Distribution network grounding grid corrosion detection method and device based on weak echo and fragment disassembly parallel acceleration

Abstract

The invention relates to the technical field of power grid technology, and discloses a distribution network grounding grid corrosion detection method and device based on weak echo and fragment disassembly parallel acceleration, and the method comprises the steps: transmitting a square wave pulse excitation signal to a to-be-detected grounding grid; weak echo signals generated by the grounding grid are collected and processed; the processing of the weak echo signal generated by the grounding grid comprises the steps of obtaining the weak echo signal through a sound sensor, amplifying the weak echo signal through a pulse receiving generator, and carrying out digital processing on the weak echo signal through an analog-to-digital converter; constructing a database, and storing a mapping relationship between the weak echo signals of the grounding grid in different corrosion states and the corrosion depth; and on the basis of a fragment disassembly parallel acceleration time sequence classification network, the weak echo signal is segmented into sub-sequence fragments, spatial-temporal features are extracted and classified as corrosion states, and a detection result is obtained. According to the invention, the corrosion state of the grounding grid is diagnosed with high efficiency and high accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of electric power technology, and in particular to a method and device for detecting corrosion of a distribution network grounding grid based on weak echo and fragment disassembly and parallel acceleration. Background Art

[0002] In the daily operation of distribution networks, the stability and safety of power supply in residential areas, commercial buildings, industrial parks, and other areas are highly dependent on the reliable operation of the grounding grid. With the development of the power industry, the coverage of distribution networks continues to expand, and the scale and capacity of systems continue to rise. Maintaining their safe and stable operation has become a core challenge in ensuring power supply. As the "safety guard" of the distribution network, the grounding grid closely connects electrical equipment to the earth and is a critical facility for protecting the safety of distribution network equipment and the personal safety of personnel.

[0003] When the distribution network is struck by lightning or equipment failure causes a short circuit accident, the grounding network responds quickly, opening up a fast discharge channel for the large current generated instantly, avoiding abnormal increase in the working potential of the power system. At the same time, it controls the transfer current and step voltage within the safety threshold, effectively playing the role of grounding protection and ensuring the continuity and reliability of the distribution network power supply.

[0004] However, current domestic distribution network grounding grids are mostly made of galvanized steel. In highly corrosive soil environments, such as those found in older urban villages and chemical industrial parks, the corrosion rate of the grounding grid conductors can be alarming, reaching up to 8mm per year. Although galvanized steel is commonly used in grounding systems designed for a ten-year lifespan, in highly corrosive environments like chemical industrial parks, grounding grids can experience seriously excessive grounding resistance within just five years of operation. This can lead to improper current discharge in the distribution network, significantly increasing the risk of safety incidents and potentially causing serious consequences such as regional power outages and equipment damage. Therefore, improvements to existing technologies are needed.

[0005] The above information is presented as background information only to assist with an understanding of the present disclosure and is not a determination or admission that any of the above may be applicable as prior art with respect to the present disclosure. Summary of the Invention

[0006] The present invention provides a distribution network grounding grid corrosion detection method and device based on weak echo and fragment disassembly parallel acceleration to solve the problems existing in the prior art.

[0007] To achieve the above objectives, the present invention provides the following technical solutions:

[0008] A distribution network grounding grid corrosion detection method based on weak echo and fragment disassembly parallel acceleration includes the following steps:

[0009] Send square wave pulse excitation signal to the grounding grid to be tested;

[0010] Collecting and processing weak echo signals generated by the grounding grid; processing the weak echo signals generated by the grounding grid includes acquiring the weak echo signals through an acoustic sensor, amplifying the weak echo signals with a pulse receiving generator, and digitizing the weak echo signals with an analog-to-digital converter;

[0011] Build a database to store the mapping relationship between weak echo signals of grounding grids in different corrosion states and corrosion depth;

[0012] Based on the fragment decomposition and parallel acceleration of the time series classification network, the weak echo signal is segmented into subsequence segments, the spatiotemporal features are extracted and classified into corrosion states to obtain the detection results;

[0013] Transmit the test results to the terminal device.

[0014] In this embodiment, the sending of the square wave pulse excitation signal includes:

[0015] Rectify and filter AC power into stable DC power, and charge it through energy storage capacitors;

[0016] The trigger square wave excitation source generates a square wave pulse signal in the range of 1μs-10μs, and controls the discharge through internal or external triggering.

