Virtual multichannel compressed sensing sampling method and system for partial discharge signals

Abstract

The invention discloses a virtual multichannel compressed sensing sampling method and system for partial discharge signals, and belongs to the field of power equipment state monitoring. The problems of high sampling cost, large data volume and low signal reconstruction precision in the existing partial discharge signal sampling method are solved, non-uniform sampling is carried out by adopting a single physical sampling channel, and enough information is obtained at a relatively low sampling rate for signal reconstruction; an observation sequence obtained by sampling is distributed to a plurality of virtual channels according to a time sequence, compressed sensing reconstruction is carried out by using a joint observation matrix matched with a random sampling rate and a virtual channel distribution mode, a complete partial discharge digital signal is recovered at one time, the hardware cost and the system power consumption are effectively reduced, and the system reliability is improved. Therefore, the precision and efficiency of signal reconstruction are improved, and an efficient and reliable solution is provided for state monitoring and fault diagnosis of power equipment.

Description

Technical Field

[0001] This invention relates to the field of power equipment condition monitoring technology, specifically to a virtual multi-channel compressed sensing sampling method and system for partial discharge signals. Background Technology

[0002] Partial discharge detection is a key technology for assessing the insulation status of power equipment (such as transformers, gas-insulated switchgear (GIS), and cables). By analyzing the characteristics of partial discharge signals, the degree and type of insulation degradation can be effectively determined, which is of great significance for preventing electrical equipment failures and ensuring the safe and stable operation of the power grid. Traditional partial discharge signal detection methods usually rely on high-frequency current transformers (HFCTs) or ultrasonic sensors for acquisition, and use high-speed data acquisition systems to uniformly sample the signals. According to the Nyquist sampling theorem, in order to capture partial discharge pulse signals (usually containing high-frequency components of 20-80MHz) without distortion, the sampling rate must be at least twice the highest frequency of the signal. This requires the acquisition system to have an extremely high sampling rate, generating massive amounts of data, which poses a huge challenge to data storage, transmission, and processing. It also makes the monitoring system expensive and power-consuming, which is not conducive to long-term online monitoring applications. Therefore, this method does not meet the current requirements. To address this, we propose a virtual multi-channel compressed sensing sampling method and system for partial discharge signals. Summary of the Invention

[0003] The purpose of this invention is to provide a virtual multi-channel compressed sensing sampling method and system for partial discharge signals. By using a single physical sampling channel for non-uniform sampling and distributing the sampled observation sequence to multiple virtual channels in chronological order, compressed sensing reconstruction is performed using a joint observation matrix that matches the random sampling rate and the virtual channel allocation method, thus recovering the complete digital partial discharge signal in one go, thereby solving the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a virtual multi-channel compressed sensing sampling method for partial discharge signals, the method comprising:

[0005] A low-speed observation sequence is obtained by non-uniformly sampling the partial discharge simulation signal using a single physical sampling channel.

[0006] The observation sequence is sequentially and cyclically distributed to N virtual channels according to the time order, resulting in N sub-observation sequences;

[0007] Based on the compressed sensing reconstruction algorithm, the N parallel sub-observation sequences are jointly reconstructed using a joint observation matrix that matches the random sampling rate and virtual channel allocation method to recover the complete partial discharge digital signal.

[0008] The reconstructed partial discharge signal is post-processed, including signal filtering and feature extraction.

[0009] Furthermore, the partial discharge simulation signal is non-uniformly sampled using a single physical sampling channel, specifically as follows:

[0010] A partial discharge simulation signal is input into a single physical sampling channel, which consists of an analog-to-digital converter and its pre-amplifier signal conditioning circuit.

[0011] Before sampling, a random sampling rate for non-uniform sampling is preset based on the characteristics of the partial discharge signal and the required compression ratio. Here, the highest frequency of the partial discharge signal is set to... The required compression ratio is It satisfies 0 < If the random sampling rate is less than 1, then the random sampling rate is... Calculated using the following formula:

[0012]

[0013] In the formula, This indicates the compression ratio, used to control the sparsity of sampling points;

[0014] When the partial discharge simulation signal is input, the physical sampling channel samples the input partial discharge simulation signal at each randomly generated sampling time point using an analog-to-digital converter according to a pre-set random sampling rate, and converts the instantaneous amplitude of the simulation signal into a digital signal. Let the set of sampling time points be denoted as . , Let be the total number of sampling points. Then the sampling process can be represented as:

[0015]

[0016] In the formula, Indicates the first The digital signal value of each sampling point;

[0017] Digital signals are collected and stored sequentially, forming a low-speed observation sequence.

