Power equipment fire identification method and system, electronic equipment and medium

Through technologies such as multi-sensor arrays and one-dimensional convolutional neural networks, the local discharge types of power transmission and distribution equipment are identified, which solves the problem of inaccurate identification in the existing technology, and improves the maintenance reliability and fire prevention and control capabilities of power equipment.

CN119089160AInactive Publication Date: 2024-12-06GUANGDONG POWER GRID CO LTD +1

Patent Information

Application Number
CN202411116363.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-14
Publication Date
2024-12-06
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The prior art cannot accurately identify the type of local discharge in power transmission and distribution equipment, resulting in the inability to effectively carry out maintenance and increase the risk of fire.

Method used

Data is collected by multi-sensor array, spatial correlation and temporal correlation models are established, spatial correlation features are extracted through one-dimensional convolutional neural network and compression-activation method, local discharge types are identified, combined with microphone array data for in-depth analysis, and discharges are identified Location and type.

Benefits of technology

It realizes accurate identification of local discharge types, reduces fire risks, and improves the maintenance reliability and detection accuracy of power equipment.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119089160A_ABST
    Figure CN119089160A_ABST
Patent Text Reader

Abstract

The invention discloses a fire identification method and system for power equipment, electronic equipment and a medium, and the method comprises a partial discharge type identification method based on time-space correlation feature fusion, a multi-sensor array can be used for identifying the partial discharge position and the discharge type, microphone array data are fully utilized, the partial discharge type is identified, and the partial discharge type is identified. Therefore, the application of the microphone array in the partial discharge detection of the power equipment is further expanded, so that the partial discharge type can be judged according to the detection, and appropriate power fire prevention and control measures can be taken to maintain the power transmission and distribution equipment. The technical problem that the power transmission and distribution equipment cannot be maintained due to the fact that the partial discharge phenomenon exists in the power transmission and distribution equipment and the discharge type cannot be accurately identified is solved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the field of electric power equipment fire identification, and in particular, relates to an electric power equipment fire identification method, system, electronic equipment and medium. Background Art

[0002] The transmission and distribution equipment is the intermediate medium to maintain the normal operation of the power system. Through the transmission and distribution of electric energy, it can realize functions such as voltage level conversion, electric energy regulation, power grid protection, current or voltage measurement, and abnormal control. Due to changes in the power consumption at the power consumption end and during operation, as well as changes in the distribution voltage, the transmission and distribution equipment is subjected to changes in the distribution pressure, causing power equipment failures.

[0003] In the prior art, among the various factors that cause power transmission and distribution equipment, direct equipment damage is mostly caused by local discharge in the power transmission and distribution equipment, which deteriorates the insulation performance of the power equipment. If it is not maintained, there is a risk of fire. Summary of the invention

[0004] The embodiments of the present invention provide a method, system, electronic device and medium for identifying fire in electric power equipment, so as to at least solve the technical problem that partial discharge exists in power transmission and distribution equipment, the discharge type cannot be accurately identified, and thus the maintenance of power transmission and distribution equipment cannot be performed.

[0005] According to a first aspect of an embodiment of the present invention, there is provided a method for identifying fires in electric equipment, comprising: using a multi-sensor array to collect partial discharge data of the electric equipment to obtain sensor data, wherein the sensor data is used to generate spatial data and temporal data, the multi-sensor array comprising a plurality of groups of sensors, each group of sensors forming a channel;

[0006] Establish a correlation model between multi-sensor data. The correlation model is used to determine the spatial correlation model and time correlation model of each sensor. The spatial correlation model is used to determine the spatial correlation of multiple sensors, and the time correlation model is used to determine the time correlation of multiple sensors. The spatial correlation is used to characterize that the signal characteristics of local signals for different discharge types are different. The redundancy and complementarity between signals from different channels further enhance the characteristics of various types of partial discharge signals. The time correlation is used to characterize that the partial discharge signal of each channel is a time series, and its changes in the time domain are affected by different signal partial discharge types, forming a partial discharge signal with time correlation characteristics.

[0007] The spatial correlation and the temporal correlation are subjected to feature fusion to obtain a feature fusion result;

[0008] According to the feature fusion results, the relationship between spatial correlation and temporal correlation is identified to obtain the type of partial discharge;

[0009] For different discharge types, signals are collected through multiple sensors and divided into n*m samples, where n is the number of channels and m is the number of samples. The time taken by the sensor to collect target data is used to determine the discharge position;

[0010] Different discharge types are classified and analyzed, and the parameters n and m are collected for the discharge types of gap discharge, corona discharge and spherical discharge respectively;

[0011] Based on local signals received by sensors at different locations;

[0012] Identify partial discharge types based on spatiotemporal correlation feature fusion;

[0013] First: Based on the one-dimensional convolutional neural network algorithm, the one-dimensional convolutional neural network uses convolution operations to extract local patterns and features in the input data, selects a specific dimension, and performs pooling processing to achieve dimensionality reduction and local feature extraction and compression;

[0014] Second: A compression-activation-based correlation mining method maps the correlation relationship on a specific dimension to a corresponding weight coefficient, and extracts feature correlation by weighting the features in this direction;

[0015] Third: Extraction and fusion of spatiotemporal correlation features based on one-dimensional convolutional neural network and compression-activation correlation mining method.

