On-load tap-changer vibration detection system and method based on wireless sensing and photonic crystal resonant cavity

Abstract

The invention discloses an on-load tap-changer vibration detection system and method based on wireless sensing and a photonic crystal resonant cavity, and relates to the technical field of intelligent power grid equipment state monitoring, and the method comprises the steps: deploying wireless monitoring nodes to collect vibration signals, constructing an adjacent matrix according to a mechanical topology connection relation, and carrying out the wireless monitoring of the vibration signals; calculating the edge weight of the adjacent matrix through the physical distance and the mechanical coupling strength between the wireless monitoring nodes; performing optical-vibration coupling enhancement on the vibration signal through the photonic crystal resonant cavity to generate an amplified signal, and compressing the amplified signal based on a quantum compressed sensing algorithm to generate a compressed signal; and performing orthogonal matching pursuit sparse reconstruction on the compressed signal, extracting a time-frequency joint feature, and inputting the time-frequency joint feature into a local convolutional neural network to generate a local convolutional neural network parameter. A mechanical vibration signal is converted into a high-sensitivity modulation optical signal through the optical-vibration coupling effect of the photonic crystal resonant cavity, and the problem that a traditional sensor is low in characteristic signal-to-noise ratio in a strong noise environment is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of smart grid equipment status monitoring, and in particular to a system and method for detecting vibration of an on-load tap changer based on wireless sensing and a photonic crystal resonant cavity. Background Art

[0002] With the increasing demand for intelligent power systems, online monitoring technology for on-load tap changers (OLTCs) is gradually evolving towards multimodal sensing and intelligent diagnosis. In recent years, wireless sensor networks (WSNs), due to their flexible deployment and robustness against electromagnetic interference, have been widely used in mechanical vibration signal acquisition. Their topology modeling capabilities offer a new path for multi-node collaborative analysis.

[0003] Traditional vibration signal enhancement mechanisms overly rely on improving the signal-to-noise ratio of a single sensor node and fail to effectively integrate mechanical topology correlations and multi-physics field coupling effects. This results in weak fault signatures being easily overwhelmed by environmental noise, limiting the effective extraction of high-frequency harmonic components. Furthermore, existing distributed diagnostic systems generally employ a fixed weight strategy during federated aggregation, failing to consider the combined impact of mechanical connection density and signal transmission quality on model convergence. This makes it difficult to achieve coordinated optimization of local features and global topology. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides an on-load tap changer vibration detection method based on wireless sensing and photonic crystal resonant cavity to solve the problems of low signal-to-noise ratio of weak fault characteristics and poor model convergence.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: In a first aspect, the present invention provides a vibration detection method for an on-load tap changer based on wireless sensing and a photonic crystal resonant cavity, which includes deploying wireless monitoring nodes to collect vibration signals, constructing an adjacency matrix according to a mechanical topological connection relationship, and calculating the edge weights of the adjacency matrix through the physical distance and mechanical coupling strength between the wireless monitoring nodes; performing optical-vibration coupling enhancement on the vibration signal through the photonic crystal resonant cavity to generate an amplified signal, and compressing the amplified signal based on a quantum compressed sensing algorithm to generate a compressed signal; performing orthogonal matching pursuit sparse reconstruction on the compressed signal, extracting time-frequency joint features, inputting the time-frequency joint features into a local convolutional neural network, and generating local convolutional neural network parameters; calculating the federation aggregation weight based on the edge weights and signal-to-noise ratio of the adjacency matrix, fusing the local convolutional neural network parameters, and constructing a global diagnostic model based on a distributed aggregation protocol of a federated learning framework; calculating the Q-value attenuation coefficient of the photonic crystal resonant cavity, inputting the mechanical coupling strength, the time-frequency joint features, and the Q-value attenuation coefficient into the global diagnostic model to generate fault classification results and maintenance decisions.

[0007] As a preferred solution of the on-load tap changer vibration detection method based on wireless sensing and photonic crystal resonant cavity of the present invention, wherein: the edge weights of the adjacency matrix are calculated according to the physical distance and mechanical coupling strength between wireless monitoring nodes, and the specific steps are as follows: Determine the mechanical topology of the on-load tap-changer, identify the physical connection of core components, mark the deployment location of wireless monitoring nodes, and collect vibration signals in real time; An adjacency matrix is ​​constructed based on the mechanical topological connection relationship, the physical distance between wireless monitoring nodes is measured, and the mechanical coupling strength between each wireless monitoring node is calibrated according to the mechanical connection type between the components; Calculate the edge weights of the adjacency matrix based on the physical distance and mechanical coupling strength.

[0008] As a preferred solution of the on-load tap changer vibration detection method based on wireless sensing and photonic crystal resonant cavity of the present invention, the specific steps of generating the compressed signal are as follows: Define the target frequency band of mechanical vibration according to the photonic crystal resonant cavity; The photonic crystal resonator is integrated with the wireless monitoring node. The photonic crystal resonator is excited by laser to generate the optical-oscillator coupling effect to convert the mechanical vibration into a modulated optical signal, which is then amplified by the bandgap resonance method. The amplified modulated optical signal is converted into an electrical signal through a photodetector, and the real-time Q value of the photonic crystal resonant cavity is recorded; A quantum random projection matrix is ​​constructed to perform linear projection on the electrical signal, compressing the vibration signal into a low-dimensional space to generate a compressed signal.

[0009] As a preferred solution of the on-load tap changer vibration detection method based on wireless sensing and photonic crystal resonant cavity of the present invention, the specific steps of generating local convolutional neural network parameters are as follows: Initialize the orthogonal matching pursuit algorithm, build a sparse basis dictionary based on the Fourier transform basis and the wavelet basis, and reconstruct the vibration signal based on the quantum random projection matrix and the sparse basis dictionary; The reconstructed vibration signal is decomposed by wavelet packets to generate a time-frequency energy distribution matrix. The instantaneous frequency gradient, energy center offset, and energy proportion of each sub-band of the time-frequency energy distribution matrix are calculated to generate a time-frequency joint feature. Construct a local convolutional neural network, input the time-frequency joint features into the local convolutional neural network, and generate the convolution kernel weights and fully connected layer parameters.

