0184] Optionally, the fault warning module 205 is specifically used to send a fault warning signal to the client of the high-voltage GIS switch cabinet based on the fault prediction results of the busbar gas chamber and the circuit breaker gas chamber, the fault prediction results of the opening coil and the closing coil, and the fault prediction results of the energy storage motor. The fault warning signal includes the number of the high-voltage GIS switch cabinet, the warning location of the high-voltage GIS switch cabinet, the predicted value of the fault parameter and the alarm time.

[0185] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A high-voltage GIS switch cabinet fault early warning method based on data mining, characterized in that: include: Obtain environmental information of the busbar gas chamber and the circuit breaker gas chamber of the high-voltage GIS switch cabinet, the operating parameters of the opening and closing coils of the secondary components of the circuit breaker, and the operating parameters of the energy storage motor; wherein the environmental information includes the gas temperature, gas density and gas pressure in the gas chamber, the opening and closing coil operating parameters include the opening coil current, the closing coil current, the opening time and the closing time, and the operating parameters include the voltage, current and operating time; The time series data generated by the environmental information is input into the LSTM prediction model to obtain the predicted values ​​of the gas pressure and gas density of the busbar gas chamber and the circuit breaker gas chamber within a preset time after the current moment, and the fault prediction results of the busbar gas chamber and the circuit breaker gas chamber are generated according to the predicted values ​​of the gas pressure and gas density of the busbar gas chamber and the circuit breaker gas chamber; Inputting the opening coil current, the closing coil current, the opening time and the closing time into a neural network model to obtain fault prediction results of the opening coil and the closing coil; The time series data of the energy storage motor waveform similarity generated by the working parameters is input into the LSTM prediction model to obtain the predicted value of the energy storage motor waveform similarity within a preset time after the current moment, and the fault prediction result of the energy storage motor is generated according to the predicted value of the energy storage motor waveform similarity; wherein the energy storage motor waveform similarity represents the similarity between the energy storage motor waveform obtained according to the working parameters and the theoretical energy storage motor waveform when the energy storage motor works normally; Based on the fault prediction results of the busbar gas chamber and the circuit breaker gas chamber, the fault prediction results of the opening coil and the closing coil, and the fault prediction results of the energy storage motor, the high-voltage GIS switchgear is given a fault warning; The step of inputting the opening coil current, the closing coil current, the opening time and the closing time into a neural network model to obtain fault prediction results of the opening coil and the closing coil includes: Sampling is performed to obtain the opening coil current, closing coil current, opening time and closing time, and generate a preliminary signal waveform; The preliminary signal waveform is denoised through EMD decomposition and reconstruction to obtain a smoothed signal waveform; Obtain a signal sequence in which the current is not zero from the smoothed signal waveform, and determine three extreme value points of the waveform for the signal sequence; Determine the time characteristic values ​​and current characteristic values ​​corresponding to the three extreme points; Calculate the mean μ, variance σ, skewness ξ and kurtosis δ of the signal waveform; where, i=1,2,…,N, N is the number of sampling points, X i is the current amplitude at the sampling point; Generate a eigenvector from the time eigenvalue, the current eigenvalue, and the mean μ, variance σ, skewness ξ, and kurtosis δ of the signal waveform; The characteristic vector is input into a neural network model to obtain fault prediction results of the opening coil and the closing coil.

2. The high-voltage GIS switch cabinet fault early warning method based on data mining according to claim 1 is characterized in that: The step of inputting the time series data generated by the environmental information into the LSTM prediction model to obtain the predicted values ​​of the gas pressure and gas density of the busbar gas chamber and the circuit breaker gas chamber within a preset time after the current moment includes: Generate a gas temperature time series of busbar gas chamber phase A, a gas density time series of busbar gas chamber phase A, and a gas pressure time series of busbar gas chamber phase A; The gas temperature time series of busbar gas chamber phase A, the gas density time series of busbar gas chamber phase A and the gas pressure time series of busbar gas chamber phase A are input into the first LSTM prediction model to obtain multiple predicted values ​​of the gas pressure of busbar gas chamber phase A within a preset time after the current moment, and are input into the second LSTM prediction model to obtain multiple predicted values ​​of the gas density of busbar gas chamber phase A within a preset time after the current moment; Generate a gas temperature time series of the busbar gas chamber B phase, a gas density time series of the busbar gas chamber B phase, and a gas pressure time series of the busbar gas chamber B phase; The gas temperature time series of the busbar gas chamber B phase, the gas density time series of the busbar gas chamber B phase and the gas pressure time series of the busbar gas chamber B phase are input into the third LSTM prediction model to obtain multiple predicted values ​​of the gas pressure of the busbar gas chamber B phase within a preset time after the current moment, and are input into the fourth LSTM prediction model to obtain multiple predicted values ​​of the gas density of the busbar gas chamber B phase within a preset time after the current moment; Generate a gas temperature time series of the busbar gas chamber C phase, a gas density time series of the busbar gas chamber C phase, and a gas pressure time series of the busbar gas chamber C phase; The gas temperature time series of the busbar gas chamber C phase, the gas density time series of the busbar gas chamber C phase and the gas pressure time series of the busbar gas chamber C phase are input into the fifth LSTM prediction model to obtain multiple predicted values ​​of the gas pressure of the busbar gas chamber C phase within a preset time after the current moment, and are input into the sixth LSTM prediction model to obtain multiple predicted values ​​of the gas density of the busbar gas chamber C phase within a preset time after the current moment; Generate a gas temperature time series, a gas density time series and a gas pressure time series of the circuit breaker gas chamber; The gas temperature time series, gas density time series and gas pressure time series of the circuit breaker gas chamber are input into the seventh LSTM prediction model to obtain multiple predicted values ​​of the gas density of the circuit breaker gas chamber within a preset time after the current moment, and are input into the eighth LSTM prediction model to obtain multiple predicted values ​​of the gas pressure of the circuit breaker gas chamber within a preset time after the current moment; wherein the gas is sulfur hexafluoride gas.