[0018] Furthermore, the random sampling rate is determined based on the sparsity of the partial discharge signal and the highest frequency component of the signal. The random sampling rate determines the time interval of each sampling, and the time interval is randomly distributed to ensure that sufficient information is obtained for subsequent signal reconstruction under non-uniform sampling conditions.

[0019] Furthermore, the observation sequence is sequentially and cyclically distributed to N virtual channels, specifically as follows:

[0020] The allocation of virtual channels is achieved using software algorithms. A loop counter is set with an initial value of 0. After each sampling point is allocated, the loop counter is incremented by 1. When the value of the loop counter reaches N, the loop counter is reset to zero and a new round of loop allocation begins.

[0021] After obtaining a low-speed observation sequence, starting from the first sampling point of the observation sequence, each sampling point is sequentially assigned to the data buffer corresponding to the virtual channel indicated by the current loop counter. This process is repeated until all sampling points in the observation sequence have been assigned.

[0022] Furthermore, the allocation period is dynamically adjusted according to the pulse repetition frequency of the partial discharge signal, specifically as follows:

[0023] By performing continuous short-time Fourier transforms on the partial discharge signal, the pulse repetition frequency characteristics of the partial discharge signal in different time segments are obtained.

[0024] Based on the monitored pulse repetition frequency characteristics, the allocation period of the observation sequence to the virtual channel is dynamically adjusted;

[0025] When the signal pulse repetition frequency is high, the allocation period is shortened, allowing the virtual channel to receive new sampled data more frequently.

[0026] When the signal pulse repetition frequency is low, extend the allocation period and reduce the data update frequency of the virtual channel;

[0027] Among them, the allocation cycle Based on the monitored pulse repetition frequency Nonlinear dynamic adjustment is performed, as shown in the following formula:

[0028]

[0029] In the formula, The base allocation period; and These are the expected minimum and maximum pulse repetition frequencies, used to define the effective adjustment range; The attenuation coefficient is controlled. Follow The rate at which growth declines; This is the gain coefficient; It is a very small positive number, used to prevent the denominator from being zero; when When height increases, Significantly reduced; when When it decreases, Smooth increase.

[0030] Furthermore, an independent data buffer is allocated to each virtual channel to store the sampling point data allocated to that virtual channel. During the allocation process, the filling status of the data buffer of each virtual channel is monitored in real time. When the data buffer of any virtual channel is filled to a predetermined length, compressed sensing reconstruction processing of the data in that data buffer is triggered.

[0031] Furthermore, based on the compressed sensing reconstruction algorithm, the N parallel sub-observation sequences are jointly reconstructed using a joint observation matrix that matches the random sampling rate and virtual channel allocation method, specifically as follows:

[0032] A joint observation matrix is ​​constructed based on the previously set random sampling rate and virtual channel allocation method. The joint observation matrix is ​​used to reflect the distribution of the observation sequence in different virtual channels and its intrinsic relationship with the original partial discharge signal. The construction of the joint observation matrix is ​​as follows:

[0033]

[0034] in, Represents the joint observation matrix; Indicates the first The observation matrix of each virtual channel, each observation matrix reflecting the distribution of sampling points in that channel;

[0035] The N parallel sub-observation sequences are used as input and substituted into the compressed sensing reconstruction algorithm. During the iteration process of the algorithm, the original partial discharge digital signal is gradually approximated through matrix operations using the determined joint observation matrix.

[0036] After multiple iterative calculations using the compressed sensing reconstruction algorithm, a complete digital signal of partial discharge is obtained in one go.

[0037] Furthermore, when constructing the joint observation matrix, the number of rows in the matrix is ​​equal to the total number of sampling points of the N virtual channels, and the number of columns is equal to the number of sampling points of the original partial discharge signal. Each element in the matrix represents the correspondence between a sampling point in the observation sequence and the original signal. The value of the joint observation matrix element is determined by the mapping relationship between the random sampling rate and the virtual channel allocation method.

[0038] Furthermore, the reconstructed partial discharge signal undergoes post-processing, specifically as follows:

[0039] After the compressed sensing reconstruction of the partial discharge signal is completed, a digital filter is used to filter the reconstructed partial discharge signal.

[0040] Feature extraction is performed on the filtered partial discharge signal to extract key parameters for characterizing the partial discharge characteristics, including discharge quantity, discharge frequency, discharge location, and discharge phase.