[0017] In this embodiment, the time series classification network based on fragment decomposition and parallel acceleration includes:

[0018] Generate subsequence segments and segment the original weak echo signal with the square wave pulse frequency as the period;

[0019] Extract the local spatial features and global temporal dependencies of subsequences through hyperconvolutional neural networks;

[0020] The fully connected layer and softmax function are used to map the features into corrosion state classification probabilities.

[0021] In this embodiment, the segmentation of the original weak echo signal using the square wave pulse frequency as a period includes:

[0022] The two-dimensional echo signal is divided into t equal-length subsequences according to the time sequence, and each subsequence shares the same corrosion state label.

[0023] In this embodiment, the operation of the hyperconvolutional neural network includes:

[0024] Dynamically filter hidden state information and retain or discard information through the sigmoid function;

[0025] Combine the tanh function to generate candidate states and update the data group state;

[0026] Information transmission is controlled by the forget gate and input gate to achieve long-term dependency modeling.

[0027] In this embodiment, the construction of the database includes:

[0028] By simulating grounding grids with different corrosion depths in the laboratory, a positive correlation mapping relationship between the sound pressure amplitude and the corrosion depth is established.

[0029] The present invention also provides a corrosion detection device for a distribution network grounding grid based on parallel acceleration of weak echo and segment disassembling, which is used to implement the corrosion detection of the distribution network grounding grid based on parallel acceleration of weak echo and segment disassembling as described in any one of the above, including:

[0030] A pulse power supply module for sending a square wave pulse excitation signal to the grounding grid to be detected;

[0031] A receiving module, including a sound sensor, a pulse receiving generator and an analog-to-digital converter connected in parallel, for collecting and processing the weak echo signal generated by the grounding grid;

[0032] A detection model, including:

[0033] A database construction unit for storing the mapping relationship between the weak echo signal of the grounding grid in different corrosion states and the corrosion depth;

[0034] A time series classification network for performing time series segmentation and feature extraction on the weak echo signal based on the segment disassembling parallel acceleration architecture, and outputting the classification result of the grounding grid corrosion state;

[0035] A data transmission module for sending the detection result to the terminal device.

[0036] In this embodiment, the pulse power supply module includes:

[0037] A rectifying and filtering unit for converting alternating current into stable direct current and charging the energy storage capacitor;

[0038] A square wave excitation source configured to control the discharge of the energy storage capacitor through internal or external triggering to generate a square wave pulse signal in the range of 1 μs - 10 μs;

[0039] The frequency band of the sound sensor is 1 Hz - 150 kHz, the sensitivity is not less than 40 dB, and the voltage gain of the pulse receiving generator is 40 dB, and the noise is lower than -100 dBm.

[0040] In this embodiment, the time series classification network includes:

[0041] A generator for dividing the original weak echo signal into subsequence segments based on the square wave pulse frequency; the division method of the generator is: taking the square wave pulse period as the reference, dividing the original two-dimensional echo signal into t equal-length subsequences, and each subsequence shares the same corrosion state label;

[0042] The backbone network uses a super convolutional neural network to extract the local spatial features and global temporal dependencies of subsequences through layer-by-layer convolution;

[0043] The discriminator uses a fully connected layer and the softmax function to map the features to the corrosion state classification probability;

[0044] The backbone network includes:

[0045] A selection module that dynamically filters the hidden state information to be retained through the sigmoid function;

[0046] A transfer module that generates candidate states in combination with the tanh function and updates the data group state by weighting the output of the selection module;

[0047] A connected channel that controls information transmission through forget gates and input gates to achieve long-term dependence modeling;

[0048] In this embodiment, the data transmission module is a Bluetooth module for real-time transmitting the corrosion state classification result to the terminal device;

[0049] The database construction unit establishes a positive correlation mapping relationship between the sound pressure amplitude and the corrosion depth by simulating grounding grids with different corrosion depths in the laboratory;

[0050] Compared with the prior art, the present invention has the following beneficial effects:

[0051] A method and device for detecting the corrosion of a distribution network grounding grid based on parallel acceleration of weak echo and segment disassembly provided by the present invention use a pulse power supply to inject a pulse current into the grounding grid through a grounding lead to excite a weak echo signal, and analyze and determine the corrosion degree of the grounding grid based on this, realizing the corrosion state diagnosis of the grounding grid efficiently and with high accuracy.