3. The high-voltage GIS switch cabinet fault early warning method based on data mining according to claim 2 is characterized in that: The generating of the fault prediction results of the busbar gas chamber and the circuit breaker gas chamber according to the predicted values ​​of the gas pressure and the gas density of the busbar gas chamber and the circuit breaker gas chamber comprises: If there is a predicted value among the multiple predicted values ​​of the gas pressure of the busbar gas chamber phase A, the multiple predicted values ​​of the gas pressure of the busbar gas chamber phase B, and the multiple predicted values ​​of the gas pressure of the busbar gas chamber phase C that exceeds the first preset gas pressure range, a busbar gas chamber gas pressure warning signal is generated; wherein the busbar gas chamber gas pressure warning signal includes the busbar gas chamber number, the warning position is phase A, phase B or phase C, the predicted value of the gas pressure and the alarm time; If a predicted value among the multiple predicted values ​​of the gas pressure of the circuit breaker gas chamber exceeds the second preset gas pressure range, a circuit breaker gas chamber gas pressure warning signal is generated; wherein the circuit breaker gas chamber gas pressure warning signal includes the circuit breaker gas chamber number, the warning location is the circuit breaker, the predicted value of the gas pressure and the alarm time; If there is a predicted value among the multiple predicted values ​​of the gas density of the busbar gas chamber phase A, the multiple predicted values ​​of the gas density of the busbar gas chamber phase B and the multiple predicted values ​​of the gas density of the busbar gas chamber phase C that is greater than the first preset gas density, a busbar gas chamber gas density warning signal is generated; wherein the busbar gas chamber gas density warning signal includes the busbar gas chamber number, the warning location is phase A, phase B or phase C, the predicted value of the gas density and the alarm time; the first preset gas density is the average value of the gas density of the three phases of the busbar gas chamber corresponding to the historical maximum value of the gas density of phase B of the busbar gas chamber during normal operation; If a predicted value among multiple predicted values ​​of the gas density of the circuit breaker gas chamber is greater than a second preset gas density, a circuit breaker gas chamber gas density warning signal is generated; wherein the circuit breaker gas chamber gas density warning signal includes the circuit breaker gas chamber number, the warning location is the circuit breaker, the predicted value of the gas density and the alarm time; the second preset gas density is the historical maximum value of the gas density of the circuit breaker gas chamber during normal operation.

4. The high-voltage GIS switch cabinet fault early warning method based on data mining according to claim 1 is characterized in that: The step of inputting the feature vector into a neural network model to obtain fault prediction results of the opening coil and the closing coil includes: Extracting principal component features from the feature vector using a PCA principal component dimensionality reduction algorithm to obtain an extracted feature vector; Based on the historical opening and closing coil current waveform database, a neural network model for circuit breaker fault diagnosis is trained; wherein the historical opening and closing coil current waveform database includes sample data of the opening and closing coil current waveform under normal conditions, sample data of the opening and closing coil current waveform when the mechanism is stuck, and sample data of the opening and closing coil current waveform when the coil is aged; The extracted feature vector is input into the trained neural network model to obtain the fault prediction results of the opening coil and closing coil, and generate a circuit breaker early warning signal, which includes the number of the opening coil or closing coil, the name of the secondary component of the circuit breaker, the predicted value and the alarm time.

5. The high-voltage GIS switch cabinet fault early warning method based on data mining according to claim 1 is characterized in that: The step of inputting the opening coil current, the closing coil current, the opening time and the closing time into a neural network model to obtain fault prediction results of the opening coil and the closing coil further includes: If there are more than three extreme points in the waveform, all extreme points are divided into three groups according to the preset grouping interval, and the extreme point with the maximum value in the first group is taken as the extreme value of the first group, the extreme point with the minimum value in the second group is taken as the extreme value of the second group, and the extreme point with the maximum value in the third group is taken as the extreme value of the third group, so as to obtain three extreme points of the waveform.