[0041] A virtual multi-channel compressed sensing sampling system for partial discharge signals, comprising:

[0042] The signal sampling module is configured as a single physical sampling channel to perform non-uniform sampling of the input partial discharge simulation signal at each randomly generated sampling time point by a pre-set non-uniform sampling rate, and convert the instantaneous amplitude of the partial discharge simulation signal into a digital signal to form a low-speed observation sequence.

[0043] The channel allocation module is configured to receive the observation sequence and allocate the sampling points in the sequence to the data buffers of N virtual channels in a cyclical manner according to the time order, forming N parallel sub-observation sequences;

[0044] The joint reconstruction module is configured to construct a joint observation matrix based on the compressed sensing reconstruction algorithm, according to the random sampling rate and the virtual channel allocation method. The joint observation matrix is ​​used to perform joint signal reconstruction on the N-way sub-observation sequences to recover the complete partial discharge digital signal.

[0045] The post-processing module is configured to perform filtering and feature parameter extraction on the reconstructed digital signal. The filtering process removes noise and interference, and the feature parameters characterizing partial discharge are extracted to provide a basis for the condition monitoring and fault diagnosis of power equipment.

[0046] Compared with the prior art, the beneficial effects of the present invention are:

[0047] 1. This invention constructs a virtual multi-channel based on a single physical sampling channel, which enables the acquisition of sufficient information for signal reconstruction at a lower sampling rate. This effectively reduces the requirements for hardware devices, not only reducing system cost and power consumption, but also solving the problem of excessive data volume in traditional high-speed sampling methods, making data storage, transmission and processing more efficient and convenient.

[0048] 2. This invention utilizes a joint observation matrix that matches the random sampling rate and virtual channel allocation method for compressed sensing reconstruction, making full use of the signal distribution information in multiple virtual channels. This improves the accuracy and reliability of signal reconstruction, especially when dealing with complex signals or situations where signal sparsity is not obvious. Compared with single-channel compressed sensing sampling, it can better recover the complete partial discharge digital signal, thus providing a more accurate basis for the condition monitoring and fault diagnosis of power equipment. Attached Figure Description

[0049] Figure 1 This is a flowchart of the virtual multi-channel compressed sensing sampling method for partial discharge signals according to the present invention;

[0050] Figure 2 This is a structural diagram of the virtual multi-channel compressed sensing sampling system for partial discharge signals according to the present invention. Detailed Implementation

[0051] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0052] To address the technical problems of high sampling cost, large data volume, and low signal reconstruction accuracy in existing partial discharge signal sampling methods, please refer to... Figures 1-2 This embodiment provides the following technical solution:

[0053] A virtual multi-channel compressed sensing sampling method for partial discharge signals, the method comprising:

[0054] A low-speed observation sequence is obtained by non-uniformly sampling the partial discharge simulation signal using a single physical sampling channel.

[0055] The observation sequence is sequentially and cyclically distributed to N virtual channels according to the time order, thereby constructing N parallel sub-observation sequences, where N is a positive integer greater than 1;

[0056] Based on the compressed sensing reconstruction algorithm, the N parallel sub-observation sequences are jointly reconstructed using a joint observation matrix that matches the random sampling rate and virtual channel allocation method, so as to recover the complete partial discharge digital signal in one go.

[0057] The reconstructed partial discharge signal is post-processed, including signal filtering and feature extraction, to extract the characteristic information of partial discharge.

[0058] The technical effects of the above solution are as follows: By using a single physical sampling channel for non-uniform sampling, the reliance on high-precision analog-to-digital converters and high-speed data storage devices found in traditional high-speed sampling methods is avoided, significantly reducing the cost of hardware devices and the power consumption of the power system. The observation sequence is sequentially and cyclically distributed to multiple virtual channels, and joint reconstruction is performed using a joint observation matrix based on a compressed sensing reconstruction algorithm. This fully utilizes the signal distribution information in multiple virtual channels, effectively integrating the dispersed sampling data. Thus, even at a low sampling rate, the complete partial discharge digital signal can be recovered in one go, improving the accuracy and efficiency of signal reconstruction. Furthermore, post-processing the reconstructed partial discharge signal can effectively extract the characteristic information of the partial discharge, providing accurate basis for the condition monitoring and fault diagnosis of power equipment. This helps to promptly detect potential equipment faults, thereby improving the operational safety and reliability of the power system.