[0010] As a preferred solution of the on-load tap changer vibration detection method based on wireless sensing and photonic crystal resonant cavity of the present invention, wherein: the federation aggregation weight is calculated based on the edge weight and signal-to-noise ratio of each wireless monitoring node in the adjacency matrix, the specific steps are as follows: Extract edge weights between wireless monitoring nodes from the adjacency matrix; Combined with the signal-to-noise ratio of each wireless monitoring node, the federated aggregation weight of each wireless monitoring node is calculated through the weight distribution rule defined by the federated learning framework.

[0011] As a preferred solution of the on-load tap changer vibration detection method based on wireless sensing and photonic crystal resonant cavity described in the present invention, wherein: the distributed aggregation protocol based on the federated learning framework constructs a global diagnostic model, and the specific steps are as follows: The local convolutional neural network parameters uploaded by each wireless monitoring node are weighted averaged according to the federated aggregation weight to generate the global diagnostic model parameters; The federated learning framework loads the global diagnostic model parameters based on the structure of the local convolutional neural network to build a global diagnostic model.

[0012] As a preferred solution of the on-load tap changer vibration detection method based on wireless sensing and photonic crystal resonant cavity of the present invention, the specific steps of generating fault classification results and maintenance decisions are as follows: The initial Q value is defined according to the frequency response curve of the photonic crystal resonator, the Q value attenuation coefficient is calculated according to the difference between the real-time Q value and the initial Q value, and the Q value attenuation coefficient threshold is defined through experimental calibration method; The mechanical coupling strength, time-frequency joint characteristics and Q-value attenuation coefficient are input into the global diagnosis model to generate fault classification results; The fault level is determined based on the fault classification results and the Q-value attenuation coefficient threshold, and a maintenance strategy is generated.

[0013] In a second aspect, the present invention provides an on-load tap changer vibration detection system based on wireless sensing and a photonic crystal resonant cavity, comprising a node deployment module, a signal compression module, a feature extraction module, a model construction module, and a maintenance decision module; the node deployment module is used to deploy wireless monitoring nodes to collect vibration signals, construct an adjacency matrix based on the mechanical topological connection relationship, and calculate the edge weights of the adjacency matrix based on the physical distance and mechanical coupling strength between the wireless monitoring nodes; the signal compression module is used to enhance the optical-vibration coupling of the vibration signal through the photonic crystal resonant cavity to generate an amplified signal, and compress the amplified signal based on the quantum compressed sensing algorithm. Generate a compressed signal; a feature extraction module is used to perform orthogonal matching pursuit sparse reconstruction on the compressed signal, extract time-frequency joint features, input the time-frequency joint features into the local convolutional neural network, and generate local convolutional neural network parameters; a model construction module is used to calculate the federated aggregation weight based on the edge weights and signal-to-noise ratio of the adjacency matrix, fuse the local convolutional neural network parameters, and build a global diagnostic model based on the distributed aggregation protocol of the federated learning framework; a maintenance decision module is used to calculate the Q-value attenuation coefficient of the photonic crystal resonant cavity, input the mechanical coupling strength, time-frequency joint features and Q-value attenuation coefficient into the global diagnostic model, and generate fault classification results and maintenance decisions.

[0014] In a third aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the on-load tap changer vibration detection method based on wireless sensing and a photonic crystal resonant cavity as described in the first aspect of the present invention is implemented.

[0015] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the on-load tap changer vibration detection method based on wireless sensing and a photonic crystal resonant cavity as described in the first aspect of the present invention is implemented.

[0016] The beneficial effects of the present invention are as follows: the mechanical vibration signal is converted into a high-sensitivity modulated light signal through the light-vibration coupling effect of the photonic crystal resonant cavity, which solves the problems of low signal-to-noise ratio of weak fault characteristics and poor data transmission efficiency at high sampling rates of traditional sensors in a strong noise environment; at the same time, the federation aggregation weight is dynamically calculated by physical distance and mechanical coupling strength, which optimizes the parameter fusion process of the global diagnostic model and solves the problems of slow convergence speed and poor global consistency of the global diagnostic model. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of 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 paying any creative work.

[0018] Figure 1 This is a flow chart of the on-load tap changer vibration detection method based on wireless sensing and photonic crystal resonant cavity.

[0019] Figure 2 Schematic diagram of the on-load tap changer vibration detection system based on wireless sensing and photonic crystal resonant cavity.

[0020] Figure 3 Flowchart for generating local convolutional neural network parameters.

[0021] Figure 4 A flowchart for generating a maintenance strategy.

[0022] Figure 5 This is the working flow diagram of the photonic crystal resonator. DETAILED DESCRIPTION

[0023] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0024] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. 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.

[0025] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0026] Reference Figures 1 to 5 , is an embodiment of the present invention, which provides a vibration detection method for an on-load tap changer based on wireless sensing and a photonic crystal resonant cavity, comprising the following steps: S1. Deploy wireless monitoring nodes to collect vibration signals, construct an adjacency matrix based on the mechanical topology connection relationship, and calculate the edge weights of the adjacency matrix based on the physical distance and mechanical coupling strength between wireless monitoring nodes.

