A method and device for analyzing and predicting equipment failure in a power distribution room based on edge computing

Abstract

The present invention discloses a method and device for analyzing and predicting equipment failure in a power distribution room based on edge computing, including collecting state quantity data of a sensing terminal in the power distribution room for normalization processing to form a sample data set; constructing a deep learning model based on a deep convolutional network, inputting the sample data set for recognition training, and stopping the training until the output recognition result is consistent with the training sample label; using the trained deep learning model to perform fusion analysis on the state quantity; respectively identifying and locating multiple local discharge sources and abnormal sounds of equipment for multiple sensing terminals; and judging the fault state of the equipment in the power distribution room based on the fusion analysis results and the identification and positioning results. The present invention effectively reduces the cost of manual inspections, and information and data are obtained in real time; in addition, the accuracy of equipment failure judgment is improved based on the integrated judgment of the fusion analysis results and the identification and positioning results.

Description

Technical Field

[0001] The present invention relates to the field of power distribution network and Internet of Things information technology, and in particular to a method and device for analyzing and predicting equipment failure in a power distribution room based on edge computing. Background Art

[0002] The operating status of equipment is crucial to ensuring the safe operation of the power grid. At present, the main methods used by power grid companies are manual inspections and live inspections. However, there are two problems: 1. Manual inspections are labor-intensive, inefficient, and costly, resulting in heavy workloads for inspectors and high inspection pressure. The contradiction between the rapid growth of the power grid scale and the allocation of equipment operation and maintenance personnel is becoming increasingly prominent; 2. Manual inspections have blind spots, and operation and maintenance personnel cannot carry out all-weather, all-time, and all-round inspections.

[0003] In recent years, in order to solve the problems existing in the operation and inspection of distribution rooms, online monitoring systems and inspection robots have been widely used. However, the existing online monitoring systems for power equipment are costly, inconvenient to install and maintain under power, and difficult to apply on a large scale. The inspection items of inspection robots are relatively small (mainly visible light and infrared temperature measurement), and their functions mainly remain at the stage of data collection and simple threshold judgment, with a low level of intelligence.

[0004] Therefore, the working method to ensure the reliability of power supply of distribution equipment should be changed from "post-emergency repair and inspection" to "pre-warning and active processing", and the status perception and fault prediction should be continuously carried out during operation to eliminate permanent faults before they occur. At present, the main problems in the live detection and diagnosis of distribution room equipment are the heavy workload of operation and inspection personnel, the low cost-effectiveness of detection methods, and the low detection accuracy of old and faulty equipment. Summary of the invention

[0005] The purpose of this section is to summarize some aspects of embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the specification abstract and the invention title of this application to avoid blurring the purpose of this section, the specification abstract and the invention title, and such simplifications or omissions cannot be used to limit the scope of the present invention.

[0006] In view of the above existing problems, the present invention is proposed.

[0007] Therefore, the present invention provides a method for analyzing and predicting equipment failures in a power distribution room based on edge computing, which can solve the problem that failures of existing power distribution equipment cannot be warned in advance.

[0008] In order to solve the above technical problems, the present invention provides the following technical solutions: including, collecting the state quantity data of the distribution room perception terminal and normalizing it to form a sample data set; constructing a deep learning model based on a deep convolutional network, inputting the sample data set for recognition training, and stopping the training until the output recognition result is consistent with the training sample label; using the trained deep learning model to perform fusion analysis on the state quantity; combining the collaborative edge strategy to identify and locate multiple local discharge sources and abnormal sounds of equipment for multiple perception terminals respectively; judging the fault status of the distribution room equipment based on the fusion analysis results and the identification and positioning results.

[0009] As a preferred solution of the method for analyzing and predicting equipment failure in a distribution room based on edge computing described in the present invention, the state quantity data includes visible light, infrared imaging, sound, partial discharge, ambient temperature, ambient humidity, and gas.

[0010] As an optimal solution of the method for analyzing and predicting equipment faults in a distribution room based on edge computing described in the present invention, the sample data set includes a training set, a test set and a verification set; the training set includes the historical state quantity data of the distribution room perception terminal in the past five years; the test set includes the state quantity data of the distribution room perception terminal to be tested; the verification set includes the state quantity data of the distribution room perception terminal in the past two years.