[0059] Non-uniform sampling of the partial discharge simulation signal is performed using a single physical sampling channel, specifically as follows:

[0060] A partial discharge simulation signal is input into a single physical sampling channel, which consists of an analog-to-digital converter and its pre-amplifier signal conditioning circuit.

[0061] Before sampling, the random sampling rate for non-uniform sampling is preset based on the characteristics of the partial discharge signal (such as the highest frequency component, sparsity, etc.) and the required compression ratio. Let the highest frequency of the partial discharge signal be... The required compression ratio is It satisfies 0 < If the random sampling rate is less than 1, then the random sampling rate is... Calculated using the following formula:

[0062]

[0063] In the formula, This represents the compression ratio, used to control the sparsity of sampling points. This formula clearly expresses the mathematical relationship between the random sampling rate, the highest frequency of the signal, and the compression ratio, making the setting of the random sampling rate a reliable basis.

[0064] When the partial discharge simulation signal is input, the physical sampling channel samples the input partial discharge simulation signal at each randomly generated sampling time point using an analog-to-digital converter according to a pre-set random sampling rate, and converts the instantaneous amplitude of the simulation signal into a digital signal. Let the set of sampling time points be denoted as . , Let be the total number of sampling points. Then the sampling process can be represented as:

[0065]

[0066] In the formula, Indicates the first The digital signal value of each sampling point; this formula defines the relationship between each sampling point and the original signal, providing the mathematical basis for the input data of the subsequent virtual channel allocation and reconstruction algorithms;

[0067] Digital signals are collected and stored sequentially, forming a low-speed observation sequence;

[0068] The random sampling rate is determined based on the sparsity of the partial discharge signal and the highest frequency component of the signal. The random sampling rate determines the time interval of each sampling, and the time interval is randomly distributed to ensure that sufficient information is obtained for subsequent signal reconstruction under non-uniform sampling conditions.

[0069] The technical effects of the above solution are as follows: by achieving non-uniform sampling through a single physical sampling channel, the sampling rate and data volume can be significantly reduced while ensuring the signal reconstruction quality. It also effectively avoids the aliasing phenomenon that may be introduced by traditional uniform sampling. At the same time, the reasonable determination of the random sampling rate ensures that sufficient information can still be obtained for subsequent signal reconstruction at a low sampling rate, thereby improving the efficiency and flexibility of signal processing and providing high-quality observation sequences for subsequent virtual multi-channel allocation and compressed sensing reconstruction.

[0070] The observation sequence is sequentially and cyclically distributed to N virtual channels, specifically as follows:

[0071] The allocation of virtual channels is achieved using software algorithms. A loop counter is set with an initial value of 0. After each sampling point is allocated, the loop counter is incremented by 1. When the value of the loop counter reaches N, the loop counter is reset to zero and a new round of loop allocation begins.

[0072] The virtual channels are allocated in chronological order, with sampling points in the observation sequence being assigned to each virtual channel sequentially. Each virtual channel receives one sampling point in each allocation period, ensuring that N virtual channels receive and process the data in the observation sequence in parallel.

[0073] After obtaining a low-speed observation sequence, starting from the first sampling point of the observation sequence, each sampling point is sequentially assigned to the data buffer corresponding to the virtual channel indicated by the current loop counter. This process is repeated until all sampling points in the observation sequence have been assigned.

[0074] The allocation period is dynamically adjusted based on the pulse repetition frequency of the partial discharge signal, specifically as follows:

[0075] By performing continuous short-time Fourier transforms on the partial discharge signal, the pulse repetition frequency characteristics of the partial discharge signal in different time segments are obtained.

[0076] Based on the monitored pulse repetition frequency characteristics, the allocation period of the observation sequence to the virtual channel is dynamically adjusted;

[0077] When the signal pulse repetition frequency is high, shortening the allocation period allows the virtual channel to receive new sampled data more frequently, thereby improving the ability to capture high-frequency signal changes and the time resolution.

[0078] When the signal pulse repetition frequency is low, the allocation period is extended, the data update frequency of the virtual channel is reduced, unnecessary oversampling of low-frequency signals is avoided, and the computational and storage burden of the system is reduced.