[0016] Optionally, the sensor includes one or more of the following: a sound sensor, a temperature sensor, a smoke sensor, and a carbon monoxide sensor.

[0017] Optionally, the local discharge type includes a gapless discharge type and a discharge type.

[0018] Optionally, the convolution operation expression in the one-dimensional convolutional neural network algorithm is Among them, s is the input feature, W and b represent the convolution kernel and bias parameters respectively.

[0019] Optionally, the compression-activation-based association mining method includes the following steps: compressing the information except the direction of the association relationship to be obtained; using a full connection and nonlinear function as the activation function; using X c '=F scale (w c ,X c )=w c ·X c Weight the feature vector X, where F scale (w c ,X c ) is the weight w c With the eigenvector X cThe product between .

[0020] Optionally, a method for extracting and fusing spatiotemporal correlation features based on a one-dimensional convolutional neural network and a compression-activation correlation mining method includes the following steps: determining the convolution direction of the one-dimensional convolutional neural network; using a compression-activation method to mine correlation relationships in a specific dimension; and mining and fusing the spatial and temporal correlations of partial discharge array data based on the convolution direction and the correlation relationship to identify the type of partial discharge.

[0021] Optionally, the method for identifying fires in power equipment also includes: analyzing the impact of imbalance on partial discharge data; processing the unbalanced data; and verifying the effect of processing the unbalanced data.

[0022] According to a second aspect of an embodiment of the present invention, there is also provided a power equipment fire identification system, comprising:

[0023] The acquisition module is used to collect partial discharge data of power equipment using a multi-sensor array to obtain sensor data, wherein the sensor data is used to generate spatial data and temporal data, and the multi-sensor array includes multiple groups of sensors, each group of sensors forming a channel; the construction module is used to establish an association relationship model between the multi-sensor data, the association relationship model is used to determine the spatial correlation model and the temporal correlation model of each sensor, the spatial correlation model is used to determine the spatial correlation of the multiple sensors, and the temporal correlation model is used to determine the temporal correlation of the multiple sensors, wherein the spatial correlation is used to characterize that the signal characteristics generated by its local signal for different discharge types are different, and the redundancy and complementarity between signals of different channels further enhance the characteristics of various types of partial discharge signals, and the temporal correlation is used to characterize that the partial discharge signal of each channel is a time series, and its change in the time domain is affected by different signal partial discharge types, forming a partial discharge signal with time correlation characteristics; the fusion module is used to perform feature fusion of the spatial correlation and the temporal correlation to obtain a feature fusion result; the identification module is used to identify the relationship between the spatial correlation and the temporal correlation according to the feature fusion result to obtain the type of partial discharge.

[0024] According to a third aspect of an embodiment of the present invention, there is also provided an electronic device, comprising a memory and a processor, wherein a computer program is stored in the memory, and the processor is configured to run the computer program to execute the method for identifying fires in electric power equipment as described in any embodiment of the first aspect above.

[0025] According to a fourth aspect of the embodiments of the present invention, a computer program product is also provided, including a computer program, which, when executed by a processor, implements the method for identifying fires in electric power equipment as described in any embodiment of the first aspect above.

[0026] According to a fifth aspect of an embodiment of the present invention, a non-volatile storage medium is further provided, in which a computer program is stored, wherein the computer program is configured to execute the method for identifying fires in electric power equipment described in any embodiment of the first aspect when running on a computer or a processor.

[0027] In an embodiment of the present invention, a partial discharge type identification method based on the fusion of spatiotemporal correlation features can be used to identify the location and type of partial discharge by using a multi-sensor array. By making full use of the microphone array data, the partial discharge type is identified, thereby further expanding the application of microphone arrays in partial discharge detection of power equipment, so that the partial discharge type can be judged according to the detection and appropriate power fire prevention measures can be taken. Through in-depth analysis of multi-sensor array data and summarizing it into spatial correlation and temporal correlation, the partial discharge type of power equipment is identified by extracting and fusing spatial and temporal correlations based on a one-dimensional convolutional neural network and a compression-activation method. For the problem of partial discharge data imbalance, the loss function adjustment method and the data distribution adjustment method are used to train the unbalanced acoustic array partial discharge data, and the proposed method is verified from the perspective of correlation analysis and imbalance processing by simulation, thereby ensuring that the detection data is more accurate and the reliability of use is improved in the later maintenance and detection process. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0029] Figure 1 is a flow chart of a method for identifying fire in electric power equipment according to one embodiment of the present invention;

[0030] Figure 2 It is various partial discharge waveform diagrams in the present invention;

[0031] Figure 3 It is a schematic diagram showing that the microphone array data in the present invention has spatial correlation and temporal correlation;