[0011] As an optimal solution of the method for analyzing and predicting equipment faults in a distribution room based on edge computing described in the present invention, the deep learning model is constructed to include a CNN layer and a coding layer; the CNN layer includes a convolutional layer, a pooling layer and a fully connected layer; the coding layer includes a convolutional coding layer and a convolutional decoding layer.

[0012] As an optimal solution of the method for analyzing and predicting equipment faults in a distribution room based on edge computing described in the present invention, wherein: training the deep learning model includes inputting the training set into the deep learning model, the encoding layer performing unsupervised pre-training on the training set to obtain preliminary features of the training set; initializing the convolution layer using the extracted features to obtain the training sample labels; the CNN layer sequentially performs convolution, pooling and full connection processing on the pre-trained training set; the encoding layer performs convolution encoding and decoding processing on the convolution-processed training set, and outputs data that is fed back to the convolution layer; after the CNN layer completes the full connection processing, it outputs the recognition result and determines whether it is consistent with the training sample label.

[0013] As an optimal solution of the method for analyzing and predicting equipment faults in a distribution room based on edge computing described in the present invention, the judgment also includes: if the recognition result is consistent with the training sample label, the iterative training is stopped and the training of the deep learning model is completed; if the recognition result is inconsistent with the training sample label, the iterative training is continued until the training is stopped when the judgment result is consistent, and the deep learning model is output.

[0014] As a preferred solution of the method for analyzing and predicting equipment faults in a distribution room based on edge computing described in the present invention, the fusion analysis includes anomaly detection based on time series clustering analysis, image recognition based on Faat-RCNN and slow mixed Gaussian model, heterogeneous image registration and fusion based on sub-pixel optimization and local intensity feature invariant description, and CNN partial discharge signal identification.

[0015] As a preferred solution of the power distribution room equipment fault analysis and prediction method based on edge computing described in the present invention, wherein: identifying and locating the multiple local discharge sources includes identifying and locating the multiple local discharge sources based on a local discharge collaborative detection strategy based on distributed synchronous measurement and signal strength distribution; based on the propagation attenuation characteristics of the electromagnetic wave signal excited by the local discharge in the power equipment and space and the sensitivity characteristics of the sensor, a local discharge signal amplitude distribution fingerprint library of different sensors is established, as follows,

[0016]

[0017] Where N is the number of measurement points, L is the number of sensors, that is, AP i , i = 1, 2, ..., L, when the local discharge occurs at RP j When the sensor AP i The measured UHF electromagnetic wave signal strength is {φ i,j (τ),τ=1,…,p,p>1}, τ is the number of measurements, and P represents the total number of measurements.

[0018] As a preferred solution of the method for analyzing and predicting equipment faults in a distribution room based on edge computing described in the present invention, identifying and locating abnormal sounds of the equipment includes identifying and locating abnormal sounds of the equipment based on a sound source localization strategy based on multi-sensor time delay sequence analysis.

[0019] As an optimal solution of the device for analyzing and predicting equipment faults in a power distribution room based on edge computing described in the present invention, it includes: an information acquisition module, which is used to collect real-time status quantities and historical status quantities of the operation of each equipment in the distribution room, and it includes a camera and a sensor, the camera is used to capture image information of each equipment, and the sensor is used to perceive the quantity data of the operation status of the equipment; a data processing center module is connected to the information acquisition module, and it includes a computing unit, a training unit and an input-output unit, the computing unit is used to calculate and process the status quantity data transmitted by the information acquisition module, the training unit is used for deep learning and convolution processing of the sample data set, and the input-output unit is used to transmit data streams between each module unit and provide communication feedback services; an analysis and judgment module is connected to the data processing center module, and it is used to receive the result information transmitted by the input-output unit, and analyze and judge whether a fault occurs in the operation status of the equipment in the distribution room.

