Air discharge decomposition product detection model training method and detection method

Abstract

The invention provides an air discharge decomposition product detection model training method and a detection method, and belongs to the field of power equipment state detection. The method comprises the following steps: acquiring normal working condition parameters, constructing a normal data set according to the normal working condition parameters, acquiring fault working condition parameters, and constructing a fault data set according to the fault working condition parameters; wherein the air chamber is filled with air under the normal working condition, and the normal working condition parameters comprise temperature data, humidity data and voltage data of the gas sensor array; target air discharge decomposers are in the air chamber under the fault working condition, and the fault working condition parameters comprise temperature data, humidity data, voltage data of the gas sensor array and the concentration of the target air discharge decomposers; constructing an integrated data set according to the normal data set and the fault data set; and training according to the integrated data set to obtain an air discharge decomposer detection model. According to the invention, the target air discharge decomposer can be accurately detected.

Description

Technical Field

[0001] The present invention relates to the technical field of power equipment state detection, and in particular to an air discharge decomposition product detection model training method and a detection method. Background Art

[0002] During long-term operation, power equipment is prone to partial discharge and other faults caused by insulation defects, which leads to the decomposition of the gas insulation medium inside the equipment. As the most common natural insulating gas, air is widely used in power equipment such as switchgear to ensure insulation performance. When a discharge fault occurs in air-insulated equipment, the solid air insulation medium inside it reacts to generate a variety of gas decomposition products represented by NO2, O3, and CO. Studies have shown that the composition and content of air discharge decomposition products are closely related to the type and severity of equipment failure. Therefore, the fault diagnosis of air-insulated power equipment can be achieved by detecting the composition and content of air discharge decomposition products.

[0003] In the prior art, such as the patent with application number 202310726978.5, when detecting the insulation status of the air cabinet, the influence of the air pressure, temperature and humidity outside the air cabinet on the air cabinet is considered, but the influence of the environment inside the air cabinet on the air cabinet is not considered, resulting in inaccurate detection of air cabinet faults. Summary of the invention

[0004] The embodiments of the present invention provide an air discharge decomposition product detection model training method and a detection method to solve the problem of accurately detecting target air discharge decomposition products.

[0005] In a first aspect, an embodiment of the present invention provides a method for training an air discharge decomposition product detection model, comprising:

[0006] Acquire normal operating condition parameters, construct a normal data set according to the normal operating condition parameters, and acquire fault operating condition parameters, construct a fault data set according to the fault operating condition parameters; wherein, under the normal operating condition, the air chamber contains air, and the normal operating condition parameters include temperature data, humidity data, and voltage data of a gas sensor array; under the fault operating condition, the air chamber contains target air discharge decomposition products, and the fault operating condition parameters include temperature data, humidity data, voltage data of a gas sensor array, and concentration of target air discharge decomposition products;

[0007] An integrated data set is constructed based on the normal data set and the fault data set; an air discharge decomposition product detection model is trained based on the integrated data set; wherein the air discharge decomposition product detection model takes the voltage data, temperature data and humidity data of the gas sensor array as input, and takes the probability and concentration of the target air discharge decomposition product as output.

[0008] In a possible implementation, constructing an integrated data set according to the normal data set and the fault data set includes:

[0009] For normal working conditions and fault working conditions respectively, the data in the normal data set and the fault data set are classified according to different temperatures, different humidities or different voltage data, and sub-data sets corresponding to different temperature, different humidity or different voltage data are constructed.

[0010] In a possible implementation, under normal operating conditions, the probability and concentration of the target air discharge decomposition product are both zero.

[0011] In a possible implementation, the gas sensor array includes multiple groups of gas sensors, each group of gas sensors detects different gases, and each gas sensor in each group has a different sensitivity to the same gas.

[0012] In a possible implementation manner, the correlation between the voltage data of different gas sensors in each group is lower than a correlation threshold.

[0013] In a possible implementation, the calculation formula of the correlation is as follows:

[0014]

[0015] Where R1 and R2 are the voltage data of two gas sensors measuring the same gas with the same concentration, D is the variance, and Cov is the covariance.