[0027] S1.1. Determine the mechanical topology of the on-load tap-changer, identify the physical connection of core components, mark the deployment location of wireless monitoring nodes, and collect vibration signals in real time. It should be noted that based on the technical drawings and three-dimensional structure of the on-load tap changer, the physical connection paths of the gearbox, transmission shaft and contact group are analyzed to generate a mechanical topology connection relationship diagram; according to the mechanical topology connection relationship diagram, the bolt connection, welding or snap-on assembly method of the core components such as the gear meshing point, contact switching mechanism and spring energy storage device is identified, and a physical connection method classification list containing connection type and position coordinates is constructed; based on the physical connection method classification list, wireless monitoring nodes are deployed at the core components such as the point with maximum radial vibration on the gearbox surface, the contact group action impact point and the transmission shaft bearing seat, and the core components are identified according to the mechanical topology connection relationship diagram. The physical connection order of the core components (such as gearbox → transmission shaft → spring mechanism → contact group) is determined, and a node number is assigned to each wireless monitoring node; the three-dimensional coordinates and node numbers of the wireless monitoring nodes are marked with a laser locator and recorded in the mechanical topology connection diagram; a photonic crystal vibration sensor group is installed on the flange surface of the transformer top cover, and the 1-5kHz vibration signal is enhanced by the built-in photonic crystal resonant cavity; the three-axis acceleration sensor of the wireless monitoring node is configured to synchronously collect vibration signals at a sampling rate of 50kHz. After compression and downsampling, the vibration signals and the coordinates of the wireless monitoring node are uploaded to the edge computing unit in real time through the LoRa protocol.

[0028] S1.2. Construct an adjacency matrix based on the mechanical topological connection relationship, measure the physical distance between wireless monitoring nodes, and calibrate the mechanical coupling strength between each wireless monitoring node according to the mechanical connection type between components; It should be noted that based on the mechanical topological connection diagram of the on-load tap changer (the physical connection path of the gear-drive shaft-contact), an N×N adjacency matrix (N is the total number of wireless monitoring nodes) is constructed, and the row and column indexes of the adjacency matrix correspond to the node numbers, and the initial value is zero; according to the mechanical topological connection diagram, the adjacency matrix positions corresponding to the wireless monitoring nodes with direct mechanical connections are marked as to be filled, and the remaining unconnected positions remain zero; the physical distance between adjacent wireless monitoring nodes is calculated using the Euclidean distance formula through the three-dimensional coordinates of the wireless monitoring nodes and a physical distance matrix is ​​generated. Combined with the classification list of physical connection methods, the mechanical coupling strength is assigned according to the mechanical connection types of bolt connection, welding, and snap-on assembly to form a mechanical coupling strength mapping table. For example, welding is the highest strength benchmark for rigid connection, and the vibration energy transfer efficiency of welding is close to 100% (no gap), and the assigned value is 1.0; due to the attenuation of preload and micron-level gaps, the vibration transfer efficiency of bolt connection is about 80%~85% of welding, and the middle value is 0.8; snap-on connection relies on elastic deformation locking, and is prone to fatigue failure due to long-term vibration. The transfer efficiency is about 50% of welding, and the assigned value is 0.5.

[0029] S1.3. Calculate the edge weights of the adjacency matrix based on the physical distance and mechanical coupling strength.

[0030] It should be noted that the edge weight values ​​of all positions to be filled in the adjacency matrix are calculated based on the physical distance matrix and the mechanical coupling strength mapping table. The expression is: ; in, is the edge weight value, indicating the wireless monitoring node and wireless monitoring nodes The strength of association; is the mechanical coupling strength, indicating that the wireless monitoring node and wireless monitoring nodes The mechanical connection rigidity is determined by the connection type; Is the physical distance, indicating the wireless monitoring node and wireless monitoring nodes The Euclidean distance between them is calculated based on the three-dimensional coordinates of the wireless monitoring nodes; is the row index in the adjacency matrix, indicating the node number; is the column index in the adjacency matrix, representing the node number.

[0031] S2. The vibration signal is enhanced by light-vibration coupling through a photonic crystal resonant cavity to generate an amplified signal, and the amplified signal is compressed based on a quantum compressed sensing algorithm to generate a compressed signal.

[0032] S2.1. Define the target frequency band for mechanical vibration based on typical mechanical faults of on-load tap changers; It should be noted that the refractive index of the basic material of the photonic crystal resonator is determined according to the material manual, the effective refractive index is calculated in combination with the photonic crystal structure, and based on the effective refractive index, the lattice constant is calculated using the photonic bandgap center frequency formula; based on the fault detection requirements of the on-load tap changer and Fourier transform analysis, the target frequency band for the design of the mechanical fault characteristic frequency band and anti-interference performance is 1-5kHz: the 1-5kHz frequency band covers the 2-3kHz characteristic range of gear loosening fault (based on the American National Standards Institute gear fault frequency band division and the 3 times gear meshing frequency harmonic enhancement effect), the 4-5kHz sensitive area of ​​contact wear (derived from the national standard "High Voltage AC Circuit Breaker") The high-frequency characteristics of arc erosion caused by spring failure (the fundamental frequency of stress waves of 1–2 kHz) and the transient impact response characteristics of the spring failure (in compliance with the national standard "Mechanical Spring Fatigue Test Method") are tested, while avoiding the electromagnetic noise interference area of ​​<1 kHz and the signal attenuation area of ​​>5 kHz. The bandgap response frequency band of the photonic crystal resonator is calculated using the formula for the center frequency of the photonic crystal bandgap based on the lattice constant and effective refractive index of the photonic crystal resonator. The transmission spectrum of the photonic crystal resonator is calculated by scanning the target frequency band of mechanical vibration using the finite element method. If the transmission spectrum shows obvious attenuation within the range of the target frequency band of mechanical vibration, it means that the bandgap response frequency band completely covers the target frequency band of mechanical vibration. It should also be noted that the Fourier transform extracts fault features by converting vibration signals into frequency-domain energy distributions. This involves two technical approaches: vibration frequency-domain spectrum analysis and high- and low-frequency envelope consistency analysis. According to test data from the high-voltage AC circuit breaker condition-based maintenance test procedure, the proportion of high-frequency energy in the 10-20kHz range for on-load tapchangers increases to 57.42% when contacts are worn, while the proportion of low-frequency energy in the 0-10kHz range rises to 96.15% when contacts are stuck. Under normal conditions, the waveforms and amplitudes of the high- and low-frequency envelopes are highly consistent, but exhibit significant deviations during faults. However, traditional Fourier transforms suffer from inherent flaws such as a lack of time-frequency localization capability, sensitivity to strong noise environments (weak features are easily submerged in electromagnetic interference >30dB), and an inability to simultaneously optimize time-frequency resolution. These flaws result in insufficient diagnostic accuracy when analyzing high-frequency harmonics >5kHz, resulting in a false negative rate exceeding 22%. The 1-5kHz target frequency band signal is enhanced through the band gap resonance of the photonic crystal resonator, forming a collaborative application mode with traditional Fourier analysis: based on the fast Fourier transform, the energy abnormal frequency band is preliminarily screened (for example, a sudden increase of 6-10dB in energy in the 2-3kHz frequency band indicates loose gears), and then the photonic crystal resonator is started to enhance the optical-vibration coupling of the energy abnormal frequency band (gain > 15dB), and finally the enhanced vibration signal is compressed through the quantum compressed sensing algorithm.