[0079] Among them, the allocation cycle Based on the monitored pulse repetition frequency Nonlinear dynamic adjustment is performed, as shown in the following formula:

[0080]

[0081] In the formula, The base allocation period; and These are the expected minimum and maximum pulse repetition frequencies, used to define the effective adjustment range; The attenuation coefficient is controlled. Follow The rate at which growth declines; This is the gain coefficient; It is a very small positive number, used to prevent the denominator from being zero; when When height increases, Significantly reduced to achieve rapid sampling; when When it decreases, The process is smoothed out to conserve resources; the formula specifies the allocation period. How to monitor the pulse repetition frequency in real time Change enables allocation strategies to be adaptive;

[0082] Each virtual channel is allocated an independent data buffer to store the sampling point data assigned to that virtual channel. During the allocation process, the filling status of each virtual channel's data buffer is monitored in real time. When the data buffer of any virtual channel is filled to a predetermined length, compressed sensing reconstruction processing of the data in that data buffer is triggered.

[0083] The technical effects of the above solution are as follows: The virtual channel allocation mechanism implemented through software algorithms can sequentially allocate sampling points in the observation sequence to various virtual channels according to the time order, ensuring that N virtual channels receive and process data in parallel. This allocation method not only improves the efficiency of data processing, but also achieves adaptive sampling of signals of different frequencies by dynamically adjusting the allocation period based on the pulse repetition frequency characteristics of the partial discharge signal. This flexible allocation strategy can better adapt to different types of partial discharge signals, improving the flexibility and adaptability of signal processing, thereby providing a higher quality data foundation for subsequent signal reconstruction and feature extraction, and thus improving the accuracy and reliability of power equipment status monitoring. In addition, by allocating an independent data buffer for each virtual channel and monitoring its filling status in real time, efficient management and timely processing of sampled data are achieved. The independent data buffer ensures that the data of each virtual channel does not interfere with each other, maintaining the integrity and accuracy of the data, while the real-time monitoring mechanism enables the system to accurately grasp the data accumulation of each buffer, avoiding excessive data accumulation and reducing the occupation of storage resources.

[0084] Based on the compressed sensing reconstruction algorithm, the N parallel sub-observation sequences are jointly reconstructed using a joint observation matrix that matches the random sampling rate and virtual channel allocation method. Specifically:

[0085] A joint observation matrix is ​​constructed based on the previously set random sampling rate and virtual channel allocation method. The joint observation matrix reflects the distribution of the observation sequence in different virtual channels and its intrinsic relationship with the original partial discharge signal. The joint observation matrix can comprehensively reflect the sampling information of N virtual channels. It is a mathematical model used to integrate the scattered information obtained by non-uniform sampling so that the original signal can be accurately reconstructed later. The construction of the joint observation matrix is ​​as follows:

[0086]

[0087] in, Represents the joint observation matrix; Indicates the first The observation matrix of each virtual channel, each observation matrix reflecting the distribution of sampling points in that channel;

[0088] In constructing the joint observation matrix, the number of rows in the matrix is ​​equal to the total number of sampling points in the N virtual channels, and the number of columns is equal to the number of sampling points in the original partial discharge signal. Each element in the matrix represents the correspondence between a sampling point in the observation sequence and the original signal. The values ​​of the elements of the joint observation matrix are determined by the mapping relationship between the random sampling rate and the virtual channel allocation method. Furthermore, the joint observation matrix can be stored in a sparse matrix manner to improve computational efficiency.

[0089] The N parallel sub-observation sequences are used as input and substituted into the compressed sensing reconstruction algorithm. During the iteration process of the algorithm, the original partial discharge digital signal is gradually approximated by matrix operations using the determined joint observation matrix. This process will continuously adjust the estimated value of the signal to match the observation sequence as closely as possible under the constraints of the joint observation matrix. At the same time, the sparsity of the signal is taken into account to ensure that the recovered signal has high accuracy and reliability.

[0090] The iterative process includes an iteration termination condition, which is either the residual is less than a preset threshold or the number of iterations reaches a preset upper limit. When either termination condition is met, the iteration stops and the final partial discharge digital signal is output.

[0091] After multiple iterative calculations using the compressed sensing reconstruction algorithm, a complete digital signal of partial discharge is obtained in one go.