[0020] The beneficial effects of the present invention are as follows: the present invention can realize full-time and all-round monitoring of parameters such as infrared, video, partial discharge, sound, temperature and humidity through an integrated sensing terminal, effectively replacing the workload of operating personnel for patrols; at the same time, the workload of some live detection personnel is replaced by automatic monitoring and analysis of status parameters, effectively reducing the cost of manual patrols, and obtaining information and data in real time; in addition, the accuracy of equipment fault judgment is improved based on the integrated judgment of fusion analysis results and identification and positioning results. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. Among them:

[0022] Figure 1 A schematic diagram of a flow chart of a method for analyzing and predicting equipment failure in a power distribution room based on edge computing according to a first embodiment of the present invention;

[0023] Figure 2 A schematic diagram of a partial discharge and interference identification process of a deep convolutional network of a method for analyzing and predicting equipment failure in a power distribution room based on edge computing according to the first embodiment of the present invention;

[0024] Figure 3 A schematic diagram of extracting basic characteristic parameters of partial discharge in a method for analyzing and predicting equipment failure in a power distribution room based on edge computing according to the first embodiment of the present invention;

[0025] Figure 4A schematic diagram of characteristic parameters of a pulse unit structure of a method for analyzing and predicting equipment failure in a power distribution room based on edge computing according to the first embodiment of the present invention;

[0026] Figure 5 This is a schematic diagram of the module structure distribution of a device for analyzing and predicting equipment failure in a power distribution room based on edge computing as described in the second embodiment of the present invention. DETAILED DESCRIPTION

[0027] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0028] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0029] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0030] The present invention is described in detail with reference to schematic diagrams. When describing the embodiments of the present invention, for the sake of convenience, the cross-sectional diagrams showing the device structure will not be partially enlarged according to the general scale, and the schematic diagrams are only examples, which should not limit the scope of protection of the present invention. In addition, in actual production, the three-dimensional dimensions of length, width and depth should be included.

[0031] At the same time, in the description of the present invention, it should be noted that the directions or positional relationships indicated by the terms "upper, lower, inner and outer" are based on the directions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as limiting the present invention. In addition, the terms "first, second or third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0032] In the present invention, unless otherwise clearly specified and limited, the terms "install, connect, connect" should be understood in a broad sense, for example: it can be a fixed connection, a detachable connection or an integral connection; it can also be a mechanical connection, an electrical connection or a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0033] Example 1

[0034] Reference Figure 1 to Figure 4 , which is the first embodiment of the present invention, and provides a method for analyzing and predicting equipment failure in a power distribution room based on edge computing, comprising:

[0035] S1: Collect the state quantity data of the sensing terminal in the distribution room and perform normalization processing to form a sample data set. It should be noted that the state quantity data includes:

[0036] Visible light, infrared imaging, sound, partial discharge, ambient temperature, ambient humidity, gas.

[0037] Furthermore, the sample dataset includes:

[0038] The training set includes the historical state data of the sensing terminals in the power distribution room for the past five years;

[0039] The test set includes the state quantity data of the sensing terminal in the power distribution room to be tested;

[0040] The validation set includes the state quantity data of the distribution room sensing terminals in the past two years.

[0041] S2: Build a deep learning model based on a deep convolutional network, input a sample data set for recognition training, and stop training when the output recognition result is consistent with the training sample label. It should be noted that building a deep learning model includes:

[0042] CNN layer and encoding layer;

[0043] CNN layers include convolutional layers, pooling layers, and fully connected layers;

[0044] The coding layer includes a convolutional coding layer and a convolutional decoding layer.

[0045] Specifically, refer to Figure 2 , training a deep learning model involves:

[0046] The training set is input into the deep learning model, and the encoding layer performs unsupervised pre-training on the training set to obtain the preliminary features of the training set;

[0047] Use the extracted features to initialize the convolution layer and obtain the training sample labels;

[0048] The CNN layer performs convolution, pooling, and full connection processing on the pre-trained training set in turn;

[0049] The encoding layer performs convolution encoding and decoding on the convolutional training set, and the output data is fed back to the convolution layer;

[0050] After the CNN layer completes the full connection processing, it outputs the recognition result and determines whether it is consistent with the training sample label.

[0051] Furthermore, the judgment also includes:

[0052] If the recognition result is consistent with the training sample label, the iterative training is stopped and the deep learning model training is completed;

[0053] If the recognition result is inconsistent with the training sample label, the iterative training will continue until the result is consistent, and the training will be stopped and the deep learning model will be output.

[0054] S3: Use the trained deep learning model to perform fusion analysis on the state quantity. It should also be noted that the fusion analysis includes:

[0055] Anomaly detection based on time series clustering analysis, image recognition using Faat-RCNN and slow mixed Gaussian models, heterogeneous image registration and fusion based on sub-pixel optimization and invariant description of local intensity features, and CNN local discharge signal identification.