[0033] S2.2. Integrate a photonic crystal resonator with a wireless monitoring node, excite the photonic crystal resonator through laser light, generate an optical-resonance coupling effect, convert mechanical vibrations into modulated optical signals, and amplify the modulated optical signals through bandgap resonance. It should be noted that the photonic crystal resonant cavity is integrated with the three-axis acceleration sensor of the wireless monitoring node, and the photonic crystal resonant cavity is illuminated by excitation light with a wavelength of 1550nm emitted by a laser; mechanical vibration causes periodic deformation of the photonic crystal, and the vibration signal is converted into a wavelength-modulated light signal through the light-vibration coupling effect; the band gap resonance effect of the photonic crystal in the target frequency band is utilized to selectively amplify the modulated light signal while suppressing out-of-band noise.

[0034] S2.3. Convert the amplified modulated optical signal into an electrical signal through a photodetector and record the real-time Q value of the photonic crystal resonant cavity; It should be noted that the modulated optical signal is converted into a voltage signal by an InGaAs photodetector, and the voltage amplitude and time series are synchronously recorded by a dynamic signal analyzer at a sampling rate of 100kHz, and the timestamp alignment is ensured by a hardware trigger signal; a micro-spectral analyzer is integrated at the output end of the photonic crystal resonator to collect the transmission spectrum of the photonic crystal resonator in real time, and the fast Fourier transform is used to extract the central wavelength and half-width of the resonance peak, and the real-time Q value is calculated using the resonance peak half-width formula.

[0035] S2.4. Construct a quantum random projection matrix to perform linear projection on the electrical signal, compress the vibration signal into a low-dimensional space, and generate a compressed signal.

[0036] It should be noted that the number of rows m of the quantum random projection matrix is ​​calculated based on the real-time Q value to determine the compression dimension; the time window length is set to n=512 according to the ISO standard, and an m×n quantum random projection matrix is ​​constructed. The electrical signal is sampled and quantized according to the time window length n through the ADC analog-to-digital converter to obtain discrete digital signal fragments (essentially the digital representation of the vibration signal). The digital signal fragments are subjected to matrix multiplication with the quantum random projection matrix to realize linear projection. The vibration signal is projected into a low-dimensional space through linear projection to generate a compressed signal.

[0037] S3. Perform orthogonal matching pursuit sparse reconstruction on the compressed signal, extract the time-frequency joint features, input the time-frequency joint features into the local convolutional neural network, and generate the local convolutional neural network parameters.

[0038] S3.1. Initialize the orthogonal matching pursuit algorithm, build a sparse basis dictionary based on the Fourier transform basis and the wavelet basis, and reconstruct the vibration signal based on the quantum random projection matrix and the sparse basis dictionary; It should be noted that the residual threshold is defined as 5% through the experimental calibration method; a joint sparse basis dictionary is constructed by combining the Fourier transform basis and the wavelet basis, wherein the Fourier transform basis is used to capture the periodic frequency components of the compressed signal, and the wavelet basis is used to extract transient impact features, and the transient impact features are used to capture short high-energy pulse signals in the mechanical vibration signal; the compressed signal and the quantum random projection matrix are input into the orthogonal matching pursuit algorithm, the initial residual is set to the compressed signal itself, the iteration counter is reset to zero, and the orthogonal matching pursuit algorithm is initialized based on the residual threshold as the stopping condition; the basis vector that contributes the most to the residual in the sparse basis dictionary is iteratively selected until the residual energy is lower than the residual threshold or the maximum number of iterations is reached. When the residual energy is lower than the residual threshold, the vibration signal is considered to have been fully restored; the error between the reconstructed vibration signal and the compressed signal is calculated, and the error allowable range is set according to the mechanical vibration evaluation standard. If the error exceeds the error allowable range, the wavelet basis type is switched, and finally a reconstructed vibration signal that meets the accuracy requirements is output.

[0039] S3.2. Perform wavelet packet decomposition on the reconstructed vibration signal to generate a time-frequency energy distribution matrix. Calculate the instantaneous frequency gradient, energy center of gravity offset, and energy proportion of each sub-band of the time-frequency energy distribution matrix to generate a time-frequency joint feature. It should be noted that the reconstructed vibration signal is decomposed into 8 sub-band nodes by performing a three-layer wavelet packet decomposition. The energy value of each sub-band node is calculated to obtain the energy sequence of each sub-band on the time axis. The energy sequence of each sub-band is arranged in two dimensions of time and frequency to construct the behavioral time points, which are listed as the time-frequency energy distribution matrix of the sub-band. Based on the time-frequency energy distribution matrix, the instantaneous frequency gradient of each sub-band is calculated by the Sobel gradient operator in the time dimension to reflect the frequency fluctuation characteristics of the reconstructed vibration signal. The energy center of gravity offset is calculated by the center of gravity formula in the frequency dimension to characterize the stability of the energy distribution. The proportion of the energy of each sub-band to the total energy is counted to generate an energy proportion vector. The instantaneous frequency gradient, energy center of gravity offset and energy proportion vector are spliced ​​into a time-frequency joint feature by time alignment.