[0016] In a possible implementation, the target air discharge decomposition products include NO2, O3 and CO.

[0017] In a possible implementation, the air discharge decomposition product detection model is an LSTM model.

[0018] In a second aspect, an embodiment of the present invention provides an air discharge decomposition product detection method, which performs detection based on an air discharge decomposition product detection model obtained by the above-mentioned air discharge decomposition product detection model training method, and is characterized in that it includes:

[0019] Acquire voltage data, temperature data and humidity data of the gas sensor array;

[0020] The voltage data, the temperature data and the humidity data of the gas sensor array are input into the air discharge decomposition product detection model, and the probability and concentration of the target air discharge decomposition product are output.

[0021] In a possible implementation, when the probability of the target air discharge decomposition product is greater than a target air discharge decomposition product threshold, it is determined that the target air discharge decomposition product is detected.

[0022] The embodiment of the present invention provides a method for training and detecting an air discharge decomposition product detection model, by obtaining normal operating condition parameters, constructing a normal data set according to the normal operating condition parameters, obtaining fault operating condition parameters, constructing a fault data set according to the fault operating condition parameters, constructing an integrated data set according to the normal data set and the fault data set, and training an air discharge decomposition product detection model according to the integrated data set. The present invention considers the normal operating condition and the fault operating condition, that is, when the power equipment is operating normally, there is only air in the air chamber and no air discharge decomposition products are generated, and when the equipment fails, air decomposition products are generated in the air chamber. The data of the normal operating condition and the data of the fault operating condition are combined to construct an integrated data set to avoid false alarms when there are no air discharge decomposition products, so that the fault detection is more accurate. When training the model, the present invention considers the influence of temperature and humidity on the detection of air chamber gas, so that the detection of air discharge decomposition products is more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, 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 invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0024] Figure 1 is a flow chart of the implementation of the air discharge decomposition product detection model training method provided by an embodiment of the present invention;

[0025] Figure 2 is a flow chart of an implementation of a method for detecting air discharge decomposition products provided by an embodiment of the present invention;

[0026] Figure 3a is a schematic diagram of temperature change provided by an embodiment of the present invention;

[0027] Figure 3b is a schematic diagram of humidity change provided by an embodiment of the present invention;

[0028] Figure 3c is a schematic diagram of voltage data changes of the sensor 1 provided in an embodiment of the present invention;

[0029] Figure 3d is a schematic diagram of voltage data changes of the sensor 2 provided in an embodiment of the present invention;

[0030] Figure 3e is a schematic diagram of voltage data changes of the sensor 3 provided in an embodiment of the present invention;

[0031] Figure 3f is a schematic diagram of voltage data changes of the sensor 4 provided in an embodiment of the present invention;

[0032] Figure 4 It is a schematic diagram of voltage data changes of a gas sensor array when air and NO2 of different concentrations are alternately introduced into a gas chamber provided by an embodiment of the present invention;

[0033] Figure 5 It is a structural schematic diagram of an air discharge decomposition product detection model training device provided by an embodiment of the present invention;

[0034] Figure 6 It is a schematic diagram of the structure of an air discharge decomposition product detection device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0035] 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 invention. However, it should be clear to those skilled in the art that the present invention may 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 invention.

[0036] In order to make the purpose, technical solutions and advantages of the present invention more clear, specific embodiments will be described below in conjunction with the accompanying drawings.

[0037] Figure 1 The implementation flow chart of the air discharge decomposition product detection model training method provided by the embodiment of the present invention is described in detail as follows:

[0038] Step 101, obtain normal operating condition parameters, construct a normal data set based on the normal operating condition parameters, and obtain fault operating condition parameters, and construct a fault data set based on the fault operating condition parameters; wherein, under normal operating conditions, the air chamber contains air, and the normal operating condition parameters include temperature data, humidity data, and voltage data of the gas sensor array; under fault conditions, the air chamber contains target air discharge decomposition products, and the fault operating condition parameters include temperature data, humidity data, voltage data of the gas sensor array, and concentration of the target air discharge decomposition products.

