A sound source positioning resolution enhancement method, apparatus, device and medium

Abstract

The present application relates to a kind of sound source positioning resolution enhancement method, device, equipment and medium, belong to the field of power equipment intelligent operation and inspection, wherein the method includes the following steps: obtaining substation equipment partial discharge sound wave signal, and generating initial image;With physical constraint, convolutional neural network is constructed, with initial image as input, enhanced image is output by learning high-dimensional features, wherein the joint loss function of convolutional neural network includes reconstruction loss, physical constraint loss and smoothing loss, and physical constraint loss is determined based on delay constraint physical model and attenuation constraint physical model;Based on enhanced image, the location of sound source is identified, and the accurate positioning of substation partial discharge sound source is realized.Compared with prior art, the present application can improve the spatial resolution of fault point in low signal-to-noise ratio environment, make the fault point range more concentrated, and easily locate the discharge sound source.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of intelligent operation and inspection of power equipment, and in particular to a sound source positioning resolution enhancement method, device, equipment and medium. BACKGROUND

[0002] Key power equipment in substations (such as transformers, GIS, switch cabinets, cable terminals, etc.) are prone to aging, dampness or manufacturing defects in the internal insulating medium under the long-term action of high voltage, large current and complex environmental stress, and thus partial discharge (PD) is triggered. Partial discharge is an important early form of equipment insulation deterioration and a warning signal, and effective detection and accurate positioning thereof are key links for assessing equipment insulation state, preventing major failures and ensuring safe and stable operation of power grids. Among numerous partial discharge detection technologies, acoustic imaging has been widely used in the field of live detection and condition monitoring of substation equipment in recent years due to its significant advantages such as non-contact, long-distance, intuitive visualization of positioning, strong anti-electromagnetic interference capability, etc. This technology is based on a microphone array to collect sound wave signals on the surface or in the space of equipment, and uses beamforming or similar algorithms to convert sound pressure information into a spatially distributed sound source intensity image, thereby realizing visual positioning of the discharge point.

[0003] Chinese patent CN109116203A discloses a power equipment partial discharge fault diagnosis method based on a convolutional neural network, which converts PRPD data of partial discharge into a grayscale image, uses a convolutional neural network with residual structure to extract recognition features of the grayscale image, and applies the features to a classic classifier, thereby effectively combining deep learning methods and traditional machine learning methods, improving the recognition degree of the extracted features and the accuracy of fault diagnosis. However, when acoustic imaging technology is applied to partial discharge detection in actual substation complex field, a series of severe challenges still exist, which seriously restricts the accuracy and reliability of the detection:

[0004] (1) Weak signal and strong noise environment: the acoustic signal generated by partial discharge itself is weak (usually in the tens to hundreds of micro-pascal level), and the frequency spectrum is widely distributed (usually in the ultrasonic frequency band, such as 20 kHz-150 kHz). The substation site is filled with various strong background noise sources, such as cooling fan noise, transformer core vibration noise, corona noise, environmental wind noise, traffic noise, etc. These noises are highly overlapped with the target PD signal in time and frequency domains, resulting in extremely low signal-to-noise ratio (SNR) of the collected signal, and conventional filtering methods are difficult to effectively separate the target signal, which seriously affects the detection sensitivity.

[0005] (2) Insufficient spatial resolution and difficulty in distinguishing dense defects: Limited by the acoustic diffraction limit (related to wavelength and array aperture) and the constraints of actual array size and the number of array elements, the spatial resolution of traditional acoustic imaging systems is limited (usually at the centimeter to decimeter level). When there are multiple adjacent partial discharge source points in the device or there are complex geometric structures (such as inside the GIS cavity, the gap between the transformer windings), the imaging results often show a fuzzy "hot spot" area, which cannot clearly distinguish the specific location and number of adjacent defects, making it difficult to accurately locate and identify defects.

[0006] (3) Imaging distortion and error caused by complex acoustic wave propagation: The path of acoustic wave propagation from the partial discharge source inside the device to the array sensor is complex and variable. When acoustic waves propagate in multiple layers of media in the device, reflection, refraction, scattering, attenuation and mode conversion occur. Most existing conventional acoustic imaging algorithms are based on simplified propagation models in free space or homogeneous media, which seriously ignore these complex physical propagation effects. This leads to a shift in the location of the reconstructed acoustic source image, distortion in intensity, and even the appearance of false acoustic sources, making it difficult to meet the high requirements of diagnostic needs, especially for the error of the discharge point deep inside the device.