6. The high-voltage GIS switch cabinet fault early warning method based on data mining according to claim 1 is characterized in that: The time series data of the energy storage motor waveform similarity generated by the working parameters is input into the LSTM prediction model to obtain the predicted value of the energy storage motor waveform similarity within a preset time after the current moment, including: Calculating the energy storage motor waveform similarity between the energy storage motor waveform obtained by the working parameters and the theoretical energy storage motor waveform when the energy storage motor is working normally; The ninth LSTM prediction model is trained using the historical time series of energy storage motor waveform similarity; The time series of the energy storage motor waveform similarity at the current moment is input into the ninth LSTM prediction model to obtain multiple predicted values ​​of the energy storage motor waveform similarity within a preset time after the current moment.

7. The high-voltage GIS switch cabinet fault early warning method based on data mining according to claim 6 is characterized in that: The method of generating a fault prediction result of the energy storage motor according to the predicted value of the energy storage motor waveform similarity includes: If there is a predicted value among multiple predicted values ​​of the energy storage motor waveform similarity that is less than the similarity threshold, an energy storage motor warning signal is generated. The energy storage motor warning signal includes the energy storage motor number, the warning part is the energy storage motor, the predicted value of the energy storage motor waveform similarity and the alarm time.

8. The high-voltage GIS switch cabinet fault early warning method based on data mining according to claim 1 is characterized in that: The method of providing a fault warning for the high-voltage GIS switch cabinet based on the fault prediction results of the busbar gas chamber and the circuit breaker gas chamber, the fault prediction results of the opening coil and the closing coil, and the fault prediction results of the energy storage motor includes: Based on the fault prediction results of the busbar gas chamber and the circuit breaker gas chamber, the fault prediction results of the opening coil and the closing coil, and the fault prediction results of the energy storage motor, a fault warning signal is sent to the client of the high-voltage GIS switchgear. The fault warning signal includes the number of the high-voltage GIS switchgear, the warning location of the high-voltage GIS switchgear, the predicted value of the fault parameter and the alarm time.

9. A high-voltage GIS switch cabinet fault warning device based on data mining, characterized in that: include: An acquisition module is used to acquire environmental information of the busbar gas chamber and the circuit breaker gas chamber of the high-voltage GIS switch cabinet, operating parameters of the opening and closing coils of the secondary components of the circuit breaker, and operating parameters of the energy storage motor; wherein the environmental information includes the gas temperature, gas density and gas pressure in the gas chamber, the opening and closing coil operating parameters include the opening coil current, the closing coil current, the opening time and the closing time, and the operating parameters include the voltage, current and operating time; A gas chamber prediction module, used to input the time series data generated by the environmental information into the LSTM prediction model, obtain the predicted values ​​of the gas pressure and gas density of the busbar gas chamber and the circuit breaker gas chamber within a preset time after the current moment, and generate the fault prediction results of the busbar gas chamber and the circuit breaker gas chamber according to the predicted values ​​of the gas pressure and gas density of the busbar gas chamber and the circuit breaker gas chamber; A circuit breaker prediction module, used for inputting the opening coil current, the closing coil current, the opening time and the closing time into a neural network model to obtain fault prediction results of the opening coil and the closing coil; An energy storage motor prediction module is used to input the time series data of the energy storage motor waveform similarity generated by the working parameters into the LSTM prediction model, obtain the predicted value of the energy storage motor waveform similarity within a preset time after the current moment, and generate a fault prediction result of the energy storage motor according to the predicted value of the energy storage motor waveform similarity; wherein the energy storage motor waveform similarity represents the similarity between the energy storage motor waveform obtained according to the working parameters and the theoretical energy storage motor waveform when the energy storage motor works normally; Fault warning module, used to provide fault warning for high-voltage GIS switchgear based on the fault prediction results of busbar gas chamber and circuit breaker gas chamber, the fault prediction results of opening coil and closing coil, and the fault prediction results of energy storage motor; Wherein, the circuit breaker prediction module is specifically used for: Sampling is performed to obtain the opening coil current, closing coil current, opening time and closing time, and generate a preliminary signal waveform; The preliminary signal waveform is denoised through EMD decomposition and reconstruction to obtain a smoothed signal waveform; Obtain a signal sequence in which the current is not zero from the smoothed signal waveform, and determine three extreme value points of the waveform for the signal sequence; Determine the time characteristic values ​​and current characteristic values ​​corresponding to the three extreme points; Calculate the mean μ, variance σ, skewness ξ and kurtosis δ of the signal waveform; where, i=1,2,…,N, N is the number of sampling points, X i is the current amplitude at the sampling point; Generate a eigenvector from the time eigenvalue, the current eigenvalue, and the mean μ, variance σ, skewness ξ, and kurtosis δ of the signal waveform; The characteristic vector is input into a neural network model to obtain fault prediction results of the opening coil and the closing coil.

Images