[0039] The air chamber used in the model training process of the present invention is a closed air chamber, and the temperature data, humidity data and voltage data of the gas sensor array in the air chamber are collected. The model is trained based on the temperature data, humidity data and voltage data of the gas sensor array. Since these data are less affected by other factors in the closed air chamber, the model trained using these data can detect the target air discharge decomposition products more accurately.

[0040] When collecting data, under normal working conditions, the air chamber is filled with air. The temperature and humidity of the air chamber are changed to collect the voltage data of the gas sensor array. The temperature, humidity, voltage data, and the presence and concentration data of the target air decomposition products are collected in the following format:

[0041]

[0042] Among them, D k is the kth data sequence. The same data set in the sequence has the same subscript i, which represents the order of the data set in the sequence; r i is the gas sensor array response (vector), t i is the temperature, h i is the humidity, e i It is a Boolean value, representing whether the target air discharge decomposition product is introduced at the current moment. If there is a target air discharge decomposition product, it is 1, otherwise it is 0. In this step, it is a normal working condition, and the air chamber is air, so e i Always 0, indicating no target air discharge decomposition product, c i is the target air discharge decomposition product concentration introduced. In this step, it is a normal working condition, so c i Always 0.

[0043] Under fault conditions, the gas chamber contains the target air discharge decomposition product, which changes the temperature and humidity of the gas chamber and collects voltage data from the gas sensor array in the same form as in normal conditions.

[0044] Among them, in this step, it is a fault condition, and the gas chamber contains the target air discharge decomposition product, so e i Always 1, indicating that there is a target air discharge decomposition product, c i is the target air discharge decomposition product concentration introduced.

[0045] Different temperatures such as 0℃, 10℃, 20℃, 30℃, etc., and different humidities such as the ratio of dry and wet air flow rates in the air chamber is 0%, 10%, 20%, 30%, 40%, 50%, etc.

[0046] The data for training the model includes the responses of gas sensors under different temperatures and humidities, which greatly reduces the impact of temperature and humidity changes on the accuracy of prediction results and avoids false alarms.

[0047] Under normal and fault conditions, specific parameters are collected to build corresponding data sets for further analysis and diagnosis. These data sets can help identify and distinguish the different performances of the air chamber under normal and fault conditions. The integrated data set is constructed by combining the data of normal conditions and the data of fault conditions to avoid false alarms when there are no air discharge decomposition products, making fault detection more accurate.

[0048] Step 102, constructing an integrated data set based on the normal data set and the fault data set; training an air discharge decomposition product detection model based on the integrated data set; wherein the air discharge decomposition product detection model takes the voltage data, temperature data and humidity data of the gas sensor array as input, and takes the probability and concentration of the target air discharge decomposition product as output.

[0049] The trained air discharge decomposition product detection model can be applied to the detection of air discharge decomposition products in switch cabinets.

[0050] The input of the air discharge decomposition product detection model is the voltage data, temperature data and humidity data of the gas sensor array, which are parameters obtained from the gas chamber in real time.

[0051] The output of the air discharge decomposition product detection model is the probability and concentration of the target air discharge decomposition product. The air discharge decomposition product detection model can predict the possibility of the existence of air discharge decomposition products and their concentration levels based on the input temperature data, humidity data and voltage data of the gas sensor.

[0052] The integrated dataset contains all relevant parameters under normal and fault conditions, providing comprehensive samples for training models.

[0053] A model that can predict air discharge decomposition products is trained using normal and fault data sets. This model can be used to monitor the status of the gas chamber in real time and detect potential faults in a timely manner, so as to achieve real-time monitoring of the gas chamber status and fault warning. By predicting the concentration and probability of air discharge decomposition products, preventive measures can be taken in advance to reduce the occurrence of faults and improve the stability and safety of the power system.

[0054] The window function is used to obtain data at n moments, and the response sequence, temperature and humidity of the gas sensor array are used as input data X, and whether the target air discharge decomposition product is introduced and the concentration of the target air discharge decomposition product are used as output data Y, as follows:

[0055]

[0056] Among them, the input data X and the output data are set correspondingly.

[0057] In a possible embodiment, after the input data X and the output data Y are normalized, 10% are randomly selected as a test set, and the remaining 90% are used as a training set.