[0052] The present invention has other characteristics and advantages that will be obvious from the accompanying drawings incorporated herein and the subsequent detailed description, or will be described in detail in the accompanying drawings incorporated herein and the subsequent detailed description, which are used together to explain the specific principles of the present invention. Description of the Drawings

[0053] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0054] Figure 1It is a flowchart of a method for detecting corrosion of a distribution network grounding grid based on parallel acceleration of weak echo and segment disassembly provided by an embodiment of the present invention;

[0055] Figure 2 It is a schematic framework diagram of a device for detecting corrosion of a distribution network grounding grid based on parallel acceleration of weak echo and segment disassembly provided by an embodiment of the present invention;

[0056] Figure 3 It is a relationship diagram between the amplitude of the sound pressure signal and the corrosion depth in a method for detecting corrosion of a distribution network grounding grid based on parallel acceleration of weak echo and segment disassembly provided by an embodiment of the present invention;

[0057] Figure 4 It is another relationship diagram between the amplitude of the sound pressure signal and the corrosion depth in a method for detecting corrosion of a distribution network grounding grid based on parallel acceleration of weak echo and segment disassembly provided by an embodiment of the present invention;

[0058] Figure 5 It is a schematic diagram of the backbone network in a device for detecting corrosion of a distribution network grounding grid based on parallel acceleration of weak echo and segment disassembly provided by an embodiment of the present invention. Detailed implementation manners

[0059] To illustrate in detail the possible application scenarios, technical principles, specific implementable solutions, achievable purposes and effects of the present application, the following will be described in detail with reference to the specific embodiments listed and in conjunction with the accompanying drawings. The embodiments described herein are only used to more clearly illustrate the technical solutions of the present application, so they are only examples and cannot be used to limit the protection scope of the present application.

[0060] Referring to "embodiment" in this article means that the specific features, structures or characteristics described in connection with the embodiment may be included in at least one embodiment of the present application. The term "embodiment" appearing in various positions in the specification does not necessarily refer to the same embodiment, nor is it particularly limited to the independence or relevance with other embodiments. In principle, in the present application, as long as there is no technical contradiction or conflict, the technical features mentioned in each embodiment can be combined in any way to form the corresponding implementable technical solution.

[0061] Unless otherwise defined, the meanings of the technical terms used in this article are the same as those generally understood by those skilled in the technical field to which the present application belongs; the use of the relevant terms in this article is only for describing specific embodiments and is not intended to limit the present application.

[0062] In the description of the present application, the term "and / or" is an expression used to describe the logical relationship between objects, indicating that there can be three relationships. For example, A and / or B means: there is A, there is B, and there is both A and B at the same time. In addition, the character " / " in this text generally represents an "or" logical relationship between the associated objects before and after.

[0063] In the present application, terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual quantitative, primary-secondary, or sequential relationships between these entities or operations.

[0064] Without further limitations, in the present application, the expressions "comprising", "including", "having", or other similar expressions used in a statement are intended to cover non-exclusive inclusion. These expressions do not exclude the possibility that there may be additional elements in the process, method, or product including the stated elements. Thus, a process, method, or product including a series of elements may include not only those defined elements, but also other elements not explicitly listed, or elements inherent to such process, method, or product.

[0065] Similar to the understanding in the "Examination Guidelines", in the present application, expressions such as "greater than", "less than", "exceeding", etc. are understood not to include the number itself; expressions such as "above", "below", "within", etc. are understood to include the number itself. In addition, in the description of the embodiments of the present application, the meaning of "a plurality of" is two or more (including two). Similar expressions related to "many", such as "multiple groups", "multiple times", etc., are understood in the same way, unless otherwise specifically defined.

[0066] In the description of the embodiments of the present application, the spatially related expressions used, such as "center", "longitudinal", "transverse", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "perpendicular", "top", "bottom", "inner", "outer", "clockwise", "counterclockwise", "axial", "radial", "circumferential", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the specific embodiment or the drawing. It is only for the convenience of describing the specific embodiments of the present application or for the reader's understanding, and does not indicate or imply that the device or component referred to must have a specific position, a specific orientation, or be constructed or operated in a specific orientation. Therefore, it should not be construed as a limitation on the embodiments of the present application.

