Electrical equipment thermal defect severity identification method and system

Abstract

The invention discloses an electrical equipment thermal defect severity identification method and system, and the method comprises the steps: collecting internal sample temperature data, housing sample temperature data and environment temperature sample data of electrical equipment corresponding to the known thermal defect severity level of the electrical equipment; obtaining steady-state sample data and transient-state sample data; constructing a steady-state sample data set and a transient-state sample data set corresponding to the known severity level of the thermal defect of the power equipment; training the first neural network by adopting the steady-state sample data set; training a second neural network by using the transient sample data set; during actual detection, the power equipment shell measured temperature and the environment measured temperature are input into the trained first neural network, the trained first neural network outputs power equipment internal temperature difference prediction data, and the power equipment shell temperature rise measured data and the environment temperature rise measured data are input into the trained second neural network. And outputting internal temperature rise prediction data of the power equipment.

Description

Technical Field

[0001] The present invention relates to a detection method and system, and in particular to a power equipment state detection method and system. Background Art

[0002] GIS equipment has been widely used in power systems due to its reliability, ease of installation, and small footprint.

[0003] At present, there is an online monitoring method for thermal defects of GIS equipment, which trains artificial neural networks based on data from simulation models, with the input node being the external measured temperature and the output node being the maximum temperature inside the GIS busbar. During the detection process, sensors are installed at the highest and lowest points on the shell surface where thermal defects are prone to occur, while monitoring the ambient temperature and inputting it into the trained neural network to achieve the purpose of monitoring internal thermal defects.

[0004] However, the input parameters of this monitoring method are in a steady state, which is relatively simple, low in accuracy, and low in applicability and versatility. Summary of the invention

[0005] One of the purposes of the present invention is to provide a method for identifying the severity of thermal defects in power equipment, which trains a dual neural network based on steady-state temperature data and transient temperature data to extract the characteristics of the severity of thermal defects in power equipment, thereby improving the detection accuracy of the severity of thermal defects in power equipment.

[0006] According to the above invention purpose, the present invention proposes a method for identifying the severity of thermal defects of power equipment, which comprises the steps of:

[0007] Corresponding to the known severity level of thermal defects of power equipment, collecting sample temperature data inside the power equipment, sample temperature data of the power equipment shell, and sample temperature data of the environment;

[0008] Based on the internal sample temperature data of the power equipment, the shell sample temperature data of the power equipment and the ambient temperature sample data, steady-state sample data and transient sample data are obtained;

[0009] Based on the acquired steady-state sample data and transient sample data, a steady-state sample data group and a transient sample data group corresponding to the known severity level of thermal defects of the power equipment are constructed, wherein the steady-state sample data group includes the sample temperature data of the power equipment shell, the ambient temperature sample data, and the internal temperature difference sample data of the power equipment corresponding to the two, and the transient sample data group includes the sample temperature rise data of the power equipment shell, the ambient temperature rise sample data, and the internal temperature rise sample data of the power equipment corresponding to the two;

[0010] The first neural network is trained using a steady-state sample data group to establish a mapping relationship between the sample temperature data of the power equipment housing, the sample data of the ambient temperature, and the sample data of the temperature difference inside the power equipment;

[0011] The second neural network is trained using the transient sample data group to establish a mapping relationship between the temperature rise sample data of the power equipment housing, the ambient temperature rise sample data and the temperature rise sample data inside the power equipment;

[0012] The actual temperature of the power equipment casing and the actual temperature of the environment are collected, and the actual temperature rise data of the power equipment casing and the actual temperature rise data of the environment are obtained. The actual temperature rise data of the power equipment casing and the actual temperature rise data of the environment are input into a trained first neural network, which outputs the prediction data of the temperature difference inside the power equipment. The actual temperature rise data of the power equipment casing and the actual temperature rise data of the environment are input into a trained second neural network, which outputs the prediction data of the temperature rise inside the power equipment. The prediction data of the temperature difference inside the power equipment and the prediction data of the temperature rise inside the power equipment are used to characterize the severity level of thermal defects of the power equipment.

[0013] Furthermore, in the method described in the present invention, the first neural network and / or the second neural network is a back propagation neural network.