[0056] S4: Combine the collaborative edge strategy to identify and locate multiple local discharge sources and abnormal device sounds on multiple sensing terminals. It should also be noted that the identification and location of multiple local discharge sources includes:

[0057] A collaborative partial discharge detection strategy based on distributed synchronous measurement and signal strength distribution is used to identify and locate multiple partial discharge sources.

[0058] Specifically, it also includes:

[0059] Based on the propagation attenuation characteristics of electromagnetic wave signals excited by partial discharge inside power equipment and in space and the sensitivity characteristics of sensors, a fingerprint library of partial discharge signal amplitude distribution of different sensors is established, as follows:

[0060]

[0061] Where N is the number of measurement points, L is the number of sensors, that is, AP i , i = 1, 2, ..., L, when the local discharge occurs at RP j When the sensor AP i The measured UHF electromagnetic wave signal strength is {φ i,j(τ),τ=1,…,p,p>1}, τ is the number of measurements, and P represents the total number of measurements.

[0062] Furthermore, identifying and locating abnormal sounds of equipment includes:

[0063] The sound source localization strategy based on multi-sensor time-delay sequence analysis is used to identify and locate abnormal sounds of equipment.

[0064] S5: Determine the fault status of the equipment in the power distribution room based on the fusion analysis results and the identification and positioning results. It should also be noted that:

[0065] The amplitude of the partial discharge UHF signal collected synchronously by different sensing terminals in the power distribution room is used to calculate the best matching discharge source position using a multi-layer neural network model to obtain the positioning result.

[0066] Using the fast matching and association search algorithm of big data analysis, similar test case data can be searched according to voltage level, equipment type, and manufacturer as needed, and sorted according to the degree of matching;

[0067] The variational Bayesian autoencoder is used to extract the features of partial discharge data, and the matching degree between partial discharge data is obtained by combining the cosine strategy.

[0068] The matching degree of the positioning results and the detection cases are screened respectively to obtain the final judgment result of the fault status of the equipment in the distribution room.

[0069] It should also be noted that the local discharge signal analysis and diagnosis algorithm based on the deep convolutional network can accurately identify the local discharge type and typical interference. Identifying and diagnosing the type and severity of local discharge and identifying various types of interference are of great significance to the local discharge data analysis. The deep convolutional network (CNN) has obvious advantages in large image processing. The PRPD spectrum and PRPS data commonly used in local discharge detection and analysis are essentially large two-dimensional matrices, which have certain similarities with the data format of digital images. Due to the influence of field conditions, the data in the matrix will also have phase offsets and different amplitudes. The deep convolutional network has the invariance of translation, scaling, and distortion of the input samples. It should be further noted that the parameter selection of the convolutional neural network and the training of the sample set are completed on the cloud platform, and the edge computing model embedded in the integrated sensor terminal is configured and updated using the cloud platform to ensure the accuracy of the computing model.

[0070] Preferably, refer to Figure 3The collection of partial discharge ultra-high frequency signals includes: the envelope detection waveform of the partial discharge pulse is conditioned and input into the high-speed ADC, and the output continuous digital signal stream is subjected to real-time feature extraction by FPGA to obtain the waveform characteristic parameters of the discharge pulse; the three basic characteristic parameters of the partial discharge pulse signal are pulse amplitude, discharge time interval and discharge interval, and the basic characteristic parameters are statistically processed to form a two-dimensional or three-dimensional spectrum of statistical parameters.

[0071] according to Figure 3 As shown in the figure, the continuous data stream output by the ADC is sent to the FPGA for pulse recognition processing. When the pulse waveform crosses the current trigger level upward, it is in the capture state. When the pulse waveform crosses the current trigger level downward, it exits the capture state. The maximum value in the capture state is the peak value of the pulse. The capture state time length is the pulse width. The distance from the capture state start point to the peak point is the rise time, and the distance from the peak point to the capture state exit time is the fall time. The pulse width is used to distinguish communication signals and effective discharge pulses. Its characteristic is that the maximum width of the effective discharge pulse is only within a certain value range, while the width of the communication interference signal is often relatively large. The data such as the rise time, fall time, peak value and trigger level are used to calculate the half-height width and triangular shape parameters of the discharge envelope pulse waveform.