[0067] Unless otherwise clearly specified or defined, in the description of the embodiments of the present application, terms such as "installation", "connection", "linkage", "fixation", and "setting" should be understood in a broad sense. For example, the "connection" may be a fixed connection, a detachable connection, or an integral setting; it may be a mechanical connection, an electrical connection, or a communication connection; it may be a direct connection or an indirect connection through an intermediate medium; it may be the communication inside two components or the interaction relationship between two components. For those skilled in the art to which the present application belongs, the specific meanings of the above terms in the embodiments of the present application can be understood according to specific circumstances.

[0068] Please refer to Figure 1 , an embodiment of the present invention provides a corrosion detection method for a distribution network grounding grid based on parallel acceleration of weak echo and fragment disassembly, including the following steps:

[0069] S1. Send a square wave pulse excitation signal to the grounding grid to be detected.

[0070] Specifically, the operation steps are as follows:

[0071] First, connect the exposed part of the grounding grid to be detected through a positioning detection device, and send a square wave pulse excitation source with a certain frequency to the grounding grid to be detected.

[0072] Then, connect the positioning detection device from the other end of the part to be grounded, and read the weak echo signal of the grounding grid.

[0073] In this embodiment, sending the square wave pulse excitation signal includes:

[0074] Rectify and filter the alternating current into stable direct current, and charge through an energy storage capacitor;

[0075] Trigger the square wave excitation source to generate a square wave pulse signal in the range of 1 μs - 10 μs, and control the discharge through an internal or external trigger method.

[0076] Specifically, rectify and filter the 380V alternating current into stable direct current, and adjust the charging voltage of the energy storage capacitor through a charging resistor;

[0077] Trigger the square wave excitation source to generate a square wave pulse signal in the range of x μs - 10 μs, and control the discharge of the energy storage capacitor through an internal or external trigger method.

[0078] S2. Collect and process the weak echo signal generated by the grounding grid.

[0079] Among them, processing the weak echo signal generated by the grounding grid includes obtaining the weak echo signal through an acoustic sensor, amplifying the weak echo signal by a pulse receiving generator, and digitizing the weak echo signal by an analog-to-digital converter.

[0080] It should be noted that there seems to be an error in the original text where "1μs - 10μs" is repeated as "x μs - 10μs" in step 32. I have translated it as presented in the original text. If this is a mistake, please correct the original text for a more accurate translation.Among them, the processing steps include: obtaining a weak echo signal through an acoustic sensor with a frequency band of 1 Hz - 150 kHz, and its sensitivity is not less than 40 dB;

[0081] Amplify the signal by a pulse receiving generator with a 40 dB voltage gain (noise is lower than -100 dBm);

[0082] Convert the analog signal into a digital signal through an analog-to-digital converter.

[0083] S3. Build a database to store the mapping relationship between the weak echo signals of the grounding grid in different corrosion states and the corrosion depth.

[0084] In this embodiment, building a database includes:

[0085] Simulate grounding grids with different corrosion depths in the laboratory to establish a positive correlation mapping relationship between the sound pressure amplitude and the corrosion depth, as shown in the figure.

[0086] S4. Based on the temporal classification network with parallel acceleration by segment disassembling, segment the weak echo signal into subsequence segments, extract spatio-temporal features and classify them into corrosion states to obtain the detection result.

[0087] In this embodiment, the temporal classification network includes:

[0088] 1) Generate subsequence segments: Taking the square wave pulse frequency as the period, divide the original two-dimensional echo signal into t equal-length subsequences according to the time relationship, and each subsequence shares the same corrosion state label;

[0089] 2) Extract spatio-temporal features through a hyper-convolutional neural network:

[0090] Dynamically screen hidden state information: Use the sigmoid function to retain or discard information;

[0091] Combine the tanh function to generate candidate states and update the data group state;

[0092] Control the information transmission through the forget gate and the input gate to achieve long-term dependence modeling, as Figure 3 shown;

[0093] 1) Use the fully connected layer and the softmax function to map the features to the corrosion state classification probability.