[0092] The technical effects of the above solution are as follows: By constructing a joint observation matrix that matches the random sampling rate and virtual channel allocation method, and using a compressed sensing reconstruction algorithm to jointly reconstruct N parallel sub-observation sequences, efficient and accurate reconstruction of partial discharge signals can be achieved. The constructed joint observation matrix can comprehensively reflect the sampling information of multiple virtual channels, integrating the scattered information obtained from non-uniform sampling, thus providing a solid data foundation for signal reconstruction. During the iterative process of the compressed sensing reconstruction algorithm, the original signal is gradually approximated through matrix operations, while considering the sparsity of the signal, ensuring the accuracy and reliability of the reconstructed signal. In addition, setting an iteration termination condition can stop the iteration in time when the accuracy requirements are met or the maximum number of iterations is reached, avoiding unnecessary calculations and further improving the efficiency of the system. Finally, a complete partial discharge digital signal is obtained in one go, providing a high-quality data foundation for subsequent signal analysis and feature extraction, significantly improving the performance and efficiency of partial discharge signal processing.

[0093] The reconstructed partial discharge signal undergoes post-processing, specifically as follows:

[0094] After the compressed sensing reconstruction of the partial discharge signal is completed, a digital filter is used to filter the reconstructed partial discharge signal. The filtering process is prior art in this field and is not an inventive solution of this application, so it will not be described in detail here.

[0095] Feature extraction is performed on the filtered partial discharge signal to extract key parameters characterizing the partial discharge characteristics. These parameters include discharge quantity, discharge frequency, discharge location, and discharge phase.

[0096] The amount of discharge can be extracted by analyzing the amplitude of the signal, finding the peak or pulse signal in the signal, and calculating its corresponding energy or amplitude.

[0097] The discharge frequency can be extracted by performing a fast Fourier transform on the signal to determine the distribution of the main frequency components in the signal.

[0098] The extraction of the discharge location usually requires combining the signal propagation characteristics and sensor layout information, and determining the location of the partial discharge by the signal arrival time difference;

[0099] Extracting the discharge phase requires synchronous analysis of the signal and the power supply voltage signal to determine the phase interval in which partial discharge occurs.

[0100] The technical effects of the above solution are as follows: using a digital filter to filter the reconstructed partial discharge signal can effectively remove noise and interference components from the partial discharge signal, improve the signal-to-noise ratio of the partial discharge signal, and make subsequent feature extraction more accurate and reliable. The subsequent feature extraction process extracts key parameters such as discharge quantity, discharge frequency, discharge location, and discharge phase from the filtered partial discharge signal. These parameters can provide reliable support for understanding the physical process of partial discharge, assessing insulation status, and locating fault locations.

[0101] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0102] Example 1

[0103] To verify the effectiveness of the method proposed in this invention, a partial discharge signal simulation platform was constructed to simulate a typical partial discharge pulse signal (pulse width approximately 100 ns, repetition frequency adjustable from 1 kHz to 10 kHz). Comparative experiments were conducted using the traditional uniform sampling method (Nyquist sampling rate) and the virtual multi-channel compressed sensing sampling method proposed in this invention.

[0104] Experimental conditions:

[0105] Signal frequency range: 20–40 MHz;

[0106] Traditional sampling rate: 80 MS / s (satisfies Nyquist theorem);

[0107] The sampling rate of this invention is 16 MS / s (compression rate of 20%).

[0108] Number of virtual channels: N=4;

[0109] Reconstruction Algorithm: Orthogonal Matching Pursuit.

[0110] index Traditional sampling methods Method of the present invention Sampling rate 80 MS / s 16 MS / s Reconstructed signal-to-noise ratio (SNR) Infinite (original signal) 30.2 dB Reconstruction Error (RMSE) 0 0.022 Data volume (MB / s) 80 MB 16 MB Number of hardware channels 1 (High-speed ADC) 1 (Low-speed ADC)

[0111] Table 1

[0112] As shown in Table 1, the method of the present invention can still achieve high-quality signal reconstruction with a signal-to-noise ratio of >30dB when the sampling rate is only 20% of that of the traditional method; the waveform characteristics of the reconstructed signal are highly consistent with the original signal, thereby ensuring that key information such as pulse position and amplitude can be accurately recovered; the amount of data is greatly reduced, which is beneficial for long-term online monitoring and remote transmission.

[0113] Specifically, this embodiment also proposes a virtual multi-channel compressed sensing sampling system for partial discharge signals. This system is used to implement a virtual multi-channel compressed sensing sampling method for partial discharge signals, including:

[0114] The signal sampling module is configured as a single physical sampling channel to perform non-uniform sampling of the input partial discharge simulation signal at each randomly generated sampling time point by a pre-set non-uniform sampling rate, and convert the instantaneous amplitude of the partial discharge simulation signal into a digital signal to form a low-speed observation sequence.