[0014] Furthermore, in the method described in the present invention, the back propagation neural network includes an input layer, a hidden layer and an output layer, wherein the hidden layer uses RELU as an activation function.

[0015] Furthermore, in the method described in the present invention, when training the back-propagation neural network, the Adam algorithm and the cross entropy loss function are used to calculate and update the network weights of the back-propagation neural network.

[0016] Furthermore, in the method described in the present invention, the internal sample temperature data of the power equipment includes contact temperature sample data of the GIS equipment and conductor temperature sample data of the GIS equipment, and the internal temperature difference sample data of the power equipment includes contact and conductor temperature difference sample data.

[0017] Furthermore, in the method described in the present invention, the internal temperature rise sample data of the power equipment includes the contact temperature rise rate and / or the temperature rise rate difference between the contact and the conductor.

[0018] Another object of the present invention is to provide a system for identifying the severity of thermal defects in power equipment, which trains a dual neural network based on steady-state temperature data and transient temperature data to extract characteristics of the severity of thermal defects in power equipment, thereby improving the detection accuracy of the severity of thermal defects in power equipment.

[0019] Based on the above purpose, the present invention also provides a system for identifying the severity of thermal defects of power equipment, which includes:

[0020] A data collection module, which collects sample temperature data inside the power equipment, sample temperature data of the power equipment shell, and sample temperature data of the environment, corresponding to the known severity level of thermal defects of the power equipment;

[0021] A data processing module, which acquires steady-state sample data and transient sample data based on the internal sample temperature data of the power equipment, the sample temperature data of the power equipment shell and the ambient temperature sample data; and constructs a steady-state sample data group and a transient sample data group corresponding to the known severity level of thermal defects of the power equipment based on the acquired steady-state sample data and transient sample data, wherein the steady-state sample data group includes the sample temperature data of the power equipment shell, the ambient temperature sample data and the internal temperature difference sample data of the power equipment corresponding to the two, and the transient sample data group includes the temperature rise sample data of the power equipment shell, the ambient temperature rise sample data and the internal temperature rise sample data of the power equipment corresponding to the two;

[0022] A first feature extraction module uses a steady-state sample data group to train a first neural network to establish a mapping relationship between the power equipment housing sample temperature data, the ambient temperature sample data, and the power equipment internal temperature difference sample data; and outputs the power equipment internal temperature difference prediction data based on the collected power equipment housing actual temperature and the ambient actual temperature;

[0023] The second feature extraction module uses a transient sample data group to train the second neural network to establish a mapping relationship between the power equipment shell temperature rise sample data, the ambient temperature rise sample data and the power equipment internal temperature rise sample data; and outputs the power equipment internal temperature rise prediction data based on the power equipment shell temperature rise actual measured data and the ambient temperature rise actual measured data.

[0024] Furthermore, in the system described in the present invention, the first neural network and / or the second neural network is a back propagation neural network.

[0025] Furthermore, in the system described in the present invention, the back propagation neural network includes an input layer, a hidden layer and an output layer, wherein the hidden layer adopts RELU as an activation function.

[0026] Furthermore, in the system described in the present invention, when the back-propagation neural network is trained, the Adam algorithm and the cross entropy loss function are used to calculate and update the network weights of the back-propagation neural network.

[0027] The method and system for identifying the severity of thermal defects in power equipment according to the present invention have the following advantages and beneficial effects:

[0028] The method and system for identifying the severity of thermal defects of power equipment described in the present invention trains a dual neural network based on steady-state temperature data and transient temperature data to extract the characteristics of the severity of thermal defects of power equipment, and identifies the severity of thermal defects of power equipment through the dual output feature data of the two neural networks, thereby improving the detection accuracy of the severity of thermal defects of power equipment.

[0029] The method and system described in the present invention are convenient for detection, widely applicable, low-cost and highly efficient. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 A flowchart of the steps of a method for identifying the severity of thermal defects in power equipment according to the present invention in one embodiment is shown.

[0031] Figure 2 From an axial perspective, a diagram showing an arrangement of sensors for detecting temperature data of electric power equipment according to an embodiment of the method for identifying severity of thermal defects of electric power equipment described in the present invention is shown.