[0094] In this embodiment, the temporal classification network with parallel acceleration by segment disassembling includes:

[0095] Generate subsequence segments and divide the original weak echo signal with the square wave pulse frequency as the period;

[0096] Extract the local spatial features and global temporal dependence relationship of the subsequences through a hyper-convolutional neural network;

[0097] The feature is mapped to the corrosion state classification probability by using a fully connected layer and a softmax function.

[0098] In this embodiment, the original weak echo signal is segmented by taking the square wave pulse frequency as the period, including:

[0099] The two-dimensional echo signal is segmented into t equal-length subsequences according to the time sequence, and each subsequence shares the same corrosion state label.

[0100] In this embodiment, the operation of the super convolutional neural network includes:

[0101] Dynamically screen the hidden state information, and retain or discard the information through the sigmoid function;

[0102] Generate a candidate state in combination with the tanh function to update the data group state;

[0103] Control the information transmission through the forget gate and the input gate to achieve long-term dependence modeling.

[0104] S5. Transmit the detection result to the terminal device.

[0105] In this embodiment, the detection result is transmitted to the staff terminal in real time through the Bluetooth module.

[0106] A method for detecting the corrosion of a distribution network grounding grid based on parallel acceleration of weak echo and segment disassembly provided by this embodiment uses a pulse power supply to inject a pulse current into the grounding grid through a grounding lead. The current generates Lorentz force and Joule heat effect through the conductor. These two forces drive the surrounding medium to generate displacement, cause particle vibration, and excite weak echo signals. At the same time, a database is constructed by experimentally simulating the corresponding relationship between different corrosion states and acoustic signals, and the corrosion state diagnosis of the grounding grid is realized efficiently and with high accuracy through a parallel architecture neural network.

[0107] Please refer to Figure 2 , based on the foregoing embodiments, the present invention further provides a device for detecting the corrosion of a distribution network grounding grid based on parallel acceleration of weak echo and segment disassembly, which is used to implement the detection of the corrosion of a distribution network grounding grid based on parallel acceleration of weak echo and segment disassembly as described in any one of the above, and includes a pulse power supply module, a receiving module, and a detection model.

[0108] The pulse power supply module is used to send a square wave pulse excitation signal to the grounding grid to be detected.

[0109] Specifically, the pulse power supply module includes:

[0110] A rectifying and filtering unit: converts 380V alternating current into stable direct current to charge the energy storage capacitor;

[0111] Square wave excitation source: Controls the discharge through internal or external triggering to generate a square wave pulse signal of 1 μs - 10 μs.

[0112] Receiving module, including a sound sensor, a pulse receiving generator, and an analog-to-digital converter connected in parallel, for collecting and processing the weak echo signals generated by the grounding grid.

[0113] Specifically, the receiving module includes:

[0114] Sound sensor: The frequency band of the sound sensor is 1 Hz - 150 kHz, the sensitivity is not less than 40 dB, and the voltage gain of the pulse receiving generator is 40 dB, with the noise lower than -100 dBm.

[0115] Pulse receiving generator: 40 dB voltage gain, with the noise lower than -100 dBm;

[0116] Analog-to-digital converter: Used to convert the analog signal of the grounding grid echo into a digital signal, providing a data basis for computer model detection.

[0117] Among them:

[0118] 1) The functions and reasons for selecting the sound sensor are as follows: Since the method of detecting the grounding body with sound waves requires the frequency band of the sensor to be as wide as possible to adapt to the frequency matching under different burial depths and have a good output response in the low-frequency band. At the same time, a high sensitivity is required to detect tiny vibration signals, reduce the output power of the excitation source, and ensure the safety of the system. Therefore, we use a sound sensor with a bandwidth of 1 Hz - 150 kHz, which has the advantages of small size, high sensitivity, and wide detection frequency band, and can effectively collect the sound signals transmitted from the grounding grid.

[0119] 2) The functions and reasons for selecting the pulse receiving generator are as follows: The pulse receiving generator with a 40 dB voltage gain has low noise, low drift, high common-mode rejection ratio, and strong anti-interference ability, and is especially suitable for amplifying the grounding grid echo signals. Therefore, we use a pulse receiving generator with a 40 dB voltage gain to solve the problem of weak grounding grid echo signals, enabling the subsequent detection model to more accurately detect the state of the grounding grid.

[0120] Detection model, including:

[0121] Database construction unit, used to store the mapping relationship between the weak echo signals of the grounding grid in different corrosion states and the corrosion depth.