[0115] The channel allocation module is configured to receive the observation sequence and allocate the sampling points in the sequence to the data buffers of N virtual channels in a cyclical manner according to the time order, forming N parallel sub-observation sequences;

[0116] The joint reconstruction module is configured to construct a joint observation matrix based on the compressed sensing reconstruction algorithm, according to the random sampling rate and the virtual channel allocation method. The joint observation matrix is ​​used to perform joint signal reconstruction on the N-way sub-observation sequences to recover the complete partial discharge digital signal.

[0117] The post-processing module is configured to perform filtering and feature parameter extraction on the reconstructed digital signal. The filtering process removes noise and interference, and the feature parameters characterizing partial discharge are extracted to provide a basis for the condition monitoring and fault diagnosis of power equipment.

[0118] Working principle: First, the partial discharge simulation signal is non-uniformly sampled through a single physical sampling channel to obtain a low-speed observation sequence. This avoids the reliance on high-precision analog-to-digital converters and high-speed data storage devices found in traditional high-speed sampling methods. Second, a software algorithm is used to sequentially and cyclically distribute the observation sequence to multiple virtual channels, constructing multiple parallel sub-observation sequences. Based on a compressed sensing reconstruction algorithm, a joint observation matrix is ​​used for joint reconstruction. This fully utilizes the signal distribution information in multiple virtual channels and effectively integrates the scattered sampled data, thereby recovering the complete partial discharge digital signal in one go at a low sampling rate. This improves the accuracy and efficiency of signal reconstruction. Finally, post-processing operations such as filtering and feature extraction are performed on the reconstructed signal to extract key parameters characterizing the partial discharge characteristics, thus providing a basis for the condition monitoring and fault diagnosis of power equipment.

[0119] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0120] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

Claims

1. A method for virtual multi-channel compressed sensing sampling of partial discharge signals, characterized in that, The method comprises: non-uniformly sampling a partial discharge analog signal using a single physical sampling channel to obtain a low-speed observation sequence; sequentially and cyclically distributing the observation sequence to N virtual channels to obtain N sub-observation sequences; based on a compressed sensing reconstruction algorithm, using a joint observation matrix matched with the random sampling rate and the virtual channel distribution mode to jointly reconstruct the N parallel sub-observation sequences to recover a complete partial discharge digital signal; post-processing the reconstructed partial discharge signal, including signal filtering and feature extraction.

2. The method of claim 1, wherein, The non-uniform sampling of the partial discharge analog signal using a single physical sampling channel comprises: connecting the partial discharge analog signal to the single physical sampling channel, which is composed of an analog-to-digital converter and its pre-signal conditioning circuit; Before sampling, the random sampling rate of the non-uniform sampling is set according to the characteristics of the partial discharge signal and the required compression rate, wherein the highest frequency of the partial discharge signal is , the required compression rate is , and 0 <1, the random sampling rate is calculated by the following formula: ； In the formula, denotes the compression rate, which is used to control the sparsity degree of the sampling points; When the partial discharge analog signal starts to input, the physical sampling channel samples the input partial discharge analog signal at each randomly generated sampling time point through an analog-to-digital converter according to a pre-set random sampling rate, and converts the instantaneous amplitude of the analog signal into a digital signal, wherein a set of sampling time points is , is the total number of sampling points, and the sampling process can be represented as: ； wherein represents the digital signal value of the th sampling point; sequentially collecting and storing the digital signal to form a low-speed observation sequence.

3. The method of claim 2, wherein, The random sampling rate is determined according to the sparsity of the partial discharge signal and the highest frequency component of the signal. The random sampling rate determines the time interval of each sampling, and the time interval is randomly distributed to ensure that sufficient information is obtained for subsequent signal reconstruction under non-uniform sampling.

4. The method of claim 1, wherein, The sequential and cyclic distribution of the observation sequence to the N virtual channels comprises: using a software algorithm to realize the distribution of the virtual channels, setting a loop counter with an initial value of 0, incrementing the loop counter by 1 after each sampling point is distributed, and resetting the loop counter to zero when the value of the loop counter reaches N, and starting a new round of cyclic distribution; after obtaining the low-speed observation sequence, starting from the first sampling point of the observation sequence, sequentially distributing each sampling point to the data buffer area corresponding to the virtual channel indicated by the current loop counter, and repeating the cycle until all sampling points in the observation sequence are distributed.