[0072] Reference Figure 4 The minimum unit of the discharge pulse characteristic data stream is the pulse characteristic unit structure, which includes basic characteristic parameters such as pulse number, pulse time, pulse peak value, rise width, fall width, pulse time-frequency characteristic parameters, trigger level, etc. The pulse characteristic unit data structure is organized in units of cycles in the time domain to form a data stream in the time domain. The pulse unit structure is of fixed length, and the cycle unit structure is of indefinite length. For example, there is no pulse characteristic unit in a certain cycle. At this time, only the pulse unit with the pane number of 0 and other fields all being 0 is output, indicating that this cycle unit is an empty cycle. The role of the empty cycle is to make the data stream continuous in time; according to Figure 4 The diagram is a description of the characteristic parameters of the pulse unit structure, where the intrinsic characteristic quantity is the half-wave width, which needs to be obtained based on the wPeak, wRisingWidth, wFallingWidth, and wTrigLevel parameters in the pulse unit structure combined with the similar triangle method.

[0073] Preferably, in order to better verify and illustrate the technical effects adopted in the method of the present invention, this embodiment chooses to compare the traditional method for judging latent faults of distribution network equipment with the method of the present invention, and compares the test results by means of scientific demonstration to verify the real effect of the method of the present invention; the traditional method for judging latent faults of distribution network equipment has low sensitivity for insulation degradation or partial discharge detection, and the judgment of equipment fault status is inaccurate. In order to verify that the method of the present invention has higher equipment judgment accuracy than the traditional method, the traditional method and the method of the present invention will be used in this embodiment to analyze and judge the fault status of the distribution room equipment in a certain area in the south.

[0074] Test environment: (1) Input the running status of the equipment in the distribution room into the PSAPAC software for simulation operation, simulating the operation of the distribution network switch cabinet, distribution network cable, and distribution network transformer respectively;

[0075] (2) Using the equipment operation status detection cases of the distribution network in the region as experimental data, 100 groups were selected for normalization and training, and finally 10 groups of data with high training degree and good normalization were determined as test sets;

[0076] (3) Two judgment methods were used to analyze the experimental data. Each method tested 10 groups of data, and the state recognition value of each group of data was calculated. The error was compared with the state value of the simulation input. The results are shown in the following table.

[0077] Table 1: Comparison of the accuracy results of the two methods.

[0078]

[0079] It should be noted that the traditional method for judging latent faults of distribution network equipment changes the phase voltage of the distribution network through the phase power supply, and combines the partial discharge monitoring device to judge whether the partial discharge detection start conditions are met based on the voltage difference of the phase voltage obtained twice in a row. When the partial discharge amount conditions of the measuring equipment are met, the measurement results are uploaded to the status evaluation device, and then the fault status judgment of the detection equipment is obtained.

[0080] Referring to Table 1, it can be seen intuitively that the traditional method has a low accuracy in identifying the partial discharge state quantity, thus generating a large error, which affects the judgment result of the equipment fault state. Compared with the traditional method, the method of the present invention performs fusion analysis and identification and positioning on the partial discharge state quantity, thereby improving the recognition accuracy of the state quantity, and thus generating a smaller error, which reflects that the method of the present invention has a high judgment accuracy.

[0081] Example 2

[0082] Reference Figure 5, which is the second embodiment of the present invention. This embodiment is different from the first embodiment in that it provides a device for analyzing and predicting equipment failure in a power distribution room based on edge computing, including:

[0083] The information collection module 100 is used to collect the real-time and historical status quantities of each device in the distribution room. It includes a camera 101 and a sensor 102. The camera 101 is used to capture the image information of each device, and the sensor 102 is used to sense the quantitative data of the device's operating status.

[0084] The data processing center module 200 is connected to the information acquisition module 100, and includes a computing unit 201, a training unit 202 and an input-output unit 203. The computing unit 201 is used to calculate and process the state quantity data transmitted by the information acquisition module 100, the training unit 202 is used for deep learning and convolution processing of sample data sets, and the input-output unit 203 is used to transmit data flows between each module unit and provide communication feedback services.

[0085] The analysis and judgment module 300 is connected to the data processing center module 200, and is used to receive the result information transmitted by the input and output unit 203, and analyze and judge whether a fault occurs in the operating status of the equipment in the power distribution room.