[0058] In a possible implementation, the target air discharge decomposition products include NO2, O3 and CO. The target air discharge decomposition products are gases generated by air discharge decomposition.

[0059] In other possible implementations, the target air discharge decomposition products also include other gases. In the present invention, NO2, O3 and CO are used as examples for illustration.

[0060] In this embodiment, the gas sensor array is divided into multiple groups of gas sensors, each group of gas sensors is responsible for detecting the corresponding gas, such as the first group of gas sensors only has an obvious response to NO2, is responsible for detecting NO2, and outputs the response value of NO2, the second group of gas sensors only has an obvious response to O3, is responsible for detecting O3, and outputs the response value of O3, the third group of gas sensors only has an obvious response to CO, is responsible for detecting CO, and outputs the response value of CO. In this way, during operation, it can be ensured that the detection of different gases will not affect each other.

[0061] Model training for each gas is performed according to the response values, temperature data and humidity data of each group of gas sensors in the gas sensor array, and the trained model combination of each gas constitutes an air discharge decomposition product detection model.

[0062] During detection, the temperature data, humidity data and the response values ​​of each group of gas sensors in the gas sensor array are respectively input into the air discharge decomposition product detection model, and the probability of each gas and its corresponding concentration are output respectively. Each gas has a corresponding gas threshold. The output order of each gas has been set in advance, and the gas name corresponding to each group of data can be determined according to the output sorting.

[0063] If the probability value of a certain gas is lower than the gas threshold of the gas, it means that the credibility of the existence of the gas is low, that is, the possibility of the existence of the gas is low, and the concentration of the gas will not be considered at this time. If the probability of a certain gas in the output result of the air discharge decomposition detection model is 0, it is considered that the gas does not exist, and the corresponding concentration value will not be considered. Correspondingly, if the probability value of a certain gas is higher than the gas threshold of the gas, it means that the possibility of the existence of the gas is high. At this time, the concentration of the gas needs to be considered, and subsequent fault judgment and other operations are performed based on the concentration of the gas in combination with other data.

[0064] If the probability of each gas in the model output results is lower than its corresponding gas threshold, and the probability of air is higher than its corresponding threshold at this time, it means that the air chamber is in normal operating conditions at this time.

[0065] The embodiment of the present invention obtains normal operating condition parameters, constructs a normal data set according to the normal operating condition parameters, obtains fault operating condition parameters, constructs a fault data set according to the fault operating condition parameters, constructs an integrated data set according to the normal data set and the fault data set, and trains the air discharge decomposition product detection model according to the integrated data set. The present invention considers the normal operating condition and the fault operating condition, that is, when the power equipment is operating normally, there is only air in the air chamber and no air discharge decomposition products are generated, and when the equipment fails, air decomposition products are generated in the air chamber. The data of the normal operating condition and the data of the fault operating condition are combined to construct an integrated data set to avoid false alarms when there are no air discharge decomposition products, so that the fault detection is more accurate. When training the model, the present invention considers the influence of temperature and humidity on the detection of air chamber gas, so that the detection of air discharge decomposition products is more accurate.

[0066] In a possible implementation, an integrated data set is constructed based on the normal data set and the fault data set, including:

[0067] For normal working conditions and fault working conditions, the data in the normal data set and the fault data set are classified according to different temperatures, different humidities or different voltage data, and sub-data sets corresponding to different temperature, different humidity or different voltage data are constructed.

[0068] In this embodiment, the purpose of constructing the sub-datasets is to more carefully analyze and understand the behavior and performance of the equipment under normal working conditions and fault working conditions under different environmental conditions.

[0069] These sub-datasets can be used to train more accurate machine learning models to predict the behavior of equipment under specific environmental conditions and improve the accuracy of fault detection. They can also study the performance of equipment under different conditions in more detail, thus providing more specific data support for equipment maintenance, fault prediction, and system optimization.

[0070] In a possible implementation, under normal operating conditions, the probability and concentration of the target air discharge decomposition products are both zero.

[0071] In a possible implementation, the gas sensor array includes multiple groups of gas sensors, each group of gas sensors detects different gases, and each gas sensor in each group has different sensitivity to the same gas.