[0007] In summary, in the application of substation partial discharge detection, the existing acoustic imaging technology is limited by low signal-to-noise ratio, limited resolution, and the lack of modeling of complex acoustic propagation physical mechanisms, and its positioning accuracy, anti-interference ability and resolution ability for dense / interior defects need to be improved. SUMMARY

[0008] The purpose of the present application is to overcome the defects of the prior art and provide a sound source positioning resolution enhancement method, device, equipment and medium.

[0009] The purpose of the present application can be achieved by the following technical solutions:

[0010] According to a first aspect of the present application, a sound source positioning resolution enhancement method is provided, the method comprising the following steps:

[0011] Acquiring a substation device partial discharge acoustic wave signal and generating an initial image;

[0012] Constructing a convolutional neural network with physical constraints, taking the initial image as input, and outputting an enhanced image through learning high-dimensional features, wherein the joint loss function of the convolutional neural network includes a reconstruction loss, a physical constraint loss and a smoothing loss, and the physical constraint loss is determined based on a delay constraint physical model and an attenuation constraint physical model;

[0013] Identifying the sound source location based on the enhanced image to achieve accurate positioning of the substation partial discharge sound source.

[0014] The delay constraint physical model is represented as: wherein, denotes the coordinate of an arbitrary point in the space, denotes the spatial position coordinate of the i-th microphone, denotes the spatial position coordinate of the i-th microphone, denotes the propagation speed of the sound wave in the medium, denotes the time required for the sound wave to propagate from the position r to the i-th microphone.

[0015] The attenuation constraint physical model is represented as: wherein, denotes the distance from the sound source position to the i-th microphone, denotes the coordinate of an arbitrary point in the space, denotes the spatial position coordinate of the i-th microphone, denotes the spatial position coordinate of the i-th microphone, denotes a preset small positive number, denotes the attenuation amount of the sound wave propagating from the position to the i-th microphone.

[0016] The joint loss function is represented as:

[0017] ,

[0018] wherein, is the joint loss function; is the reconstruction loss, obtained by calculating the L2 norm of the enhanced image and the initial image; is the physical constraint loss, determined based on the enhanced image, the delay constraint physical model and the attenuation constraint physical model; is the smoothing loss, obtained by calculating the L1 norm of the gradient of the enhanced image; , , are weight coefficients of the reconstruction loss, the physical constraint loss and the smoothing loss, respectively.

[0019] The physical constraint loss is represented as:

[0020] ,

[0021] wherein, is the enhanced image value corresponding to the point in the space; denotes the attenuation amount of the sound wave propagating from the position to the i-th microphone, determined by the attenuation constraint physical model; ​​​​The number of microphones of an acoustic detection device used for acquiring a partial discharge acoustic wave signal of a substation equipment; is a convolution operation, simulating the time evolution effect of acoustic wave propagation; represents the time required for the acoustic wave to propagate from the position r to the first microphone, determined by a delay constraint physical model.

[0022] The method for identifying the acoustic source position based on the enhanced image is:

[0023] A peak detection algorithm is used to screen out local maximum points from the enhanced image, and the peak point coordinates are determined as the partial discharge acoustic source position:

[0024] , ,

[0025] wherein, represents the enhanced image, represents the neighborhood of , represents the coordinate point position of the real acoustic source, determined based on the resolution of the acoustic detection device.

[0026] The method for generating the initial image is: estimating the acoustic source spatial spectrum by using a traditional beamforming algorithm to generate the initial image.

[0027] According to a second aspect of the present application, there is provided an acoustic source positioning resolution enhancement device for implementing the method, the device comprising:

[0028] A data acquisition and initial image generation module: acquiring a partial discharge acoustic wave signal of a substation equipment and generating an initial image;

[0029] An enhanced image generation module: constructing a convolutional neural network with physical constraints, taking the initial image as input, and outputting an enhanced image by learning high-dimensional features, wherein the joint loss function of the convolutional neural network includes a reconstruction loss, a physical constraint loss, and a smoothing loss, and the physical constraint loss is determined based on a delay constraint physical model and an attenuation constraint physical model;

[0030] A positioning module: identifying the acoustic source position based on the enhanced image to realize accurate positioning of the partial discharge acoustic source of the substation.

[0031] According to a third aspect of the present application, there is provided an electronic device comprising a memory and a processor, the memory having a computer program stored thereon, and the processor implementing the method when executing the program.

[0032] According to a fourth aspect of the present application, a computer readable storage medium is provided, having stored thereon a computer program which, when executed by a processor, implements the method.