[0040] S3.3. Construct a local convolutional neural network, input the time-frequency joint features into the local convolutional neural network, and generate the convolution kernel weights and fully connected layer parameters.

[0041] It should be noted that the hierarchical structure of the local convolutional neural network is designed based on the joint time-frequency features, including the input layer, the first local convolutional layer, the second local convolutional layer, the maximum pooling layer, the flattening layer, the fully connected layer, and the output layer; the activation function of the first local convolutional layer, the second local convolutional layer, and the fully connected layer is ReLU, and the activation function of the output layer is Softmax; the parameters of the local convolutional neural network (convolution kernel weights and fully connected layer parameters) are randomly initialized using a normal distribution (mean 0, variance 0.01), and L2 regularization terms are added to the first local convolutional layer and the second local convolutional layer to prevent overfitting. The structure of the local convolutional neural network is implemented using the Keras framework; The joint time-frequency features are input into the local convolutional neural network and forward propagation is performed. After the input layer receives the joint time-frequency features, the first local convolutional layer extracts the local correlation pattern between adjacent time points and feature dimensions through a 3×3 convolution kernel sliding, and outputs a feature map. The second local convolutional layer further compresses the feature dimensions and enhances the abstraction ability to generate a high-order feature map. The significant feature map is downsampled by the maximum pooling layer, flattened, and input into the fully connected layer to calculate the fault probability distribution. The fault category with the highest fault probability is taken as the predicted label. The current fault of the on-load tap changer is determined as the true label based on the equipment inspection report. The error between the predicted label and the true label is calculated using the cross-entropy loss function, and the Adam optimizer is used for back propagation to iteratively update the convolution kernel weights and the fully connected layer parameters until the loss function converges or the maximum number of iterations is reached.

[0042] S4. Calculate the federated aggregation weight based on the edge weight and signal-to-noise ratio of the adjacency matrix, fuse the local convolutional neural network parameters, and build a global diagnostic model based on the distributed aggregation protocol of the federated learning framework.

[0043] S4.1. Extract the edge weights between each wireless monitoring node from the adjacency matrix, combine the signal-to-noise ratio of each wireless monitoring node, and calculate the federated aggregation weight of each wireless monitoring node using the weight distribution rule defined by the federated learning framework; It should be noted that the adjacency matrix contains off-diagonal elements and main diagonal elements. The main diagonal elements refer to the current wireless monitoring node itself having no communication link; each element represents the quality of the communication link between wireless monitoring nodes, where if the element value is 1, it indicates an ideal channel with no packet loss and low latency; if the element value is 0, it indicates communication interruption; extract all off-diagonal elements in the adjacency matrix, arrange them into an edge weight list according to the node number, and arrange all element values ​​into a one-dimensional list to generate an edge weight vector; calculate the real-time signal-to-noise ratio measurement value of each wireless monitoring node through the SNR formula to generate a signal Noise ratio vector; based on the weight distribution rule defined in the federated learning framework, the weight distribution rule is to determine the federated aggregation weight of each wireless monitoring node by multiplying the mean of the edge weight vector and the normalized signal-to-noise ratio vector; for each wireless monitoring node, calculate the arithmetic mean of the edge weight vectors of the communication links with all adjacent wireless monitoring nodes; linearly map the signal-to-noise ratio vectors of all wireless monitoring nodes to the range of 0-1 to generate a normalized signal-to-noise ratio vector; multiply the arithmetic mean of the edge weight vector of each wireless monitoring node by the normalized signal-to-noise ratio vector to calculate the federated aggregation weight of each wireless monitoring node.

[0044] S4.2. Perform weighted averaging on the local convolutional neural network parameters uploaded by each wireless monitoring node according to the federated aggregation weight to generate global diagnostic model parameters. It should be noted that the local convolutional neural network parameters (including convolution kernel weights and fully connected layer parameters) uploaded by all wireless monitoring nodes are received, and the federated aggregation weights are obtained synchronously; in the federated learning framework, the local convolutional neural network parameters uploaded by each wireless monitoring node are standardized; for the convolution layer weights, if the number of convolution kernel output channels of a wireless monitoring node is less than the number of convolution kernel channels of the local convolutional neural network, the corresponding number of all-zero channels are added to the output channel dimension, otherwise the redundant channels are truncated; for the fully connected layer parameters, if the number of input or output neurons of the wireless monitoring node is inconsistent, the corresponding dimension of the fully connected layer of the local convolutional neural network is padded with zeros or truncated; the convolution kernel weights and fully connected parameters of all wireless monitoring nodes have exactly the same dimensions, thereby supporting layer-by-layer weighted averaging calculation to generate global diagnostic model parameters; according to the weight value in the federated aggregation weight, the aligned local convolutional neural network parameters are weighted summed layer by layer to generate the weighted averaged global diagnostic model parameters.

[0045] S4.3. The federated learning framework loads the global diagnostic model parameters based on the structure of the local convolutional neural network to build a global diagnostic model.