[0086] It should be appreciated that embodiments of the present invention may be implemented or enforced by computer hardware, a combination of hardware and software, or by computer instructions stored in a non-transitory computer-readable memory. The method may be implemented in a computer program using standard programming techniques - including a non-transitory computer-readable storage medium configured with a computer program, wherein the storage medium so configured causes the computer to operate in a specific and predefined manner - according to the methods and drawings described in the specific embodiments. Each program may be implemented in a high-level procedural or object-oriented programming language to communicate with a computer system. However, if desired, the program may be implemented in an assembly or machine language. In any case, the language may be a compiled or interpreted language. In addition, the program may be run on a programmed ASIC for this purpose.

[0087] Furthermore, the operations of the processes described herein may be performed in any suitable order unless otherwise indicated herein or otherwise clearly contradicted by context. The processes described herein (or variations and / or combinations thereof) may be performed under the control of one or more computer systems configured with executable instructions, and may be implemented as code (e.g., executable instructions, one or more computer programs, or one or more applications) that is executed collectively on one or more processors, by hardware, or a combination thereof. The computer program includes a plurality of instructions that may be executed by one or more processors.

[0088] Further, the method can be implemented in any type of computing platform that is operably connected to a suitable computer, including but not limited to a personal computer, a minicomputer, a mainframe, a workstation, a network or distributed computing environment, a separate or integrated computer platform, or in communication with a charged particle tool or other imaging device, etc. Aspects of the present invention can be implemented in machine-readable code stored on a non-transitory storage medium or device, whether removable or integrated into a computing platform, such as a hard disk, an optical read and / or write storage medium, a RAM, a ROM, etc., so that it can be read by a programmable computer, and when the storage medium or device is read by the computer, it can be used to configure and operate the computer to perform the process described herein. In addition, the machine-readable code, or part thereof, can be transmitted via a wired or wireless network. When such media includes instructions or programs that implement the steps described above in conjunction with a microprocessor or other data processor, the invention described herein includes these and other different types of non-transitory computer-readable storage media. When programmed according to the methods and techniques described in the present invention, the present invention also includes the computer itself. The computer program can be applied to input data to perform the functions described herein, thereby converting the input data to generate output data stored in a non-volatile memory. The output information can also be applied to one or more output devices such as a display. In a preferred embodiment of the present invention, the converted data represents a physical and tangible object, including a specific visual depiction of the physical and tangible object produced on a display.

[0089] As used in this application, the terms "component", "module", "system", etc. are intended to refer to a computer-related entity, which can be hardware, firmware, a combination of hardware and software, software, or software in operation. For example, a component can be, but is not limited to: a process running on a processor, a processor, an object, an executable file, a thread in execution, a program, and / or a computer. As an example, an application running on a computing device and the computing device can both be components. One or more components can exist in a process and / or thread in execution, and a component can be located in a computer and / or distributed between two or more computers. In addition, these components can be executed from various computer-readable media having various data structures thereon. These components can communicate in a local and / or remote process manner, such as according to a signal having one or more data packets (e.g., data from a component that interacts with another component in a local system, a distributed system, and / or interacts with other systems in a signal manner through a network such as the Internet).

[0090] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for analyzing and predicting equipment failure in a power distribution room based on edge computing, characterized in that: include, Collect the state quantity data of the sensing terminal in the power distribution room and perform normalization processing to form a sample data set; Building a deep learning model based on a deep convolutional network, inputting the sample data set for recognition training, and stopping the training when the output recognition result is consistent with the training sample label; Using the trained deep learning model to perform fusion analysis on the state quantity; Combined with the collaborative edge strategy, multiple sensing terminals are respectively used to identify and locate multiple local discharge sources and abnormal sounds of equipment; Determine the fault status of the equipment in the power distribution room based on the fusion analysis results and identification and positioning results; Among them, the amplitude of the partial discharge ultra-high frequency signal collected synchronously by different sensing terminals in the distribution room is used to calculate the most matching discharge source position using a multi-layer neural network model to obtain the positioning result; Using the fast matching and association search algorithm of big data analysis, similar test case data can be searched according to voltage level, equipment type, and manufacturer as needed, and sorted according to the degree of matching; The variational Bayesian autoencoder is used to extract the features of partial discharge data, and the matching degree between partial discharge data is obtained by combining the cosine strategy. The matching degree of the positioning results and the detection cases are screened respectively to obtain the final judgment result of the fault status of the equipment in the distribution room.

2. The method for analyzing and predicting equipment failure in a power distribution room based on edge computing according to claim 1, characterized in that: The state quantity data includes: Visible light, infrared imaging, sound, partial discharge, ambient temperature, ambient humidity, gas.