[0122] Specifically, the database construction unit simulates grounding grids with different corrosion depths in the laboratory to establish a positive correlation mapping relationship between the sound pressure amplitude and the corrosion depth.

[0123] Place the waveforms of the sound pressure varying with time extracted in the same coordinate for comparison, fromFigure 3 , Figure 4 It can be seen that the amplitude of the sound pressure generated at the corroded part is consistent with the change trend of the corrosion depth. As the corrosion of the galvanized steel grounding body deepens, the amplitude of the generated sound pressure signal increases, that is, the amplitude of the sound pressure signal is positively correlated with the corrosion depth.

[0124] A temporal classification network is used to perform temporal segmentation and feature extraction on weak echo signals based on a fragment disassembling parallel acceleration architecture, and output the classification results of the grounding grid corrosion state.

[0125] Among them, the temporal classification network learns the spatial and temporal dependencies between subsequences and is used for the classification (normal, mild, moderate, and severe) detection of different states of grounding grid corrosion. This network includes three modules: a generator, a backbone network, and a discriminator.

[0126] Specifically, in this embodiment, the temporal classification network includes:

[0127] A generator that divides the original weak echo signal into subsequence segments based on the square wave pulse frequency; solves the problem of low training efficiency of the weak echo signal data of the grounding grid with long time series.

[0128] The segmentation method of the generator is: based on the square wave pulse period, the original two-dimensional echo signal is divided into t equal-length subsequences, and each subsequence shares the same corrosion state label; that is, the time series of the original two-dimensional grounding grid echo signal is divided into separate subsequence segments according to the time sequence relationship.

[0129] Implementation steps: For the time series of the two-dimensional grounding grid echo signal, sub-time series based on the transmitting square wave pulse frequency can be generated (i.e., sub-sequence segments with the transmitting square wave pulse frequency as the period), which means that each segment represents.

[0130] Therefore, we assume that the input of the segment generator is the time series sample X(i), and the output is the segment set B(i). And B(i) = {bi1, bi2,..., bit}, where bit is a single segment, t represents the index, which is the number of segments of X(i). By inheriting X(i), each b(i) ∈ B(i) shares the same label (normal, mild, moderate, and severe) y(i).

[0131] .

[0132] A backbone network that uses a super convolutional neural network to extract the local spatial features and global temporal dependencies of the subsequences through layer-by-layer convolution.

[0133] The backbone network is the super convolutional neural network, and its purpose is to learn the correlation between each variable of the input segment.

[0134] The backbone network aims to model the local spatial dependencies of each segment and the global temporal dependencies between segments to generate valuable hidden representations. By adopting a multi-layer convolutional network to learn the correlations between each variable of the input segments, and then capturing the hidden spatial states, in order to propagate information between different segments, the hidden spatial states generated by the convolutional network can be utilized to generate spatio-temporal hidden states.

[0135] This network, as Figure 5 shown, has three input parameters. C represents the state of the previous data group (if it is the first step, it is the two-dimensional grounding grid echo signal subsequence), H represents the hidden state at the current time step, and X represents the input at the current time step. Among them, * represents the element-wise convolution operation, and Wf, Wi, Wc, Wo are weight matrices, and bf, bi, bc, bo are bias terms.

[0136] Among them, the backbone network includes:

[0137] A selection module that dynamically filters the hidden state information to be retained through the sigmoid function.

[0138] In the network proposed in the present invention, the selection module combines the information of the previous hidden state and the current input, and passes this information to the sigmoid function to determine which information should be discarded from the data group state. The output value of the sigmoid function ranges between 0 and 1. The closer it is to 0, the more the corresponding information should be discarded, and the closer it is to 1, the more it should be retained, thereby dynamically adjusting the information in the data group state to adapt to different task requirements.

[0139] .

[0140] A transfer module that combines the tanh function to generate a candidate state and updates the data group state by weighting the output of the selection module.

[0141] The information of the previous layer hidden state and the information of the current input are jointly input into the sigmoid function. By adjusting the output value between 0 and 1, it is determined which information is important for the update (where 0 represents unimportant information and 1 represents important information). Immediately afterwards, the information of the previous layer hidden state and the information of the current input are also input into the tanh function to generate a new candidate value vector. Finally, the output value of the sigmoid function is multiplied by the output value of the tanh function. This process is actually to determine which key information in the output value of the tanh function needs to be retained by the output value of the sigmoid function, thereby realizing the update of the data group state.