[0032] Figure 3 From an axial perspective, a diagram showing an arrangement of sensors for detecting ambient temperature data according to an implementation of the method for identifying severity of thermal defects in electric power equipment described in the present invention is shown.

[0033] Figure 4 The structure diagram of the BPNN used in the method for identifying the severity of thermal defects of power equipment described in the present invention in one embodiment is shown.

[0034] Figure 5 The diagram shows the architecture of a system for identifying the severity of thermal defects in power equipment according to the present invention in one embodiment. DETAILED DESCRIPTION

[0035] The following will further explain and illustrate the method and system for identifying the severity of thermal defects in power equipment described in the present invention in conjunction with the drawings and specific embodiments of the specification. However, such explanation and illustration do not constitute an undue limitation on the technical solution of the present invention.

[0036] Figure 1 A flowchart of the steps of a method for identifying the severity of thermal defects in power equipment according to the present invention in one embodiment is shown.

[0037] like Figure 1 As shown, in one embodiment, the method for identifying the severity of thermal defects of power equipment may include the following steps:

[0038] S100: Corresponding to the known severity level of thermal defects of the power equipment, collecting sample temperature data inside the power equipment, sample temperature data of the power equipment shell, and sample temperature data of the environment.

[0039] In some more specific implementations, the severity levels of thermal defects of the power equipment may include four levels: normal, mild, moderate, and severe.

[0040] In a more specific embodiment, the power device may be a GIS device.

[0041] Based on this, the internal temperature and shell temperature of the GIS equipment are collected in real time through the temperature sensors arranged on the GIS equipment. Figure 2 As shown, in a specific example, the sensor for collecting the internal temperature data of the GIS device includes a first temperature sensor 12 arranged on the surface of the contact seat 2, a second temperature sensor 13 arranged on the surface of the conductor 3, and the sensor for collecting the outer shell temperature data of the GIS device includes a third temperature sensor 11 arranged on the surface of the outer shell 1. Among them, the first temperature sensor 12 on the surface of the contact seat can be set at the highest point of the contact, and the second temperature sensor 13 can be separated from the first temperature sensor 12 by a certain axial distance, for example, 0.5m. Due to gravity, the temperature at the highest point of the contact seat surface is the highest, and the temperature change is most obvious. The third temperature sensor 11 on the surface of the outer shell 1 can be set at the lowest position of the outer shell surface, because it will not be disturbed by solar radiation at the bottom, and the influence of wind speed at the bottom is also greatly reduced.

[0042] In addition, if Figure 3 As shown, in the surrounding environment E of the GIS equipment, for example, within a range greater than 1m and less than 2m from the surface of the casing 1, a fourth temperature sensor 14 is installed, whose axial position corresponds to the third temperature sensor on the casing, and the position is preferably away from solar radiation and wind, for monitoring ambient temperature data.

[0043] S200: Acquire steady-state sample data and transient sample data based on the internal sample temperature data of the power equipment, the shell sample temperature data of the power equipment, and the ambient temperature sample data.

[0044] S300: Based on the acquired steady-state sample data and transient sample data, a steady-state sample data group and a transient sample data group corresponding to the known severity levels of thermal defects of the power equipment are constructed, wherein the steady-state sample data group includes the power equipment casing sample temperature data, the ambient temperature sample data, and the power equipment internal temperature difference sample data corresponding to the two, and the transient sample data group includes the power equipment casing temperature rise sample data, the ambient temperature rise sample data, and the power equipment internal temperature rise sample data corresponding to the two.

[0045] In some more specific embodiments, the internal sample temperature data of the power equipment includes the contact temperature sample data of the GIS equipment and the conductor temperature sample data of the GIS equipment, the internal temperature difference sample data of the power equipment includes the contact and conductor temperature difference sample data, and the internal temperature rise sample data of the power equipment includes the contact temperature rise rate and / or the temperature rise rate difference between the contact and the conductor.

[0046] Since the poor contact thermal defect is a transient process, the contact resistance may be very small at the beginning, but under long-term load current operation, the contact resistance gradually increases and generates heat, eventually leading to equipment failure. Based on this, the present invention divides the data into two categories: one is steady-state data, including shell temperature, internal temperature and ambient temperature and temperature difference; the other is transient data, including temperature rise rate and / or temperature rise rate difference.