5. The method of claim 4, wherein, The distribution period is dynamically adjusted according to the pulse repetition frequency of the partial discharge signal, specifically: by performing continuous short-time Fourier transform on the partial discharge signal, obtaining the pulse repetition frequency characteristics of the partial discharge signal in different time segments; based on the monitored pulse repetition frequency characteristics, dynamically adjusting the distribution period of the observation sequence to the virtual channels; when the signal pulse repetition frequency is high, shorten the distribution period to make the virtual channels receive new sampling data more frequently; when the signal pulse repetition frequency is low, extend the distribution period to reduce the data update frequency of the virtual channels; wherein the allocation period According to the monitored pulse repetition frequency A non-linear dynamic adjustment is made, as follows: ； wherein is the reference distribution period; and are the minimum and maximum expected pulse repetition frequencies, respectively, defining the effective adjustment range; is the decay coefficient, controlling the decay rate of the increase of ; and is the gain coefficient; is a small positive number, preventing the denominator from being zero; when increases, it decreases significantly; when decreases, it increases smoothly.

6. The method of claim 4, wherein, assigning an independent data buffer area to each virtual channel for storing the sampling point data distributed to the virtual channel, and monitoring the filling state of the data buffer area of each virtual channel in real time. When the data buffer area of any virtual channel is filled to a predetermined length, the data in the data buffer area is triggered for compressed sensing reconstruction processing.

7. The method of claim 1, wherein, based on the compressed sensing reconstruction algorithm, using a joint observation matrix matched with the random sampling rate and the virtual channel distribution mode to jointly reconstruct the N parallel sub-observation sequences, specifically: According to the random sampling rate and the allocation mode of the virtual channel, a joint observation matrix is constructed, which is used to reflect the distribution of the observation sequence in different virtual channels and the internal relationship with the original partial discharge signal. The construction of the joint observation matrix is as follows: ； wherein, represents a joint observation matrix; represents an observation matrix of the virtual channel, each observation matrix reflecting the sampling point distribution of the channel; The N parallel sub-observation sequences are taken as input and substituted into the compressed sensing reconstruction algorithm. In the iteration process of the algorithm, the joint observation matrix is determined, and the original partial discharge digital signal is gradually approximated through matrix operation. After multiple iterations of the compressed sensing reconstruction algorithm, the complete partial discharge digital signal is obtained at one time. The iteration formula of the compressed sensing reconstruction algorithm is as follows.

8. The method of claim 7, wherein, In the construction of the joint observation matrix, the number of rows of the matrix is equal to the total number of sampling points of the N virtual channels, and the number of columns is equal to the number of sampling points of the original partial discharge signal. Each element in the matrix represents the corresponding relationship between a sampling point in the observation sequence and the original signal. The value of the element in the joint observation matrix is determined by the mapping relationship between the random sampling rate and the virtual channel allocation mode.

9. The method of claim 1, wherein, The reconstructed partial discharge signal is post-processed, specifically: After the compressed sensing reconstruction of the partial discharge signal is completed, a digital filter is used to filter the reconstructed partial discharge signal; The filtered partial discharge signal is feature extracted, and the key parameters for representing the characteristics of the partial discharge are extracted from the partial discharge signal. The parameters include discharge amount, discharge frequency, discharge position, and discharge phase.

10. A virtual multi-channel compressed sensing sampling system of partial discharge signals, characterized in that, The system is used to implement a virtual multi-channel compressed sensing sampling method for partial discharge signals as claimed in any one of claims 1-9, comprising: A signal sampling module is configured to perform non-uniform sampling on the input partial discharge analog signal at each randomly generated sampling time point through a pre-set random sampling rate of non-uniform sampling in a single physical sampling channel, and convert the instantaneous amplitude of the partial discharge analog signal into a digital signal to form a low-speed observation sequence; A channel allocation module is configured to receive the observation sequence and sequentially allocate the sampling points in the sequence to the data buffer area of the N virtual channels in a time-sequential cyclic allocation mode to form N parallel sub-observation sequences; A joint reconstruction module is configured to construct a joint observation matrix based on the compressed sensing reconstruction algorithm according to the random sampling rate and the virtual channel allocation mode, and perform joint signal reconstruction on the N parallel sub-observation sequences through the joint observation matrix to restore the complete partial discharge digital signal. A post-processing module is configured to perform filtering and feature parameter extraction on the reconstructed digital signal. The filtering process removes noise and interference, and the feature parameters representing the characteristics of the partial discharge are extracted to provide a basis for the state monitoring and fault diagnosis of the power equipment.

Images