[0142] ;

[0143] 。

[0144] Connected channels control information transmission through forget gates and input gates to achieve long-term dependence modeling.

[0145] During the update process, the connected channels first discard the information that is no longer needed through the product operation with the output of the forget gate; then, through the product operation with the output of the input gate, new important information is added. This ensures the effective transmission of information between the selection module and the transfer module, and at the same time dynamically adjusts the information it carries according to task requirements to achieve effective modeling of long-term dependence relationships.

[0146] 。

[0147] Derivation module: To determine the value of the next hidden state, we first input the previous hidden state and the current input into the sigmoid function together to evaluate the importance of the information, and transfer the updated data group state to the tanh function to generate a candidate hidden state vector. Subsequently, the output of the tanh is multiplied by the output of the sigmoid, and the "gating" effect of the sigmoid is used to determine the information that the hidden state should carry. Finally, the obtained hidden state is used as the output of the current hyper-convolutional neural network module unit and is transferred to the next time step together with the new data group state to continue processing the sequence information and generate the corresponding output sequence.

[0148] ;

[0149] 。

[0150] Discriminator, using fully connected layers and the softmax function to map features to the probability of corrosion state classification.

[0151] A time series is generated by mapping the hidden state Ht (i.e., the sum of the final output sequence matrix vectors of the hyper-convolutional neural network) to a low-dimensional space using a fully connected network, and then the obtained low-dimensional vector is normalized to the probability of the label Y (normal, mild, moderate, and severe) through the softmax function. The equation is as follows:

[0152] 。

[0153] Data transmission module, used to send the detection results to the terminal device.

[0154] In this embodiment, the data transmission module is a Bluetooth module, which is used to transmit the corrosion state classification result to the terminal device in real time.

[0155] Compared with the prior art, the present invention has the following beneficial effects:

[0156] A method and device for detecting corrosion of a distribution network grounding grid based on parallel acceleration of weak echo and segment disassembly provided by the present invention use a pulse power supply to inject a pulse current into the grounding grid through a grounding lead. The current generates Lorentz force and Joule heat effect when passing through the conductor. These two forces drive the surrounding medium to generate displacement, cause particle vibration, and excite weak echo signals. At the same time, a database is constructed by experimentally simulating the corresponding relationship between different corrosion states and acoustic signals, and the corrosion state diagnosis of the grounding grid is realized efficiently and with high accuracy through a parallel architecture neural network.

[0157] Finally, it should be noted that although the above embodiments have been described in the text and drawings of the specification of the present application, the patent protection scope of the present application cannot be limited thereby. Any technical solution obtained by equivalent structure or equivalent process substitution or modification based on the essential concept of the present application and using the content recorded in the text and drawings of the specification of the present application, as well as any technical solution directly or indirectly implementing the above embodiments in other related technical fields, is included in the patent protection scope of the present application.

Claims

1. A corrosion detection method for a distribution network grounding grid based on parallel acceleration of weak echo and fragment disassembly, characterized in that, It includes the following steps: Send a square-wave pulse excitation signal to the grounding grid to be detected; Collect and process the weak echo signal generated by the grounding grid; the processing of the weak echo signal generated by the grounding grid includes obtaining the weak echo signal through a sound sensor, amplifying the weak echo signal by a pulse receiving generator, and digitizing the weak echo signal by an analog-to-digital converter; Construct a database to store the mapping relationship between the weak echo signal of the grounding grid in different corrosion states and the corrosion depth; Based on the temporal classification network with parallel acceleration by segment disassembling, segment the weak echo signal into subsequence segments, extract spatio-temporal features and classify them into corrosion states to obtain the detection result; Transmit the detection result to the terminal device.

2. The method for detecting the corrosion of the distribution network grounding grid based on parallel acceleration of weak echo and fragment disassembly according to claim 1, wherein, The sending of the square-wave pulse excitation signal includes: Rectify and filter the alternating current into stable direct current and charge through an energy storage capacitor; Trigger the square-wave excitation source to generate a square-wave pulse signal in the range of 1 μs - 10 μs, and control the discharge through internal or external triggering.