[0072] In this embodiment, the gas sensors can respond to the target air discharge decomposition products, and the response of each gas sensor to the target air discharge decomposition product should differ in indicators such as sensitivity, response rate, and response time, so that the gas sensor array can respond to the target air discharge decomposition product.

[0073] In a possible implementation manner, the correlation between the voltage data of different gas sensors in each group is lower than a correlation threshold.

[0074] In this embodiment, the correlation threshold is 90%, and the performance of each gas sensor is different, so that the gas sensor array can respond to the gas well.

[0075] In a possible implementation, the calculation formula of the correlation is as follows:

[0076]

[0077] Where R1 and R2 are the voltage data of two gas sensors measuring the same gas with the same concentration, D is the variance, and Cov is the covariance.

[0078] In this embodiment, at least one gas sensor should have a good response time, because the response time of the LSTM model will depend on the gas sensor with the shortest response time; at least one gas sensor is not saturated at the maximum concentration of the target air discharge decomposition product; at least one gas sensor can produce a response at the minimum concentration of the target air discharge decomposition product; all gas sensors should have good repeatability, so that the gas sensor array can respond to the target air discharge decomposition product.

[0079] In a possible implementation, the air discharge decomposition product detection model is an LSTM model.

[0080] In this embodiment, the LSTM model is a recurrent neural network that is good at processing time series. The voltage data, temperature and humidity data of the gas sensor array are input into the trained LSTM model to detect whether the target air discharge decomposition product currently exists and estimate its concentration.

[0081] The LSTM model includes two LSTM hidden layers and ReLU activation function, and the output layer is a fully connected layer and a Sigmoid activation function. During the detection, the voltage data, temperature and humidity data of the gas sensor array are input into the model, and the probability of the target air discharge decomposition product and its corresponding concentration are output.

[0082] The width of the two hidden layers is 32. The reason for selecting the activation function as ReLU is that the accuracy of identifying the target air discharge decomposition products is extremely low when using other activation functions, such as tanh.

[0083] The ReLU activation function and the Sigmoid activation function are used to ensure that the capacity of the model is just suitable for the prediction of one type of air discharge decomposition, without obvious underfitting or overfitting.

[0084] In order to reduce overfitting, the Dropout strategy is used, and the forward propagation during training randomly disables 50% of the nodes in the two-layer LSTM hidden layer. The optimizer uses adam and the loss function uses mean square error. The amount of data and the number of iterations for each iteration need to be determined by comparing the prediction accuracy of the model on the test set.

[0085] Figure 2 The implementation flow chart of the air discharge decomposition product detection method provided by the embodiment of the present invention is described in detail as follows:

[0086] Step 201, acquiring voltage data, temperature data and humidity data of the gas sensor array.

[0087] Temperature data and humidity data can be collected according to their respective sensors. The voltage data, temperature data and humidity data of the gas sensor array are all data in the gas chamber.

[0088] Voltage data is a key indicator of gas sensor response and can be acquired through the circuit connected to the gas sensor.

[0089] The voltage data, temperature data and humidity data of the gas sensor array provide a data basis for subsequent gas detection.

[0090] Step 202 : input the voltage data, temperature data and humidity data of the gas sensor array into the air discharge decomposition product detection model, and output the probability and concentration of the target air discharge decomposition product.

[0091] The probability represents the model's prediction of the likelihood of a particular gas being present, and is a value between 0 and 1, with each gas having its own gas threshold.

[0092] In a possible implementation, when the probability of a target air discharge decomposition product is greater than a target air discharge decomposition product threshold, it is determined that the target air discharge decomposition product is detected.

[0093] If the probability value of a certain gas output by the model is greater than or equal to the gas threshold of the gas, it indicates that the possibility of the existence of the gas is relatively high, and it is considered that the gas exists in the air chamber at this time. If the probability value of a certain gas output by the model is less than the gas threshold of the gas, it indicates that the possibility of the existence of the gas is relatively low, and it is considered that the gas does not exist in the air chamber at this time. If the results of each gas are less than their corresponding gas thresholds, and the probability value of air is higher than its corresponding gas threshold, it is considered that this is a normal operating condition.