[0047] S400: Using the steady-state sample data group to train the first neural network to establish a mapping relationship between the power equipment housing sample temperature data, the ambient temperature sample data and the power equipment internal temperature difference sample data.

[0048] In some more specific embodiments, the first neural network may be a back propagation neural network (BNPP).

[0049] like Figure 4 As shown, the back propagation neural network can include an input layer, a hidden layer (the number of layers is greater than or equal to 1), and an output layer.

[0050] The first neural network is a steady-state neural network, and the data of the input layer are the sample temperature data of the power equipment shell and the sample data of the ambient temperature. The hidden layer uses RELU as the activation function. The output layer outputs the temperature difference inside the power equipment, and in a more specific embodiment, outputs the temperature difference between the contact and the conductor.

[0051] During the training process, the Adam algorithm and cross entropy loss function can be used to calculate and update the network weights of BPNN.

[0052] S500: Using the transient sample data group to train the second neural network to establish a mapping relationship between the power equipment housing temperature rise sample data, the ambient temperature rise sample data and the power equipment internal temperature rise sample data.

[0053] In some more specific embodiments, the second neural network may also be a back propagation neural network (BNPP).

[0054] The second neural network is a transient neural network. The data input to the input layer are the temperature rise sample data of the power equipment shell and the ambient temperature rise sample data. The hidden layer uses RELU as the activation function. The output layer outputs the internal temperature rise data of the power equipment. In a more specific embodiment, the temperature rise rate of the contact and the difference between the temperature rise rate of the contact and the conductor are output.

[0055] S600: Collect the actual temperature of the power equipment shell and the actual temperature of the environment, and obtain the actual temperature rise data of the power equipment shell and the actual temperature rise data of the environment, input the actual temperature rise data of the power equipment shell and the actual temperature rise data of the environment into a trained first neural network, which outputs the prediction data of the temperature difference inside the power equipment, input the actual temperature rise data of the power equipment shell and the actual temperature rise data of the environment into a trained second neural network, which outputs the prediction data of the temperature rise inside the power equipment.

[0056] Then, the severity level of thermal defects of the power equipment is determined based on the outputted prediction data of internal temperature difference and internal temperature rise of the power equipment.

[0057] Table 1 lists the correspondence between the temperature difference between the internal contacts and conductors of the GIS equipment, which is steady-state data, and the contact temperature rise rate and the difference in the temperature rise rate between the contacts and the conductor, which are transient data, and the severity level of thermal defects of the power equipment in some more specific implementations.

[0058] Table 1

[0059]

[0060] In another embodiment, the present invention further provides a system for identifying the severity of thermal defects in electric power equipment.

[0061] like Figure 5 As shown, the system may include:

[0062] A data collection module 502 collects sample temperature data inside the power equipment, sample temperature data of the power equipment housing, and sample ambient temperature data corresponding to known severity levels of thermal defects of the power equipment;

[0063] The data processing module 504 acquires steady-state sample data and transient sample data based on the sample temperature data inside the power equipment, the sample temperature data of the power equipment shell, and the ambient temperature sample data; and constructs a steady-state sample data group and a transient sample data group corresponding to the known severity level of thermal defects of the power equipment based on the acquired steady-state sample data and transient sample data, wherein the steady-state sample data group includes the sample temperature data of the power equipment shell, the ambient temperature sample data, and the temperature difference sample data inside the power equipment corresponding to the two, and the transient sample data group includes the temperature rise sample data of the power equipment shell, the ambient temperature rise sample data, and the temperature rise sample data inside the power equipment corresponding to the two;

[0064] The first feature extraction module 506 uses the steady-state sample data group to train the first neural network to establish a mapping relationship between the power equipment shell sample temperature data, the ambient temperature sample data and the power equipment internal temperature difference sample data; and outputs the power equipment internal temperature difference prediction data based on the collected power equipment shell measured temperature and the ambient measured temperature;

[0065] The second feature extraction module 508 uses a transient sample data group to train the second neural network to establish a mapping relationship between the power equipment casing temperature rise sample data, the ambient temperature rise sample data and the power equipment internal temperature rise sample data; and outputs the power equipment internal temperature rise prediction data based on the power equipment casing temperature rise actual measured data and the ambient temperature rise actual measured data.