[0046] It should be noted that the local convolutional neural network structure of each wireless monitoring node is analyzed, and the hierarchical structure shared by the wireless monitoring nodes is dynamically selected as the global diagnostic model skeleton benchmark. The input layer, convolution layer, pooling layer, fully connected layer and output layer are created in the order of the skeleton benchmark structure. Among them, the convolution layer parameter is the convolution kernel weight, the pooling layer type is maximum pooling, and the fully connected layer parameter is the weight matrix. The rows of the weight matrix represent the number of neurons in the previous layer, and the columns represent the number of neurons in the current layer. A parameter container is created for each layer, with the dimension consistent with the skeleton benchmark and the initial value being empty. The global diagnostic model parameters are loaded into the global diagnostic model skeleton and filled into the parameter container of the corresponding level of the global diagnostic model skeleton. The rationality of the global diagnostic model is verified by forward propagation, and the complete global diagnostic model is output. Initialize the global diagnostic model parameters and distribute them to each wireless monitoring node. Each wireless monitoring node performs local training based on the local vibration signal and uses stochastic gradient descent or Adam optimizer to update the local global diagnostic model parameters (including convolution kernel weights and fully connected layer weights). Gradient clipping and dynamic learning rate adjustment are added during the local training process. After each round of training, the wireless monitoring node uploads the updated global diagnostic model parameters to the server. After receiving the data, the server performs weighted aggregation based on the federated averaging algorithm, where the weights are dynamically allocated according to the signal-to-noise ratio of the wireless monitoring node data, and adds differential privacy protection to the aggregated global diagnostic model parameters to generate global diagnostic model parameters that meet privacy constraints. The server then distributes the updated global diagnostic model parameters to all wireless monitoring nodes and starts the next round of local training. At the same time, it performs global diagnostic model verification and calculates the cross-entropy loss. If the verification loss does not decrease for N consecutive rounds, the early stopping mechanism is triggered, and the converged global diagnostic model is finally output. The training termination condition is that the preset maximum number of iterations is reached. The expression of the global diagnostic model is, ; in, Indicates the fault classification score, which is used to quantify the fault probability of the current state of the equipment; Indicates time From the initial moment To the end time The integral operation of Represents the time-frequency joint characteristic function, which describes the vibration signal in time The time-frequency eigenvalue at ; represents the complex exponential function, yes , is an imaginary singular number, represents the angular frequency, Indicates a moment; is the time differential variable, which represents the time Make infinitesimal divisions; represents the mechanical coupling strength; Indicates the Q value attenuation coefficient; The strength of mechanical coupling directly determines the effective transmission efficiency of vibration energy from the internal structure to the outer shell. Strong coupling ensures that the vibration signal in the 1.5-15kHz frequency band can maintain sufficient energy base when passing through the steel shell, preventing the time-frequency joint feature from being overwhelmed by background noise. The time-frequency joint feature relies on the high-Q value photonic crystal resonator to achieve narrowband resonance enhancement, but the Q value is affected by mechanical damping and produces an attenuation coefficient. The three exist in a nonlinear balance relationship: when the mechanical coupling strength is increased to 0.9, the shell stiffness increases, resulting in an increase in the Q value attenuation coefficient. At this time, the time-frequency ridge tracking algorithm is required to compensate for the loss of frequency resolution caused by the decrease in Q value. Conversely, under the condition of loose flange bolts, although the Q value attenuation coefficient drops to 0.01, causing the Q value to recover, the insufficient energy transfer efficiency causes the time-frequency joint feature to decrease by 12dB. In this case, the time-frequency joint feature is reconstructed using wavelet packet energy entropy. It should also be noted that the wavelet packet energy entropy is calculated based on the time-frequency energy distribution matrix of each wireless monitoring node generated by wavelet packet decomposition of the vibration signal and the Shannon information entropy formula, and the expression is: ; in, represents the energy entropy of wavelet packet; Indicates the number of wavelet packet decomposition layers; Indicates the total number of sub-bands; Indicates the subband index, ranging from 1- ; Indicates the The energy probability in each sub-band; S5. Calculate the Q-value attenuation coefficient of the photonic crystal resonant cavity, input the mechanical coupling strength, time-frequency joint characteristics, and Q-value attenuation coefficient into the global diagnosis model, and generate fault classification results and maintenance decisions.

[0047] S5.1. Define an initial Q value based on the frequency response curve of the photonic crystal resonator, calculate the Q value attenuation coefficient based on the difference between the real-time Q value and the initial Q value, and define the Q value attenuation coefficient threshold through experimental calibration; It should be noted that a vector network analyzer is used to measure the frequency response curve of the photonic crystal resonant cavity in a fault-free state, and the ratio of the resonant frequency to the half-power bandwidth is extracted as the initial Q value; the Q value attenuation coefficient is generated based on the percentage difference between the real-time Q value and the initial Q value; typical faults are artificially simulated and the Q value attenuation coefficient data is recorded, and the normal fluctuation range is determined in combination with the statistical distribution, and the critical value exceeding three times the standard deviation of the mean is selected as the Q value attenuation coefficient threshold. If the Q value attenuation coefficient exceeds the Q value attenuation coefficient threshold range, it indicates that there is a risk of equipment failure.

[0048] S5.2. Input the mechanical coupling strength, time-frequency joint characteristics, and Q-value attenuation coefficient into the global diagnosis model to generate a fault classification result; It should be noted that the one-dimensional vectors of the flattened mechanical coupling strength and the time-frequency joint features are spliced ​​and merged with the Q-value attenuation coefficient to form a comprehensive input vector and Z-score normalized. The normalized comprehensive input vector is then input into the global diagnosis model. The global diagnosis model extracts the high-order mode of the time-frequency features through the convolution layer, combines the physical correlation of the mechanical coupling strength and the Q-value attenuation coefficient with the fully connected layer, and finally generates the fault classification score by the output layer; the Sigmoid function is used to independently output the probability of each type of fault, and the fault probability range is within 0-1; through 3 The principle is to determine the fault probability threshold. Based on the historical operating data and known faults of the equipment, the fault classification scores output by the global diagnostic model are statistically analyzed and converted into fault probabilities. The mean and standard deviation are calculated, and the fault probability threshold is determined to be the mean + 3 standard deviations. If the probability of a certain type of fault exceeds the fault probability threshold, it is determined that a fault exists. For example, if the gear crack is 0.9 and the lubrication failure is 0.6 (the fault probability thresholds are both 0.5), it is determined to be "gear crack + lubrication failure".