[0033] Compared with the prior art, the present application has the following beneficial effects:

[0034] The present application realizes sound source positioning resolution enhancement by combining a physically constrained convolutional neural network, adopts three loss cooperative joint loss functions in the convolutional neural network, uses a reconstruction loss to ensure positioning accuracy, uses a physical constraint loss to force compliance with acoustic laws, and uses a smoothing loss to suppress artifacts, thereby solving the problems of energy leakage and false positioning points in traditional methods, and the introduction of the physical constraint can solve the resolution limitation of the traditional beamforming algorithm under a low signal-to-noise ratio, and still maintain physical consistency under a strong electromagnetic interference environment of a substation. Therefore, the present application can effectively compensate for the defects of low signal-to-noise ratio, low resolution, and insufficient estimation performance in conventional algorithms, thereby significantly improving the spatial resolution of fault points, and quickly positioning the defect fault points of power equipment. BRIEF DESCRIPTION OF DRAWINGS

[0035] Figure 1 A flowchart of the method of the present application is shown in Figure 1.

[0036] Figure 2 A flowchart of the output of the sound source positioning enhancement image of the physically constrained convolutional neural network of the present application is shown in Figure 2.

[0037] Figure 3 A conventional spectral acoustic imaging positioning map is shown in Figure 3.

[0038] Figure 4 An enhanced spectral acoustic imaging positioning map output by the present application is shown in Figure 4.

[0039] Figure 5 A positioning calculation rate comparison chart of the conventional spectral acoustic imaging and the enhanced spectral acoustic imaging of the present application is shown in Figure 5. DETAILED DESCRIPTION

[0040] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor should fall within the scope of protection of the present application.

[0041] Unless otherwise defined, technical terms and scientific terms used in the present application shall have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. The terms "a", "an", "one", "this", and similar referents in the context of describing the application are to be construed to be open-ended, referring to one or more than one, unless otherwise noted. The terms "including", "comprising", "having" and variations thereof in this application are meant to encompass the possibility that there are additional steps, units, etc. that are not listed, or that are not otherwise inherent to the process, method, product, or apparatus. The terms "connected", "coupled", and "pathway" are not restricted to direct or physical connections, but can include indirect connections, unless otherwise noted. The term "plurality" means two or more. The term "and / or" describes associated objects, and can mean that three conditions exist, for example, "A and / or B" can mean A exists alone, B exists alone, or both A and B exist. The term " / " generally means "or". The terms "first", "second", "third", etc. are used to distinguish similar objects, and do not necessarily indicate a specific order.

[0042] Embodiment 1

[0043] This embodiment provides a sound source positioning resolution enhancement method, as shown in Figure 1 The method comprises the following steps:

[0044] S1, obtaining a substation equipment partial discharge sound wave signal and generating an initial image.

[0045] S11, acoustic data acquisition.

[0046] The substation equipment partial discharge sound wave signal is collected using an acoustic array to form a time domain signal matrix:

[0047] ,

[0048] wherein, represents a signal vector collected by the acoustic array at time , which is a column vector containing the signal of each microphone in the array; is the number of microphones of the acoustic detection device; is the sound wave signal received by the th microphone at time .

[0049] S12, generating an initial image.

[0050] In this embodiment, the traditional beamforming algorithm is used to estimate the spatial spectrum of the sound source (i.e. the initial image):

[0051] ,

[0052] wherein, represents the sound source intensity at the spatial position , i.e. the output power of the beamformer; represents the coordinates of any point in space, usually three-dimensional coordinates; represents the spatial position coordinates of the th microphone; represents the propagation speed of sound waves in the medium, represents the Euclidean distance from the position to the position of the th microphone .

[0053] S2, a convolutional neural network with physical constraints is constructed, taking the initial image as input, and outputting an enhanced image through learning high-dimensional features.

[0054] First, a delay constraint physical model is established:

[0055] ,

[0056] wherein, represents the time required for sound waves to propagate from position r to the th microphone.

[0057] Second, an attenuation constraint physical model is established:

[0058] ,

[0059] wherein , represents the distance from the sound source position to the th microphone, represents a preset positive minimum number, represents the attenuation amount of sound waves propagating from position to the th microphone.