[0066] In some more specific embodiments, the first neural network and / or the second neural network is a back-propagation neural network.

[0067] In some more specific embodiments, the back propagation neural network includes an input layer, a hidden layer and an output layer, wherein the hidden layer uses RELU as an activation function.

[0068] In some more specific embodiments, when training the back-propagation neural network, the Adam algorithm and the cross entropy loss function are used to calculate and update the network weights of the back-propagation neural network.

[0069] In the system described in the present invention, the data acquisition module, the data processing module, the first feature extraction module and the second feature extraction module can be implemented in any appropriate manner, for example, they can take the form of a microprocessor or a processor and a computer-readable medium storing computer-readable program code (such as software or firmware) that can be executed by the (micro)processor, logic gates, switches, application specific integrated circuits (ASIC), programmable logic controllers and embedded microcontrollers. Examples of controllers include but are not limited to the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20 and Silicone Labs C8051F320.

[0070] In addition, those skilled in the art also know that, in addition to implementing the controller in a purely computer-readable program code, the control module can be made to implement the same function in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers by logically programming the method steps. Therefore, such a control module can be considered as a hardware component, and the devices for implementing various functions included therein can also be considered as structures within the hardware component. Or even, the devices for implementing various functions can be considered as both software modules for implementing the method and structures within the hardware component.

[0071] In order to verify the effectiveness of the method proposed in the present invention, an experiment was carried out on a 72.5KVGIS device. The steady-state and transient temperature data corresponding to its normal operation and the presence of different degrees of poor contact defects were collected on-site. 100 groups of temperature signals were collected under each thermal defect state in a clean indoor environment, of which 70 groups were used as training sets and the remaining 30 groups were used as test sets.

[0072] In order to verify the superiority of the method proposed in the present invention over the comparative method (GIS equipment contact temperature monitoring based on a single BP neural network) in discrimination performance and the influence of ambient temperature on the recognition results, the recognition results of the two methods were compared under indoor temperatures of 0°C, 20°C and 40°C, respectively. The recognition results under different thermal defect severity levels are shown in Table 2.

[0073] Table 2

[0074]

[0075] It can be seen from Table 2 that the present invention has better recognition accuracy. It can be seen that the method and system for identifying the severity of thermal defects of power equipment according to the present invention trains dual neural networks based on steady-state temperature data and transient temperature data to extract the characteristics of the severity of thermal defects of power equipment, and identifies the severity of thermal defects of power equipment through the dual output feature data of the two neural networks, thereby improving the detection accuracy of the severity of thermal defects of power equipment.

[0076] It should be noted that the prior art in the protection scope of the present invention is not limited to the embodiments given in the present application documents. All prior art that does not contradict the scheme of the present invention, including but not limited to prior patent documents, prior public publications, prior public uses, etc., can be included in the protection scope of the present invention.

[0077] In addition, the combination of the various technical features in this case is not limited to the combination described in the claims of this case or the combination described in the specific embodiments. All technical features recorded in this case can be freely combined or combined in any way unless there is a contradiction between them.

[0078] It should also be noted that the above-listed embodiments are only specific embodiments of the present invention. Obviously, the present invention is not limited to the above embodiments, and similar changes or modifications made therewith can be directly derived or easily associated with the contents disclosed by those skilled in the art from the present invention, and all should belong to the protection scope of the present invention.