[0049] S5.3. Determine the fault level based on the fault classification result and the Q-value attenuation coefficient threshold, and generate a maintenance strategy.

[0050] It should be noted that the fault level is determined based on the fault probability and the Q-value attenuation coefficient threshold. When the fault probability is 30%-70% and the Q-value attenuation coefficient is less than 0.2, it indicates a minor equipment fault. When the fault probability is 70%-90% and the Q-value attenuation coefficient is 0.2≤<0.4, it indicates a moderate equipment fault. When the fault probability is greater than 90% and the Q-value attenuation coefficient is ≥0.4, it indicates a serious equipment fault. Based on the fault level and real-time equipment operating conditions (such as load, temperature, and operating time), rules matching the current fault type and level are retrieved from the maintenance strategy knowledge base, such as "serious fault + high load → immediate shutdown" and "moderate fault + low loss → planned maintenance", to generate specific maintenance strategies (such as component replacement, lubrication, and enhanced monitoring). Fault information is generated based on the fault type, fault level, fault location, and fault time. The maintenance strategy is bound to the fault information and pushed to the operation and maintenance management unit through the interface. The equipment health file is also updated synchronously to record the fault handling trajectory.

[0051] This embodiment also provides an on-load tap changer vibration detection system based on wireless sensing and photonic crystal resonant cavity, comprising: a node deployment module, a signal compression module, a feature extraction module, a model construction module, and a maintenance decision module; The node deployment module is used to deploy wireless monitoring nodes to collect vibration signals, build an adjacency matrix based on the mechanical topological connection relationship, and calculate the edge weights of the adjacency matrix based on the physical distance and mechanical coupling strength between the wireless monitoring nodes; A signal compression module is used to enhance the optical-resonant coupling of the vibration signal through a photonic crystal resonant cavity to generate an amplified signal, and to compress the amplified signal based on a quantum compressed sensing algorithm to generate a compressed signal; The feature extraction module is used to perform orthogonal matching pursuit sparse reconstruction on the compressed signal, extract the time-frequency joint features, input the time-frequency joint features into the local convolutional neural network, and generate the local convolutional neural network parameters; A model building module is used to calculate the federated aggregation weight based on the edge weights and signal-to-noise ratio of the adjacency matrix, fuse the local convolutional neural network parameters, and build a global diagnostic model based on the distributed aggregation protocol of the federated learning framework; The maintenance decision module is used to calculate the Q-value attenuation coefficient of the photonic crystal resonator, input the mechanical coupling strength, time-frequency joint characteristics and Q-value attenuation coefficient into the global diagnosis model, and generate fault classification results and maintenance decisions.

[0052] This embodiment further provides a computer device applicable to the on-load tap changer vibration detection method based on wireless sensing and a photonic crystal resonant cavity, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the on-load tap changer vibration detection method based on wireless sensing and a photonic crystal resonant cavity as proposed in the above embodiment.

[0053] The computer device may be a terminal, comprising a processor, memory, a communication interface, a display, and an input device connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores an operating system and computer programs. The internal memory provides an environment for the operating system and computer programs stored in the non-volatile storage media. The communication interface of the computer device is used to communicate with external terminals via wired or wireless communication. Wireless communication may be achieved via Wi-Fi, a carrier network, NFC (near-field communication), or other technologies. The display of the computer device may be a liquid crystal display or an electronic ink display. The input device may be a touchscreen overlay on the display, buttons, a trackball, or a touchpad on the computer device housing, or an external keyboard, touchpad, or mouse.

[0054] This embodiment further provides a storage medium having a computer program stored thereon. When the program is executed by a processor, the program implements the on-load tap changer vibration detection method based on wireless sensing and a photonic crystal resonant cavity as proposed in the above embodiment. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0055] In summary, the present invention converts mechanical vibration signals into high-sensitivity modulated light signals through the optical-vibration coupling effect of the photonic crystal resonant cavity, thereby solving the problems of low signal-to-noise ratio of weak fault characteristics and poor data transmission efficiency at high sampling rates in traditional sensors under strong noise environments. At the same time, the federation aggregation weights are dynamically calculated through physical distance and mechanical coupling strength, thereby optimizing the parameter fusion process of the global diagnostic model and solving the problems of slow convergence speed and poor global consistency of the global diagnostic model.

[0056] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. 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 detecting vibration of an on-load tap changer based on wireless sensing and a photonic crystal resonant cavity, characterized by: include, Wireless monitoring nodes are deployed to collect vibration signals. An adjacency matrix is ​​constructed based on the mechanical topology connection relationship. The edge weights of the adjacency matrix are calculated based on the physical distance between the wireless monitoring nodes and the mechanical coupling strength. The vibration signal is enhanced by optical-resonant coupling through a photonic crystal resonant cavity to generate an amplified signal, and the amplified signal is compressed based on a quantum compressed sensing algorithm to generate a compressed signal. Perform orthogonal matching pursuit sparse reconstruction on the compressed signal to extract the time-frequency joint features, input the time-frequency joint features into the local convolutional neural network, and generate the local convolutional neural network parameters; The federated aggregation weight is calculated based on the edge weight and signal-to-noise ratio of the adjacency matrix, the local convolutional neural network parameters are fused, and a global diagnostic model is constructed based on the distributed aggregation protocol of the federated learning framework; The Q-value attenuation coefficient of the photonic crystal resonator is calculated, and the mechanical coupling strength, time-frequency joint characteristics and Q-value attenuation coefficient are input into the global diagnosis model to generate fault classification results and maintenance decisions.