[0094] The present invention takes into account that in actual situations, temperature and humidity will change with time and produce regular changes with a daily cycle, thereby causing the gas sensor response to change accordingly. Therefore, a temperature sensor and a humidity sensor are added to collect temperature data and humidity data using the temperature sensor and the humidity sensor. The influence of temperature and humidity on the target air discharge decomposition product is comprehensively considered, so that the detected target air discharge decomposition product is more accurate.

[0095] For example, there are four gas sensors in the gas sensor array for measuring NO2. Under the condition of 0% wet air flow rate, the temperature is increased from 0℃ to 60℃ and then cooled to 0℃, and the response of the gas sensors to the air is as follows: Figure 3a-3f It can be seen that changes in temperature and humidity will affect the response of the gas sensor.

[0096] At 0℃ and 0% humid air flow rate, the gas sensor responds to different concentrations of NO2 as shown below: Figure 4 . Air and a certain concentration of NO2 were alternately introduced into the gas chamber for 30 minutes and 30 minutes, and the concentration increased from 5ppm to 50ppm according to the increase rate of 5ppm in each group of experiments. Gas sensor 1 and gas sensor 2 have shorter response time and recovery time than gas sensor 3 and gas sensor 4; gas sensor 1 is more sensitive to lower concentrations of NO2 than gas sensor 2 and is easily saturated with higher concentrations of NO2. The same is true for gas sensor 4 compared to gas sensor 3.

[0097] It should be understood that the order of execution of the steps in the above embodiment does not necessarily 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 embodiment of the present invention.

[0098] The following is an embodiment of the device of the present invention. For details not described in detail therein, reference may be made to the corresponding method embodiment described above.

[0099] Figure 5 The structure diagram of the air discharge decomposition product detection model training device provided by the embodiment of the present invention is shown. For the convenience of explanation, only the part related to the embodiment of the present invention is shown, which is described in detail as follows:

[0100] like Figure 5 As shown, the air discharge decomposition product detection model training device 5 includes:

[0101] The parameter acquisition module 51 is used to acquire normal operating condition parameters, construct a normal data set according to the normal operating condition parameters, and acquire fault operating condition parameters, and construct a fault data set according to the fault operating condition parameters; wherein, under normal operating conditions, the air chamber contains air, and the normal operating condition parameters include temperature data, humidity data, and voltage data of the gas sensor array; under fault operating conditions, the air chamber contains target air discharge decomposition products, and the fault operating condition parameters include temperature data, humidity data, voltage data of the gas sensor array, and concentration of the target air discharge decomposition products;

[0102] The air discharge decomposition product detection model training module 52 is used to construct an integrated data set based on the normal data set and the fault data set; and to obtain the air discharge decomposition product detection model based on the integrated data set training; wherein the air discharge decomposition product detection model takes the voltage data, temperature data and humidity data of the gas sensor array as input, and takes the probability and concentration of the target air discharge decomposition product as output.

[0103] In a possible implementation, the parameter acquisition module 51 is used to:

[0104] Based on the normal data set and the fault data set, an integrated data set is constructed, including:

[0105] For normal working conditions and fault working conditions, the data in the normal data set and the fault data set are classified according to different temperatures, different humidities or different voltage data, and sub-data sets corresponding to different temperature, different humidity or different voltage data are constructed.

[0106] Figure 6 As shown, the air discharge decomposition product detection device 6 includes:

[0107] A data acquisition module 61 is used to acquire voltage data, temperature data and humidity data of the gas sensor array;

[0108] The air discharge decomposition product detection module 62 is used to input the voltage data, temperature data and humidity data of the gas sensor array into the air discharge decomposition product detection model, and output the probability and concentration of the target air discharge decomposition product.

[0109] In a possible implementation, the air discharge decomposition product detection module 62 is used to:

[0110] When the probability of the target air discharge decomposition product is greater than the target air discharge decomposition product threshold, it is determined that the target air discharge decomposition product is detected.