[0032] Figure 4 It is a schematic diagram of a one-dimensional convolutional neural network in the present invention;

[0033] Figure 5 It is a structural schematic diagram of the compression-activation association relationship extraction method of the present invention;

[0034] Figure 6 It is a schematic diagram of a partial discharge type identification model based on spatiotemporal feature fusion in the present invention;

[0035] Figure 7The figure is a structural block diagram of a power equipment fire identification system according to one embodiment of the present invention. DETAILED DESCRIPTION

[0036] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0037] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the terms used in this way can be interchangeable where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0038] According to an embodiment of the present invention, an embodiment of a method for identifying fire in electric power equipment is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system containing at least one set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0039] The method embodiment can also be executed in an electronic device including a memory and a processor, a similar control device or a cloud. Taking an electronic device as an example, the electronic device may include one or more processors and a memory for storing data. Optionally, the electronic device may also include a communication device and a display device for communication functions. It can be understood by those of ordinary skill in the art that the above structural description is only illustrative and does not limit the structure of the above electronic device. For example, the electronic device may also include more or fewer components than the above structural description, or have a configuration different from the above structural description.

[0040] The processor may include one or more processing units. For example, the processor may include a central processing unit (CPU), a graphics processing unit (GPU), a digital signal processing (DSP) chip, a microcontroller unit (MCU), a field-programmable gate array (FPGA), a neural network processor (NPU), a tensor processing unit (TPU), an artificial intelligence (AI) type processor, and the like. Among them, different processing units may be independent components or integrated into one or more processors. In some instances, the electronic device may also include one or more processors.

[0041] The memory can be used to store computer programs, for example, to store the computer program corresponding to the method for identifying fires in power equipment in the embodiment of the present invention. The processor implements the above-mentioned method for identifying fires in power equipment by running the computer program stored in the memory. The memory may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory may further include a memory remotely arranged relative to the processor, and these remote memories may be connected to the electronic device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0042] The communication device is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of the mobile terminal. In one example, the communication device includes a network adapter (network interface controller, NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the communication device can be a radio frequency (radio frequency, RF) module, which is used to communicate with the Internet wirelessly. In some embodiments of the present solution, the communication device is used to connect to mobile devices such as mobile phones and tablets, and instructions can be sent to electronic devices through the mobile device.

[0043] The display device may be a touch screen type liquid crystal display (LCD) and a touch display (also referred to as a "touch screen" or "touch display screen"). The liquid crystal display may enable a user to interact with a user interface of the electronic device. In some embodiments, the electronic device has a graphical user interface (GUI), and a user may interact with the GUI by touching a finger release and / or gestures on a touch-sensitive surface, and executable instructions for executing the above-mentioned human-computer interaction functions are configured / stored in a computer program product or a readable storage medium executable by one or more processors.

[0044] Figure 1 FIG. 1 is a flow chart of a method for identifying a fire in an electric power device according to one embodiment of the present invention. Figure 1 As shown, the method comprises the following steps:

[0045] Step 1: Using a multi-sensor array to collect partial discharge data, the collected data is used to generate spatial data and temporal data;

[0046] Step 2: Establish a correlation model between multi-sensor data, analyze and determine the model of spatial and temporal correlation of each sensor, and analyze the spatial correlation and temporal correlation of multi-sensors;

[0047] Step 3: Fuse the spatial correlation and temporal correlation features;

[0048] Step 4: Identify the relationship between spatial correlation and temporal correlation, and identify the type of partial discharge, including gap discharge, corona discharge and spherical discharge;

[0049] Step 5: Based on step 4, the multi-sensor array is set into n groups to form n channels;

[0050] Step 6: For different discharge types, multiple sensors are used to collect signals and divide them into n*m samples, where n is the number of channels and m is the number of samples. The time taken by the sensor to collect target data is used to determine the discharge position;

[0051] Step 7: Classify and analyze different discharge types, and collect parameters n and m for the discharge types of gap discharge, corona discharge and spherical discharge respectively;

[0052] Step 8: Based on the local signals received by sensors at different locations;

[0053] Spatial correlation: The signal characteristics of partial discharge signals for different discharge types are different. The redundancy and complementarity between signals from different channels enhance the characteristics of various partial discharge signals.

[0054] Time correlation: The partial discharge signal of each channel is a time series, and its change in the time domain is affected by different types of partial discharge signals, forming a partial discharge signal with time correlation characteristics;

[0055] Step 9: Identify the type of partial discharge based on the fusion of spatiotemporal correlation features;

[0056] Step 10: Data imbalance analysis and processing;

[0057] Step 11: Verify the effect of spatiotemporal correlation feature fusion.

[0058] Partial discharge brings great harm and requires frequent maintenance. The present application realizes the detection of partial discharge release signals based on the accompanying sound, heat, light and electromagnetic wave changes when partial discharge occurs. The detection means used for partial discharge include but are not limited to the detection of partial discharge signals through sound sensors, temperature sensors, smoke sensors and CO sensors.