2. The on-load tap changer vibration detection method based on wireless sensing and photonic crystal resonant cavity according to claim 1, characterized in that: The edge weights of the adjacency matrix are calculated based on the physical distance and mechanical coupling strength between wireless monitoring nodes. The specific steps are as follows: Determine the mechanical topology of the on-load tap-changer, identify the physical connection of core components, mark the deployment location of wireless monitoring nodes, and collect vibration signals in real time; An adjacency matrix is ​​constructed based on the mechanical topological connection relationship, the physical distance between wireless monitoring nodes is measured, and the mechanical coupling strength between each wireless monitoring node is calibrated according to the mechanical connection type between the components; Calculate the edge weights of the adjacency matrix based on the physical distance and mechanical coupling strength.

3. The on-load tap changer vibration detection method based on wireless sensing and photonic crystal resonant cavity according to claim 2, characterized in that: The specific steps of generating the compressed signal are as follows: Define the target frequency band of mechanical vibration according to the photonic crystal resonant cavity; The photonic crystal resonator is integrated with the wireless monitoring node. The photonic crystal resonator is excited by laser to generate the optical-oscillator coupling effect to convert the mechanical vibration into a modulated optical signal, which is then amplified by the bandgap resonance method. The amplified modulated optical signal is converted into an electrical signal through a photodetector, and the real-time Q value of the photonic crystal resonant cavity is recorded; A quantum random projection matrix is ​​constructed to perform linear projection on the electrical signal, compressing the vibration signal into a low-dimensional space to generate a compressed signal.

4. The on-load tap changer vibration detection method based on wireless sensing and photonic crystal resonant cavity according to claim 3, characterized in that: The specific steps of generating local convolutional neural network parameters are as follows: Initialize the orthogonal matching pursuit algorithm, build a sparse basis dictionary based on the Fourier transform basis and the wavelet basis, and reconstruct the vibration signal based on the quantum random projection matrix and the sparse basis dictionary; The reconstructed vibration signal is decomposed by wavelet packets to generate a time-frequency energy distribution matrix. The instantaneous frequency gradient, energy center offset, and energy proportion of each sub-band of the time-frequency energy distribution matrix are calculated to generate a time-frequency joint feature. Construct a local convolutional neural network, input the time-frequency joint features into the local convolutional neural network, and generate the convolution kernel weights and fully connected layer parameters.

5. The on-load tap changer vibration detection method based on wireless sensing and photonic crystal resonant cavity according to claim 4, characterized in that: The federation aggregation weight is calculated based on the edge weight and signal-to-noise ratio of each wireless monitoring node in the adjacency matrix. The specific steps are as follows: Extract edge weights between wireless monitoring nodes from the adjacency matrix; Combined with the signal-to-noise ratio of each wireless monitoring node, the federated aggregation weight of each wireless monitoring node is calculated through the weight distribution rule defined by the federated learning framework.

6. The on-load tap changer vibration detection method based on wireless sensing and photonic crystal resonant cavity according to claim 5, characterized in that: The distributed aggregation protocol based on the federated learning framework builds a global diagnostic model. The specific steps are as follows: The local convolutional neural network parameters uploaded by each wireless monitoring node are weighted averaged according to the federated aggregation weight to generate the global diagnostic model parameters; The federated learning framework loads the global diagnostic model parameters based on the structure of the local convolutional neural network to build a global diagnostic model.

7. The on-load tap changer vibration detection method based on wireless sensing and photonic crystal resonant cavity according to claim 6, characterized in that: The specific steps for generating fault classification results and maintenance decisions are as follows: The initial Q value is defined according to the frequency response curve of the photonic crystal resonator, the Q value attenuation coefficient is calculated according to the difference between the real-time Q value and the initial Q value, and the Q value attenuation coefficient threshold is defined through experimental calibration method; The mechanical coupling strength, time-frequency joint characteristics and Q-value attenuation coefficient are input into the global diagnosis model to generate fault classification results; The fault level is determined based on the fault classification results and the Q-value attenuation coefficient threshold, and a maintenance strategy is generated.

8. A system for detecting vibration of an on-load tap changer based on wireless sensing and a photonic crystal resonant cavity, based on the method for detecting vibration of an on-load tap changer based on wireless sensing and a photonic crystal resonant cavity according to any one of claims 1 to 7, characterized in that: include, Node deployment module, signal compression module, feature extraction module, model building module and maintenance decision module; The node deployment module is used to deploy wireless monitoring nodes to collect vibration signals, build an adjacency matrix based on the mechanical topological connection relationship, and calculate the edge weights of the adjacency matrix based on the physical distance and mechanical coupling strength between the wireless monitoring nodes; A signal compression module is used to enhance the optical-resonant coupling of the vibration signal through a photonic crystal resonant cavity to generate an amplified signal, and to compress the amplified signal based on a quantum compressed sensing algorithm to generate a compressed signal; The feature extraction module is used to perform orthogonal matching pursuit sparse reconstruction on the compressed signal, extract the time-frequency joint features, input the time-frequency joint features into the local convolutional neural network, and generate the local convolutional neural network parameters; A model building module is used to calculate the federated aggregation weight based on the edge weights and signal-to-noise ratio of the adjacency matrix, fuse the local convolutional neural network parameters, and build a global diagnostic model based on the distributed aggregation protocol of the federated learning framework; The maintenance decision module is used to calculate the Q-value attenuation coefficient of the photonic crystal resonator, input the mechanical coupling strength, time-frequency joint characteristics and Q-value attenuation coefficient into the global diagnosis model, and generate fault classification results and maintenance decisions.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the on-load tap changer vibration detection method based on wireless sensing and photonic crystal resonant cavity according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the on-load tap changer vibration detection method based on wireless sensing and a photonic crystal resonant cavity according to any one of claims 1 to 7 are implemented.

Images