[0060] After that, a convolutional neural network with physical constraints is constructed, and its joint loss function includes reconstruction loss, physical constraint loss and smoothing loss, represented as:

[0061] ,

[0062] wherein, is the joint loss function; The reconstruction loss is obtained by calculating the L2 norm of the enhanced image and the initial image. The physical constraint loss is obtained by calculating the L2 norm of the enhanced image and the initial image. The smoothing loss is obtained by calculating the L1 norm of the gradient of the enhanced image. are the weight coefficients of the reconstruction loss, the physical constraint loss, and the smoothing loss, respectively.

[0063]

[0064]

[0065]

[0066] wherein, is the spatial midpoint corresponding to the enhanced image value; represents the attenuation amount of the sound wave propagating from position to the first microphone, which is determined by the attenuation constraint physical model; is a convolution operation that simulates the time evolution effect of sound wave propagation; represents the time required for the sound wave to propagate from position r to the first microphone, which is determined by the delay constraint physical model; represents the gradient of .

[0067] As shown in Figure 2 , according to the physical constraint modeling, the joint loss function is calculated, and when the joint loss function value of the convolutional neural network tends to 0, the network coefficients of each point on the neural network training set tend to the true value. If not, the learning parameters of the network are updated through error backpropagation.

[0068] After training the network on the labeled data set, it is deployed to an embedded edge device to realize real-time output of high-resolution sound source images.

[0069] S3, based on the enhanced image, identifies the sound source position, and realizes accurate positioning of the partial discharge sound source in the substation.

[0070] Specifically, for the enhanced image , a peak detection algorithm is used to screen out local maximum points, and the peak point coordinates are determined as the partial discharge sound source position:

[0071] ,​​​​​​

[0072] wherein, represents an enhanced image, represents a neighborhood of, represents a coordinate point position of a real sound source, is determined based on a resolution of an acoustic detection device, and is usually less than half of the target minimum resolution distance.

[0073] Embodiment 2

[0074] This embodiment is based on the scheme of Embodiment 1, and 2 sound source signal generators are arranged in a laboratory, and the horizontal spatial distance between the 2 sound sources is set to 30 cm. The sound source positions are identified by using a conventional spectrum positioning method and an enhanced spectrum positioning method of the application respectively, and the spatial positioning effects of the above two algorithms are compared through a matlab simulation program. The simulation diagram using the conventional spectrum positioning method is as shown in Figure 3 , the simulation diagram using the enhanced spectrum positioning method is as shown in Figure 4 , and the positioning rate comparison diagram of the two is as shown in Figure 5 . From the simulation results, it can be concluded that the sound source positioning resolution enhancement method based on the physical constraint convolutional neural network of the application not only improves the spatial resolution of the sound positioning method, but also reduces the computational workload.

[0075] Embodiment 3

[0076] The embodiment provides a sound source positioning resolution enhancement device for implementing the method described in Embodiment 1. The device comprises:

[0077] a data acquisition and initial image generation module for acquiring substation equipment partial discharge sound wave signals and generating an initial image;

[0078] an enhanced image generation module for constructing a convolutional neural network with physical constraints, taking the initial image as input, and outputting an enhanced image by learning high-dimensional features, wherein the joint loss function of the convolutional neural network comprises a reconstruction loss, a physical constraint loss and a smoothing loss, and the physical constraint loss is determined based on a delay constraint physical model and an attenuation constraint physical model;

[0079] a positioning module for identifying the sound source position based on the enhanced image to realize accurate positioning of the substation partial discharge sound source.

[0080] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the described modules can refer to the corresponding process in the foregoing method embodiments, which will not be described here.

[0081] Embodiment 4

[0082] The electronic device of the present application includes a central processing unit (CPU) that can perform various appropriate actions and processes in accordance with computer program instructions stored in a read-only memory (ROM) or computer program instructions loaded from a storage unit into a random access memory (RAM). Various programs and data required for device operation can also be stored in the RAM. The CPU, ROM, and RAM are connected to each other by a bus. An input / output (I / O) interface is also connected to the bus.

[0083] A plurality of components in the device are connected to the I / O interface, including: an input unit such as a keyboard, mouse, etc.; an output unit such as various types of displays, speakers, etc.; a storage unit such as a magnetic disk, optical disk, etc.; and a communication unit such as a network card, modem, wireless communication transceiver, etc. The communication unit allows the device to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0084] The processing unit performs the various methods and processes described above, such as methods S1-S3. For example, in some embodiments, methods S1-S3 can be implemented as a computer software program that is tangibly embodied in a machine-readable medium, such as the storage unit. In some embodiments, part or all of the computer program can be loaded and / or installed on the device via the ROM and / or the communication unit. When the computer program is loaded into the RAM and executed by the CPU, one or more steps of methods S1-S3 described above can be performed. Alternatively, in other embodiments, the CPU can be configured to perform methods S1-S3 by any other appropriate means, such as by means of firmware.