3. The corrosion detection method for the distribution network grounding grid based on parallel acceleration of weak echo and fragment disassembly according to claim 1, wherein The temporal classification network with parallel acceleration by segment disassembling includes: Generate subsequence segments, and segment the original weak echo signal with the square-wave pulse frequency as the period; Extract the local spatial features and global temporal dependencies of the subsequences through a super convolutional neural network; Use the fully connected layer and the softmax function to map the features to the corrosion state classification probability.

4. The method for detecting corrosion of a distribution network grounding grid based on parallel acceleration of weak echo and segment disassembly according to claim 3, characterized in that, The segmentation of the original weak echo signal with the square-wave pulse frequency as the period includes: Segment the two-dimensional echo signal into t equal-length subsequences according to the time sequence, and each subsequence shares the same corrosion state label.

5. The method for detecting the corrosion of the distribution network grounding grid based on parallel acceleration of weak echo and segment disassembly according to claim 4, wherein, The operation of the super convolutional neural network includes: Dynamically screen the hidden state information, and retain or discard the information through the sigmoid function; Combine the tanh function to generate candidate states and update the data group state; Control the information transmission through the forget gate and the input gate to achieve long-term dependence modeling.

6. The corrosion detection method for the distribution network grounding grid based on parallel acceleration of weak echo and segment disassembly according to claim 1, characterized in that The construction of the database includes: Simulate the grounding grid with different corrosion depths in the laboratory to establish a positive correlation mapping relationship between the sound pressure amplitude and the corrosion depth.

7. A corrosion detection device for a distribution network grounding grid based on parallel acceleration of weak echo and segment disassembly, characterized in that, For implementing the corrosion detection of the distribution network grounding grid based on weak echo and parallel acceleration by segment disassembling as described in any one of claims 1 to 6, it includes: A pulse power supply module for sending a square-wave pulse excitation signal to the grounding grid to be detected; A receiving module, including a sound sensor, a pulse receiving generator and an analog-to-digital converter connected in parallel, for collecting and processing the weak echo signal generated by the grounding grid; A detection model, including: A database construction unit for storing the mapping relationship between the weak echo signal of the grounding grid in different corrosion states and the corrosion depth; A temporal classification network for performing temporal segmentation and feature extraction on the weak echo signal based on the parallel acceleration architecture by segment disassembling, and outputting the classification result of the corrosion state of the grounding grid; A data transmission module for sending the detection result to the terminal device.

8. The corrosion detection device for the distribution network grounding grid based on parallel acceleration of weak echo and segment disassembly according to claim 7, wherein, The pulse power supply module includes: A rectification and filtering unit for converting the alternating current into stable direct current and charging the energy storage capacitor; A square-wave excitation source configured to control the discharge of the energy storage capacitor through internal or external triggering to generate a square-wave pulse signal in the range of 1 μs - 10 μs; The frequency band of the acoustic sensor is 1 Hz - 150 kHz, the sensitivity is not less than 40 dB, and the voltage gain of the pulse receiving generator is 40 dB, with the noise lower than -100 dBm.

9. The corrosion detection device for distribution network grounding grid based on parallel acceleration of weak echo and segment disassembly according to claim 7, characterized in that, The timing classification network includes: A generator that divides the original weak echo signal into subsequence segments based on the square wave pulse frequency; the division method of the generator is: based on the square wave pulse period, the original two-dimensional echo signal is divided into t equal-length subsequences, and each subsequence shares the same corrosion state label; A backbone network that uses a super convolutional neural network to extract the local spatial features and global timing dependencies of the subsequences through layer-by-layer convolution; A discriminator that maps the features to the corrosion state classification probability using a fully connected layer and a softmax function; The backbone network includes: A selection module that dynamically filters the hidden state information to be retained through a sigmoid function; A transfer module that generates candidate states in combination with a tanh function and updates the data group state by weighting the output of the selection module; A connected channel that controls information transmission through a forget gate and an input gate to achieve long-term dependence modeling.

10. The corrosion detection device for the distribution network grounding grid based on parallel acceleration of weak echo and fragment disassembly according to claim 7, characterized in that, The data transmission module is a Bluetooth module for real-time transmitting the corrosion state classification result to a terminal device; The database construction unit establishes a positive correlation mapping relationship between the sound pressure amplitude and the corrosion depth by simulating grounding grids with different corrosion depths in the laboratory.

Images