Claims

1. A method for identifying the severity of thermal defects in power equipment, characterized in that: Includes steps: Corresponding to the known severity level of thermal defects of the power equipment, collecting sample temperature data inside the power equipment, sample temperature data of the power equipment shell, and sample temperature data of the environment; Based on the internal sample temperature data of the power equipment, the shell sample temperature data of the power equipment and the ambient temperature sample data, steady-state sample data and transient sample data are obtained; Based on the acquired steady-state sample data and transient sample data, a steady-state sample data group and a transient sample data group corresponding to the known severity level of thermal defects of the power equipment are constructed, wherein the steady-state sample data group includes the sample temperature data of the power equipment shell, the ambient temperature sample data, and the internal temperature difference sample data of the power equipment corresponding to the two, and the transient sample data group includes the sample temperature rise data of the power equipment shell, the ambient temperature rise sample data, and the internal temperature rise sample data of the power equipment corresponding to the two; The first neural network is trained using a steady-state sample data group to establish a mapping relationship between the sample temperature data of the power equipment housing, the sample data of the ambient temperature, and the sample data of the temperature difference inside the power equipment; The second neural network is trained using the transient sample data group to establish a mapping relationship between the temperature rise sample data of the power equipment housing, the ambient temperature rise sample data and the temperature rise sample data inside the power equipment; The actual temperature of the power equipment casing and the actual temperature of the environment are collected, and the actual temperature rise data of the power equipment casing and the actual temperature rise data of the environment are obtained. The actual temperature rise data of the power equipment casing and the actual temperature rise data of the environment are input into a trained first neural network, which outputs the prediction data of the temperature difference inside the power equipment. The actual temperature rise data of the power equipment casing and the actual temperature rise data of the environment are input into a trained second neural network, which outputs the prediction data of the temperature rise inside the power equipment. The prediction data of the temperature difference inside the power equipment and the prediction data of the temperature rise inside the power equipment are used to characterize the severity level of thermal defects of the power equipment.

2. The method according to claim 1, characterized in that The first neural network and / or the second neural network is a back propagation neural network.

3. The method according to claim 2, characterized in that The back propagation neural network includes an input layer, a hidden layer and an output layer, wherein the hidden layer adopts RELU as an activation function.

4. The method according to claim 2, characterized in that When the back propagation neural network is trained, the Adam algorithm and the cross entropy loss function are used to calculate and update the network weights of the back propagation neural network.

5. The method according to claim 1, characterized in that The internal sample temperature data of the electric power equipment includes contact temperature sample data of the GIS equipment and conductor temperature sample data of the GIS equipment, and the internal temperature difference sample data of the electric power equipment includes contact and conductor temperature difference sample data.

6. The method according to claim 5, characterized in that The internal temperature rise sample data of the electric power equipment includes a contact temperature rise rate and / or a temperature rise rate difference between a contact and a conductor.

7. A system for identifying the severity of thermal defects in power equipment, characterized in that: include: A data collection module, which collects sample temperature data inside the power equipment, sample temperature data of the power equipment shell, and sample temperature data of the environment, corresponding to the known severity level of thermal defects of the power equipment; A data processing module, which acquires steady-state sample data and transient sample data based on the internal sample temperature data of the power equipment, the sample temperature data of the power equipment shell and the ambient temperature sample data; and constructs a steady-state sample data group and a transient sample data group corresponding to the known severity level of thermal defects of the power equipment based on the acquired steady-state sample data and transient sample data, wherein the steady-state sample data group includes the sample temperature data of the power equipment shell, the ambient temperature sample data and the internal temperature difference sample data of the power equipment corresponding to the two, and the transient sample data group includes the temperature rise sample data of the power equipment shell, the ambient temperature rise sample data and the internal temperature rise sample data of the power equipment corresponding to the two; A first feature extraction module uses a steady-state sample data group to train a first neural network to establish a mapping relationship between the power equipment housing sample temperature data, the ambient temperature sample data, and the power equipment internal temperature difference sample data; and outputs the power equipment internal temperature difference prediction data based on the collected power equipment housing actual temperature and the ambient actual temperature; The second feature extraction module uses a transient sample data group to train the second neural network to establish a mapping relationship between the power equipment shell temperature rise sample data, the ambient temperature rise sample data and the power equipment internal temperature rise sample data; and outputs the power equipment internal temperature rise prediction data based on the power equipment shell temperature rise actual measured data and the ambient temperature rise actual measured data.

8. The system of claim 1, wherein: The first neural network and / or the second neural network is a back propagation neural network.

9. The system according to claim 8, characterized in that The back propagation neural network includes an input layer, a hidden layer and an output layer, wherein the hidden layer adopts RELU as an activation function.

10. The system according to claim 8, characterized in that When the back propagation neural network is trained, the Adam algorithm and the cross entropy loss function are used to calculate and update the network weights of the back propagation neural network.

Images