3. The method for analyzing and predicting equipment failure in a power distribution room based on edge computing according to claim 1 or 2, characterized in that: The sample data set includes a training set, a test set and a validation set; The training set includes historical state quantity data of the sensing terminal in the power distribution room in the past five years; The test set includes the state quantity data of the sensing terminal in the power distribution room to be tested; The verification set includes the state quantity data of the distribution room sensing terminal in the past two years.

4. The method for analyzing and predicting equipment failure in a power distribution room based on edge computing according to claim 3, characterized in that: Constructing the deep learning model includes a CNN layer and a coding layer; The CNN layer includes a convolutional layer, a pooling layer and a fully connected layer; The coding layer includes a convolutional coding layer and a convolutional decoding layer.

5. The method for analyzing and predicting equipment failure in a power distribution room based on edge computing according to claim 4, characterized in that: Training the deep learning model includes, The training set is input into the deep learning model, and the encoding layer performs unsupervised pre-training on the training set to obtain preliminary features of the training set; Initializing the convolution layer using the extracted features to obtain the training sample labels; The CNN layer sequentially performs convolution, pooling and full connection processing on the pre-trained training set; The encoding layer performs convolution encoding and decoding processing on the training set subjected to convolution processing, and outputs data which is fed back to the convolution layer; After completing the full connection processing, the CNN layer outputs the recognition result and determines whether it is consistent with the training sample label.

6. The method for analyzing and predicting equipment failure in a power distribution room based on edge computing according to claim 5, characterized in that: The judgment also includes: If the recognition result is consistent with the training sample label, the iterative training is stopped and the deep learning model training is completed; If the recognition result is inconsistent with the training sample label, the iterative training is continued until the judgment result is consistent, and the training is stopped and the deep learning model is output.

7. The method for analyzing and predicting equipment failure in a power distribution room based on edge computing according to claim 6, characterized in that: The fusion analysis includes: Anomaly detection based on time series clustering analysis, image recognition using Faat-RCNN and slow mixed Gaussian models, heterogeneous image registration and fusion based on sub-pixel optimization and invariant description of local intensity features, and CNN local discharge signal identification.

8. The method for analyzing and predicting equipment failure in a power distribution room based on edge computing according to claim 7, characterized in that: Identifying and locating the multiple local discharge sources includes: Identifying and locating the multiple local discharge sources based on a local discharge collaborative detection strategy based on distributed synchronous measurement and signal strength distribution; Based on the propagation attenuation characteristics of electromagnetic wave signals excited by partial discharge inside power equipment and in space and the sensitivity characteristics of sensors, a fingerprint library of partial discharge signal amplitude distribution of different sensors is established, as follows: Where N is the number of measurement points, L is the number of sensors, that is, AP i , i = 1, 2, ..., L, when the local discharge occurs at RP j When the sensor AP i The measured UHF electromagnetic wave signal strength is τ is the number of measurements, and P is the total number of measurements.

9. The method for analyzing and predicting equipment failure in a power distribution room based on edge computing according to claim 8, characterized in that: Identifying and locating abnormal sounds of the device includes: The abnormal sound of the equipment is identified and located by a sound source localization strategy based on multi-sensor time-delay sequence analysis.

10. A device for analyzing and predicting equipment failure in a power distribution room based on edge computing, based on the method for analyzing and predicting equipment failure in a power distribution room based on edge computing according to any one of claims 1 to 9, characterized in that: include, An information collection module (100) is used to collect real-time state quantities and historical state quantities of each device in the power distribution room, and comprises a camera (101) and a sensor (102), wherein the camera (101) is used to capture image information of each device, and the sensor (102) is used to sense quantitative data of the device's operating state; The data processing center module (200) is connected to the information acquisition module (100), and comprises a calculation unit (201), a training unit (202) and an input-output unit (203); the calculation unit (201) is used for calculating and processing the state quantity data transmitted by the information acquisition module (100); the training unit (202) is used for deep learning and convolution processing of the sample data set; and the input-output unit (203) is used for transmitting data streams between the module units and providing communication feedback services; The analysis and judgment module (300) is connected to the data processing center module (200) and is used to receive the result information transmitted by the input and output unit (203) and analyze and judge whether a fault occurs in the operating status of the equipment in the power distribution room.

Images