[0059] Among them, the detection method based on acoustic signals has a wide range of applications due to its good anti-electromagnetic interference, non-destructiveness, low cost, etc. The acoustic detection method can be divided into two categories: distributed measurement method and centralized measurement method. The former requires multiple acoustic sensors to be attached to different positions on the outside of the power equipment for data collection and analysis, and is mostly used for internal partial discharge detection of power equipment. It has high accuracy, but the operation is time-consuming and inconvenient; the latter integrates multiple acoustic sensors together to form an acoustic array, which is mostly used for external partial discharge detection of substations, GIS and other equipment. The advantages are non-contact and flexible operation.

[0060] This application preferentially uses acoustic signal detection as an example, and realizes the monitoring of partial discharge signals through a multi-acoustic sensor array. On the one hand, the signal collected by the microphone array is a multi-channel signal, and each channel reflects the partial discharge characteristic information from a different position; on the other hand, the signal of each channel is a time series, which reflects the partial discharge characteristic information from the time domain change of the signal.

[0061] Partial discharge data acquisition based on microphone array:

[0062] When the acoustic sensor monitors different types of partial discharge, different discharge types produce different waveforms, such as Figure 2 As shown, the collected signals are segmented to obtain acoustic array data of different types of partial discharge phenomena, which are combined with video monitoring to determine the location of the partial discharge.

[0063] Based on the above content, by selecting a more obvious sound signal and converting the binary value into an amplification signal, such as Figure 2 The waveform shown.

[0064] Analysis based on microphone array data correlation:

[0065] like Figure 3 As shown, the microphone array is composed of multi-channel acoustic sensors. The partial discharge signals received by sensors at different positions are correlated with their spatial layout, which reflects the characteristics of partial discharge signals from different spatial positions. The redundancy and complementarity between signals from different channels can further enhance the characteristics of each category of partial discharge signals. Therefore, the partial discharge signals of different channels have spatial correlation.

[0066] The partial discharge acoustic signal of each channel is a time series, and its changes in the time domain reflect the characteristics of different types of partial discharge signals. Therefore, the partial discharge signals of the same channel at different sampling times have time correlation.

[0067] In summary, the multi-channel data collected by the microphone array has spatial correlation and temporal correlation.

[0068] In the identification of partial discharge types, in order to realize the identification of partial discharge types, the partial discharge type identification method based on the fusion of spatiotemporal correlation features includes:

[0069] First: Based on the one-dimensional convolutional neural network algorithm, the one-dimensional convolutional neural network uses convolution operations to extract local patterns and features in the input data, selects a specific dimension, and performs pooling processing to achieve dimensionality reduction and local feature extraction and compression;

[0070] According to the formula And refer to Figure 4 As shown in the figure, s is the input feature, W and b represent the convolution kernel and bias parameters respectively. In a one-dimensional convolutional neural network, the size of W is only optional in the convolution dimension.

[0071] In a one-dimensional convolutional neural network, the size of W is only selectable in the convolution dimension. Taking the two-dimensional input feature as an example, Figure 4 Describes the convolution operation of a one-dimensional convolutional neural network of size 2 in different dimensions. When the convolution direction is dimension 1, d1(W) = 2, d2(W) = d2(s); when the convolution direction is dimension 2, d1(W) = d1(s), d2(W) = 2. i (W) represents the size of W in dimension i.

[0072] Therefore, the acoustic array data has two dimensions, which have spatial correlation and temporal correlation respectively. The one-dimensional convolutional neural network can extract corresponding features in a specific dimension, so the one-dimensional convolutional neural network can be used to extract correlation in the time or space dimension respectively.

[0073] Second: A compression-activation-based correlation mining method maps the correlation relationship on a specific dimension to a corresponding weight coefficient, and extracts feature correlation by weighting the features in this direction;

[0074] The specific steps include:

[0075] Step 91: Compression. The compression process compresses the information except the direction of the association relationship to be obtained, so that only the direction feature is retained. Suppose the feature to be analyzed is a three-dimensional vector It is necessary to mine the correlation between C channels in dimension 3, so the information of other dimensions is compressed so that it only has information in dimension 3. Commonly used compression methods include global average pooling, global maximum pooling, etc. Taking average pooling as an example, use

[0076] c=1,2…C compresses the information on dimension 1 and dimension 2. After compression, the feature vector is obtained Corresponding to C channels in dimension 3;

[0077] Step 92: Activation, in order to extract the correlation information between the channels and map it to the corresponding weight information, a full connection and nonlinear function activation method is usually used. In order to extract the correlation information between the channels and map it to the corresponding weight information, a full connection and nonlinear function activation method is used;

[0078] like Figure 5 As shown, the activation module is based on the feature vector As input, with weight information The output is a fully connected layer, and the association relationship is extracted in the middle. Since each output node in the fully connected layer is the weighted and nonlinear mapping of all nodes in the input layer, the output weight of the activation module reflects the association relationship of the input channel information. In the activation process, two fully connected layers are usually used to extract the association relationship and map the association relationship weight coefficient. In order to reduce the number of parameters, the number of nodes in the first fully connected layer is C / r, and r is the parameter adjustment coefficient, which can be adjusted according to the size of C;

[0079] Step 93: Using X' c =F scale (w c ,X c )=w c ·X c Weight the feature vector X, where F scale (w c ,X c ) is the weight w c With the eigenvector X c The product between is the feature of dimension 3 channel c; w c is the cth element in W, which is X c The weight of .