[0085] The functions described above in the specification can be implemented entirely by one or more hardware logic components. For example, and without limitation, non-limiting examples of hardware logic components that can be used include field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), system on a chip (SOCs), complex programmable logic devices (CPLDs), etc.

[0086] Program code for carrying out methods of the present application can be written in any combination of one or more programming languages. The program code can be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the program code, when executed by the processor or controller, produces a means for implementing the functions / acts specified in the flowcharts and / or block diagrams. The program code can be executed entirely on a machine, partially on a machine, partially on a machine and partially on a remote machine or entirely on a remote machine or server.

[0087] In the context of the present application, a machine-readable medium can be a tangible medium that can contain or store a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. Machine-readable storage medium can include, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of machine-readable storage medium would include one or more lines of electrical wire, portable computer diskette, hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or Flash memory), optical fiber, portable compact disc read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination of the foregoing.

[0088] The above description is only specific embodiments of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present application, and these modifications or replacements should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for enhancing the resolution of sound source localization, characterized in that, The method includes the following steps: Acquire partial discharge acoustic signals from substation equipment and generate initial images; A convolutional neural network with physical constraints is constructed. Taking the initial image as input, the network enhances the image by learning high-dimensional features. The joint loss function of the convolutional neural network includes reconstruction loss, physical constraint loss, and smoothing loss. The physical constraint loss is determined based on a delay constraint physical model and a decay constraint physical model. Based on the enhanced image recognition of the sound source location, accurate localization of the partial discharge sound source in the substation can be achieved; The physical model for attenuation constraints is expressed as follows: ,in , indicating the location of the sound source To the The distance of one microphone, Represents the coordinates of any point in space. Indicates the first The spatial coordinates of each microphone Represents a pre-defined small positive number. Indicates the position of the sound wave Spread to the Attenuation of each microphone; The physical constraint loss is expressed as: ， in, For physical constraint loss, Midpoint of space The corresponding enhanced image values; The number of microphones in the acoustic detection device used to acquire partial discharge acoustic signals from substation equipment; This is a convolution operation used to simulate the time-varying effect of sound wave propagation. Indicates the position of the sound wave r Spread to the The time required for each microphone is determined by the delay-constrained physical model; The method for identifying the sound source location based on the enhanced image is as follows: The enhanced image is processed using a peak detection algorithm to identify local maxima, and the coordinates of these peaks are determined as the locations of the partial discharge sound sources. ， ， in, Indicates an enhanced image. express The neighborhood, This represents the coordinates of the actual sound source. Resolution determination based on acoustic detection device.

2. The method for enhancing sound source localization resolution according to claim 1, characterized in that, The delay-constrained physical model is represented as follows: ,in, This indicates the speed at which sound waves propagate in a medium.

3. The method for enhancing sound source localization resolution according to claim 1, characterized in that, The joint loss function is expressed as: ， in, For the joint loss function; The reconstruction loss is obtained by calculating the L2 norm of the enhanced image and the original image; The physical constraint loss is determined based on the enhanced image, the delay-constrained physical model, and the decay-constrained physical model. To smooth the loss, the L1 norm of the gradient of the enhanced image is calculated. , , These are the weighting coefficients for reconstruction loss, physical constraint loss, and smoothing loss, respectively.

4. The method for enhancing sound source localization resolution according to claim 1, characterized in that, The method for generating the initial image is as follows: the spatial spectrum of the sound source is estimated using a traditional beamforming algorithm to generate the initial image.

5. A sound source localization resolution enhancement device, characterized in that, For implementing the method as described in any one of claims 1 to 4, the apparatus comprises: Data acquisition and initial image generation module: acquires partial discharge acoustic signals from substation equipment and generates initial images; Enhanced image generation module: Constructs a convolutional neural network with physical constraints, takes the initial image as input, and outputs an enhanced image by learning high-dimensional features. The joint loss function of the convolutional neural network includes reconstruction loss, physical constraint loss and smoothing loss. The physical constraint loss is determined based on a delay constraint physical model and a decay constraint physical model. Positioning module: Based on the enhanced image recognition of the sound source location, it realizes the accurate positioning of the partial discharge sound source in the substation.

6. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the program, it implements the method as described in any one of claims 1 to 4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1 to 4.

Images