[0111] The embodiment of the present invention obtains normal operating condition parameters, constructs a normal data set according to the normal operating condition parameters, obtains fault operating condition parameters, constructs a fault data set according to the fault operating condition parameters, constructs an integrated data set according to the normal data set and the fault data set, and trains the air discharge decomposition product detection model according to the integrated data set. The present invention considers the normal operating condition and the fault operating condition, that is, when the power equipment is operating normally, there is only air in the air chamber and no air discharge decomposition products are generated, and when the equipment fails, air decomposition products are generated in the air chamber. The data of the normal operating condition and the data of the fault operating condition are combined to construct an integrated data set to avoid false alarms when there are no air discharge decomposition products, so that the fault detection is more accurate. When training the model, the present invention considers the influence of temperature and humidity on the detection of air chamber gas, so that the detection of air discharge decomposition products is more accurate.

[0112] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0113] Those of ordinary skill in the art will appreciate that the templates, units, and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0114] If the module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, the steps of the above-mentioned air discharge decomposition product detection model training method and air discharge decomposition product detection method embodiment can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form, etc. The computer-readable medium may include: any entity or device that can carry the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory, random access memory, electric carrier signal, telecommunication signal and software distribution medium, etc.

[0115] 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 method for training an air discharge decomposition product detection model, characterized in that: include: Normal operating condition parameters are obtained, a normal data set is constructed according to the normal operating condition parameters, and fault operating condition parameters are obtained, and a fault data set is constructed according to the fault operating condition parameters; wherein, under the normal operating condition, the air chamber contains air, and the normal operating condition parameters include temperature data, humidity data, and voltage data of a gas sensor array; under the fault operating condition, the air chamber contains target air discharge decomposition products, and the fault operating condition parameters include temperature data, humidity data, voltage data of a gas sensor array, and concentration of target air discharge decomposition products; an integrated data set is constructed according to the normal data set and the fault data set; an air discharge decomposition product detection model is obtained by training according to the integrated data set; wherein, the air discharge decomposition product detection model uses the voltage data, temperature data, and humidity data of the gas sensor array as input, and uses the probability and concentration of the target air discharge decomposition products as output.

2. The air discharge decomposition product detection model training method according to claim 1, characterized in that: The step of constructing an integrated data set according to the normal data set and the fault data set includes: For normal working conditions and fault working conditions respectively, the data in the normal data set and the fault data set are classified according to different temperatures, different humidities or different voltage data, and sub-data sets corresponding to different temperature, different humidity or different voltage data are constructed.

3. The air discharge decomposition product detection model training method according to claim 1, characterized in that: Under normal operating conditions, the probability and concentration of the target air discharge decomposition products are both zero.

4. The air discharge decomposition product detection model training method according to claim 1, characterized in that: The gas sensor array includes multiple groups of gas sensors, each group of gas sensors detects different gases, and each gas sensor in each group has different sensitivity to the same gas.

5. The air discharge decomposition product detection model training method according to claim 4, characterized in that: The correlation between the voltage data of different gas sensors in each group is lower than a correlation threshold.

6. The air discharge decomposition product detection model training method according to claim 5, characterized in that: The calculation formula of the correlation is as follows: Where R1 and R2 are the voltage data of two gas sensors measuring the same gas with the same concentration, D is the variance, and Cov is the covariance.

7. The air discharge decomposition product detection model training method according to claim 1, characterized in that: The target air discharge decomposition products include NO2, O3 and CO.

8. The air discharge decomposition product detection model training method according to claim 1, characterized in that: The air discharge decomposition product detection model is an LSTM model.

9. A method for detecting air discharge decomposition products, the method performing detection based on an air discharge decomposition product detection model obtained by the air discharge decomposition product detection model training method according to any one of claims 1 to 8, characterized in that: include: Acquire voltage data, temperature data and humidity data of the gas sensor array; The voltage data, the temperature data and the humidity data of the gas sensor array are input into the air discharge decomposition product detection model, and the probability and concentration of the target air discharge decomposition product are output.

10. The method for detecting air discharge decomposition products according to claim 9, characterized in that: When the probability of the target air discharge decomposition product is greater than the target air discharge decomposition product threshold, it is determined that the target air discharge decomposition product is detected.

Images