[0080] Third: Extraction and fusion of spatiotemporal correlation features based on one-dimensional convolutional neural network and compression-activation correlation mining method;

[0081] The one-dimensional convolutional neural network can select a specific convolution direction, and the compression-activation method can effectively mine the correlation relationship in a specific dimension. Therefore, this embodiment combines the two to mine and fuse the spatial and temporal correlations of the partial discharge array data in turn to realize the partial discharge type recognition method.

[0082] like Figure 6 As shown in the figure, the proposed method first extracts the spatial correlation relationship of the microphone array data, and then extracts the temporal correlation relationship based on it. Finally, the feature vector after the fusion of spatiotemporal correlation features is further analyzed to realize the partial discharge type identification.

[0083] Figure 6 In the figure, conv1d(n,f,s) represents n one-dimensional convolution operations with size f and stride s; avgpool1d(f,s) represents a one-dimensional global pooling operation with size f and stride s; FC(x) represents a fully connected layer with x neurons; Relu and Sigmoid are different activation methods.

[0084] like Figure 6 As shown, since the sampling frequency of the acoustic array data is high and the data volume is large, the time dimension in each sample has rich information. Therefore, in order to obtain as much information as possible under different compression methods, this embodiment proposes a time dimension information compression method using one-dimensional convolution + global average pooling.

[0085] Example 1: Use 32 one-dimensional convolution kernels to perform convolution in the spatial dimension, that is, use 32 different methods to compress the 4000 sampling data of each sound sensor to obtain a feature vector of size 16*32, and then use one-dimensional global average pooling to perform information statistics on the 32 types of compression results of each sound sensor to obtain spatial feature information of size 16. Then, in order to activate the spatial features, two fully connected layers of size 16 are used to extract the correlation between each sensor, and the second fully connected layer is used as the output of the spatial correlation weight. Considering that the number of channels C = 16 and the amount of data is small, the parameter adjustment coefficient r is set to 1. Finally, use X' c =F scale (w c ,X c )=w c ·X c The input information of size 16*4000 is weighted in spatial dimension so that it has associated information in spatial dimension.

[0086] Example 2: This embodiment uses a microphone array with a size of 4000 in the time dimension. The data volume is large, and considering that the time series is correlated in adjacent time periods, feature extraction of data in adjacent time periods is beneficial to extracting local information. Therefore, 32 one-dimensional convolution kernels with a size of 5 and a step size of 5 are used to perform convolution operations on the feature vector extracted by spatial correlation in the time dimension, thereby extracting information from 5 adjacent sampling points and obtaining a feature vector with a size of 32*800.

[0087] For the 32*800 feature vector, we first use the one-dimensional global average pooling method to compress the 32 information of each channel in the time dimension to obtain a feature vector of size 800. Then, we use the fully connected layers of size 400 and 800 to mine the association between 800 time periods (each time period corresponds to 5 original sampling moments), and obtain the time association weight information of size 800. Considering that the size of the time dimension information is 800 and the amount of data is large, the parameter adjustment coefficient r=2 is set, so the size of the first fully connected layer in the activation method is 400. Finally, X' c =F scale (w c ,X c )=w c ·X c The information of size 32*800 is weighted by time relevance so that it has relevant information in the time dimension.

[0088] Since the 32*800 feature vector is a feature vector mined through spatial correlation, the spatial and temporal correlation feature fusion of the microphone array data is achieved here.

[0089] In summary, the acoustic array data is used to extract spatial correlation relations, and on this basis, the temporal correlation relations are extracted to achieve the fusion of spatiotemporal correlation features, so that the two types of correlation features can complement and enhance each other.

[0090] In order to further extract and analyze the temporal and spatial correlation features, the feature vector after temporal and spatial correlation feature extraction and fusion is used as input, and two fully connected layers of size 1024 are used to further extract features, and then a fully connected layer of size 4 and softmax are used to identify the PD type.

[0091] The data imbalance analysis and processing method includes the following steps:

[0092] Step 101: Analyze the impact of imbalance on partial discharge data;

[0093] In practice, partial discharge is affected by the production process, operation and maintenance conditions, so the collected samples are often unbalanced. The partial discharge type recognition model based on spatiotemporal feature fusion in this embodiment is trained by using the training environment of Tensorflow2.11+Python 3.8.6, the optimization function of RMSprop, the learning rate of 0.0001, and the rho parameter of 0.9.

[0094] Step 102: Processing unbalanced data;

[0095] Among them, the processing methods include a loss function adjustment method and a data distribution adjustment method.

[0096] like As shown in Figure 1, in multi-classification models, cross loss entropy is usually used to measure the difference between the prediction and the true label, and the model is trained by minimizing the cross loss entropy. However, since the cross loss entropy is averaged over all samples, when the categories are unbalanced, the loss function is dominated by the category with more samples, causing the model to overfit to such samples, while the classification performance of the category with fewer samples decreases.

[0097] As FL(p i )=-(1-p i ) γ log(p i ), this embodiment uses Focal loss to train unbalanced partial discharge data. Focal loss solves the data imbalance problem by introducing a modulation factor γ (modulation coefficient), thereby improving the model's classification ability for minority categories. For the i-th sample, the probability that the model correctly predicts it is p i , then when its predicted probability is high, (1-p i ) γ is smaller, so the weight of the sample in the loss is smaller; conversely, when the model has a smaller ability to correctly predict it, (1-p i ) γ The larger the sample, the larger the corresponding weight in the loss function. Therefore, the model can allocate loss weights according to the predictive ability of the samples, and pay more attention to samples with poor predictive ability, thereby avoiding the phenomenon that the loss function is dominated by a large number of categories.

[0098] To address the imbalance of partial discharge data, the purpose of data distribution adjustment can also be achieved by adjusting the number of various samples used for training. In practice, partial discharge and no partial discharge phenomena need to be clearly distinguished, and the discharge phenomenon is more important than the no discharge phenomenon, so no adjustment is made to the no partial discharge phenomenon.

[0099] Step 103: Verify the effect of unbalanced data processing.

[0100] The proposed PD type recognition model is trained by combining the methods of step 101 and step 102, and the test results are compared with the situation in section 3.1 where sample imbalance is not considered, to verify the necessity of considering sample imbalance.

[0101] For the verification of the effect of the fusion of time-space correlation features in step 11, by comparing the partial discharge type recognition effects of considering both time correlation and space correlation, considering only time correlation, considering only space correlation, and not considering correlation characteristics, the necessity of time-space correlation fusion in the identification of partial discharge types of acoustic arrays is verified.

[0102] It can be understood that the above-mentioned embodiments of the present invention can identify the partial discharge type of the power equipment through the algorithm, and then monitor and maintain the power equipment according to the discharge type of the power equipment to avoid fire in the power equipment.

[0103] Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus a necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), including a number of instructions for a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods described in each embodiment of the present invention.

[0104] In this embodiment, a power equipment fire identification system is also provided, which is used to implement the above-mentioned embodiments and preferred implementation modes, and the descriptions that have been made are not repeated here. As used below, the term "module" is a combination of software and / or hardware that can implement a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0105] Figure 7 FIG. 2 is a structural block diagram of a power equipment fire identification system 200 according to one embodiment of the present invention. Figure 7As shown, an example of a fire identification system 200 for electric equipment is provided, which includes: an acquisition module 201, which is used to acquire partial discharge data of electric equipment using a multi-sensor array to obtain sensor data, wherein the sensor data is used to generate spatial data and temporal data, and the multi-sensor array includes multiple groups of sensors, each group of sensors forming a channel; a construction module 202, which is used to establish an association relationship model between the multi-sensor data, and the association relationship model is used to determine the spatial correlation model and the temporal correlation model of each sensor, and the spatial correlation model is used to determine the spatial correlation of multiple sensors, and the temporal correlation model is used to determine the temporal correlation of multiple sensors. Among them, the spatial correlation is used to characterize that the signal characteristics of the local signal for different discharge types are different, and the redundancy and complementarity between signals of different channels further enhance the characteristics of various types of partial discharge signals. The time correlation is used to characterize that the partial discharge signal of each channel is a time series, and its change in the time domain is affected by different signal partial discharge types, forming a partial discharge signal with time correlation characteristics; the fusion module 203 is used to perform feature fusion of the spatial correlation and the time correlation to obtain a feature fusion result; the identification module 204 is used to identify the relationship between the spatial correlation and the time correlation according to the feature fusion result to obtain the local discharge type.

[0106] An embodiment of the present invention further provides an electronic device, including a memory and a processor, wherein a computer program is stored in the memory, and the processor is configured to run the computer program to execute the method for identifying fire in electric power equipment described in any of the above embodiments.

[0107] Optionally, in this embodiment, the processor is configured to run the computer program to perform the following steps:

[0108] Step 1: Use a multi-sensor array to collect partial discharge data of power equipment to obtain sensor data, wherein the sensor data is used to generate spatial data and time data. The multi-sensor array includes multiple groups of sensors, and each group of sensors forms a channel;

[0109] Step 2: Establish a correlation model between multi-sensor data. The correlation model is used to determine the spatial correlation model and temporal correlation model of each sensor. The spatial correlation model is used to determine the spatial correlation of multiple sensors, and the temporal correlation model is used to determine the temporal correlation of multiple sensors. The spatial correlation is used to characterize that the signal characteristics of local signals for different discharge types are different. The redundancy and complementarity between signals from different channels further enhance the characteristics of various types of partial discharge signals. The temporal correlation is used to characterize that the partial discharge signal of each channel is a time series, and its changes in the time domain are affected by different signal partial discharge types, forming a partial discharge signal with time correlation characteristics.

[0110] Step 3: Fuse the spatial correlation and temporal correlation features to obtain the feature fusion result;

[0111] Step 4: Identify the relationship between spatial correlation and temporal correlation based on the feature fusion results to obtain the local discharge type.

[0112] An embodiment of the present invention further provides a computer program product, including a computer program, which, when executed by a processor, implements the method for identifying a fire in electric power equipment as described in any of the above embodiments.

[0113] Optionally, in this embodiment, the above computer program implements the following steps when executed by a processor:

[0114] Step 1: Use a multi-sensor array to collect partial discharge data of power equipment to obtain sensor data, wherein the sensor data is used to generate spatial data and time data. The multi-sensor array includes multiple groups of sensors, and each group of sensors forms a channel;

[0115] Step 2: Establish a correlation model between multi-sensor data. The correlation model is used to determine the spatial correlation model and temporal correlation model of each sensor. The spatial correlation model is used to determine the spatial correlation of multiple sensors, and the temporal correlation model is used to determine the temporal correlation of multiple sensors. The spatial correlation is used to characterize that the signal characteristics of local signals for different discharge types are different. The redundancy and complementarity between signals from different channels further enhance the characteristics of various types of partial discharge signals. The temporal correlation is used to characterize that the partial discharge signal of each channel is a time series, and its changes in the time domain are affected by different signal partial discharge types, forming a partial discharge signal with time correlation characteristics.

[0116] Step 3: Fuse the spatial correlation and temporal correlation features to obtain the feature fusion result;

[0117] Step 4: Identify the relationship between spatial correlation and temporal correlation based on the feature fusion results to obtain the local discharge type.

[0118] Optionally, the specific examples in this embodiment may refer to the examples described in the above embodiments and optional implementation modes, and this embodiment will not be described in detail here.

[0119] An embodiment of the present invention further provides a non-volatile storage medium, in which a computer program is stored, wherein the computer program is configured to execute the power equipment fire identification method described in any of the above embodiments when running on a computer or a processor.

[0120] Optionally, in this embodiment, the above-mentioned computer program may be configured to store a computer program for performing the following steps:

[0121] Step 1: Use a multi-sensor array to collect partial discharge data of power equipment to obtain sensor data, wherein the sensor data is used to generate spatial data and time data. The multi-sensor array includes multiple groups of sensors, and each group of sensors forms a channel;

[0122] Step 2: Establish a correlation model between multi-sensor data. The correlation model is used to determine the spatial correlation model and temporal correlation model of each sensor. The spatial correlation model is used to determine the spatial correlation of multiple sensors, and the temporal correlation model is used to determine the temporal correlation of multiple sensors. The spatial correlation is used to characterize that the signal characteristics of local signals for different discharge types are different. The redundancy and complementarity between signals from different channels further enhance the characteristics of various types of partial discharge signals. The temporal correlation is used to characterize that the partial discharge signal of each channel is a time series, and its changes in the time domain are affected by different signal partial discharge types, forming a partial discharge signal with time correlation characteristics.

[0123] Step 3: Fuse the spatial correlation and temporal correlation features to obtain the feature fusion result;

[0124] Step 4: Identify the relationship between spatial correlation and temporal correlation based on the feature fusion results to obtain the local discharge type.

[0125] Optionally, the specific examples in this embodiment may refer to the examples described in the above embodiments and optional implementation modes, and this embodiment will not be described in detail here.

[0126] In the above embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0127] In some embodiments provided in the present application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the modules can be a logical function division. There may be other division methods in actual implementation. For example, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, modules or indirect coupling or communication connection of modules, which can be electrical or other forms.

[0128] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed on multiple modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment.

[0129] In addition, each functional module in each embodiment of the present invention may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of software functional modules.

[0130] If the integrated module is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, disk or optical disk and other media that can store program codes.

[0131] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for identifying fire in electric power equipment, characterized in that: include: A multi-sensor array is used to collect partial discharge data of power equipment to obtain sensor data, wherein the sensor data is used to generate spatial data and temporal data, and the multi-sensor array includes multiple groups of sensors, each group of sensors forming a channel; Establishing a correlation model between multi-sensor data, wherein the correlation model is used to determine the spatial correlation model and time correlation of each sensor, the spatial correlation model is used to determine the spatial correlation of multiple sensors, and the time correlation model is used to determine the time correlation of multiple sensors, wherein the spatial correlation is used to characterize that the signal characteristics of the partial discharge signal for different discharge types are different, and the redundancy and complementarity between signals of different channels enhance the characteristics of various types of partial discharge signals, and the time correlation is used to characterize that the partial discharge signal of each channel is a time series, and its change in the time domain is affected by different signal partial discharge types, forming a partial discharge signal with time correlation characteristics; Performing feature fusion on the spatial correlation and the temporal correlation to obtain a feature fusion result; Identifying the relationship between the spatial correlation and the temporal correlation according to the feature fusion result to obtain a partial discharge type; First: Based on the one-dimensional convolutional neural network algorithm, the one-dimensional convolutional neural network uses convolution operations to extract local patterns and features in the input data, selects a specific dimension, and performs pooling processing to achieve dimensionality reduction and local feature extraction and compression; Second: A compression-activation-based correlation mining method maps the correlation relationship on a specific dimension to a corresponding weight coefficient, and extracts feature correlation by weighting the features in this direction; Third: Based on the one-dimensional convolutional neural network and the compression-activation correlation mining method, the spatiotemporal correlation features are extracted and fused to realize the identification of local discharge types.

2. The method for identifying fire in electric power equipment according to claim 1, characterized in that: The sensor includes one or more of the following: a sound sensor, a temperature sensor, a smoke sensor, and a carbon monoxide sensor.

3. The method for identifying fire in electric power equipment according to claim 1, characterized in that: The partial discharge types include a gapless discharge type and a partial discharge type.

4. The method for identifying fire in electric power equipment according to claim 1, characterized in that: The convolution operation expression in the one-dimensional convolutional neural network algorithm is: Among them, s is the input feature, W and b represent the convolution kernel and bias parameters respectively.

5. The method for identifying fire in electric power equipment according to claim 1, characterized in that: The compression-activation based association mining method includes the following steps: Compress the information except the direction of the association relationship to be obtained; The activation function uses full connection and nonlinear function; Using X' c =F scale (w c ,X c )=w c ·X c Weight the feature vector X, where F scale (w c ,X c ) is the weight w c With the eigenvector X c The product between .

6. The method for identifying fire in electric power equipment according to claim 1, characterized in that: The method for extracting and fusing spatiotemporal correlation features based on a one-dimensional convolutional neural network and a compression-activation correlation mining method includes the following steps: Determine the convolution direction of the one-dimensional convolutional neural network; Use compression-activation method to mine associations on specific dimensions; According to the convolution direction and the correlation relationship, the spatial and temporal correlation of the partial discharge array data is mined and fused to identify the partial discharge type.

7. The method for identifying fire in electric power equipment according to claim 1, characterized in that: Also includes: Analysis of the impact of imbalance on partial discharge data; Handling imbalanced data; Verify the effect of unbalanced data processing.

8. The method for identifying fire in electric power equipment according to claim 6, characterized in that: The specific dimensions include spatial correlation and temporal correlation respectively, and the one-dimensional convolutional neural network extracts correlation in the temporal or spatial dimension respectively.

9. A fire identification system for electric power equipment, characterized in that: include: A collection module, used to collect partial discharge data of power equipment using a multi-sensor array to obtain sensor data, wherein the sensor data is used to generate spatial data and time data, and the multi-sensor array includes multiple groups of sensors, each group of sensors forming a channel; A construction module is used to establish an association relationship model between multi-sensor data, wherein the association relationship model is used to determine the spatial association model and the temporal association model of each sensor, wherein the spatial association model is used to determine the spatial association of the multi-sensors, and the temporal association model is used to determine the temporal association of the multi-sensors, wherein the spatial association is used to characterize that the signal characteristics of the partial discharge signal for different discharge types are different, and the redundancy and complementarity between signals of different channels enhance the characteristics of various types of partial discharge signals, and the temporal association is used to characterize that the partial discharge signal of each channel is a time series, and its change in the time domain is affected by the partial discharge types of different signals, thereby forming a partial discharge signal with a time correlation characteristic; A fusion module, used for performing feature fusion on the spatial correlation and the temporal correlation to obtain a feature fusion result; An identification module is used to identify the relationship between the spatial correlation and the temporal correlation according to the feature fusion result to obtain the local discharge type.

10. An electronic device comprising a memory and a processor, characterized in that: The memory stores a computer program, and the processor is configured to run the computer program to execute the method for identifying a fire in electric power equipment as described in any one of claims 1 to 8.

11. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the computer program implements the steps of the method for identifying a fire in electric power equipment according to any one of claims 1 to 8.

12. A non-volatile storage medium, characterized in that: The non-volatile storage medium stores a computer program, wherein the computer program is configured to execute the method for identifying a fire in electric power equipment as described in any one of claims 1 to 8 when running on a computer or a processor.

Citation Information

Patent Citations

  • Different insulation defect partial discharge type distinguishing method, system and device and storage medium

    CN113449803A

  • Partial discharge diagnosing apparatus

    JP1989182765A

  • Method and system for implementing temporal convolution in spatiotemporal neural networks

    WO2023250093A1

Cited By

  • A method for analyzing the root causes of business application system failures based on deep learning

    CN119782029A