Phase locking method and system for local place wireless same frequency based on big data intelligent analysis

Abstract

The application provides a partial discharge wireless same-frequency phase locking method and system fusing big data intelligent analysis, which comprises the following steps: obtaining a standard phase angle standard vector from the cloud based on the type of a partial discharge detector; generating a two-dimensional partial discharge matrix from the obtained discrete data and projecting the two-dimensional partial discharge matrix to a current phase angle density vector; calculating a cross power spectrum in a frequency domain to determine an original phase slip; evaluating a network time delay variance to match target filtering parameters, and removing network time delay phase offsets through a tracking filter to output a predicted phase compensation step; performing reverse cyclic translation on the two-dimensional partial discharge matrix based on the step to lock to a same-frequency standard phase domain; and extracting a predicted residual error to adjust a color rendering threshold rule and performing cumulative rendering on a mobile terminal. The application effectively eliminates the influence of wireless network transmission time delay on partial discharge phase synchronization, and realizes high-precision same-frequency phase locking and clear rendering display.

Description

Technical Field

[0001] This invention belongs to the field of partial discharge detection technology, specifically relating to a partial discharge wireless phase-locking method and system that integrates big data intelligent analysis. Background Technology

[0002] Partial discharge detection is a core method for assessing the insulation status of power equipment. With the development of mobile internet and intelligent detection technology, using mobile terminals to connect to partial discharge detectors via wireless networks for on-site data reading and real-time spectrum display has become the mainstream application mode. However, in power plants or industrial sites, wireless communication links are easily affected by strong electromagnetic interference and multipath effects. The plotting of partial discharge phase-resolved maps depends on the synchronization of the discharge signal with the power grid's power frequency phase. When the partial discharge detector transmits high-density discrete partial discharge data to the mobile terminal via wireless network, multi-user contention and network fluctuations in the wireless channel will generate network latency, causing a shift in the time when the mobile terminal receives the data. This results in a misalignment of the originally ordered discharge phase sequence on the time axis. If this phase slip caused by dynamic network latency cannot be effectively identified and eliminated, the spectrum displayed by the mobile terminal will show phase distribution divergence and waveform distortion, ultimately leading to inaccurate diagnostic conclusions. Summary of the Invention

[0003] This invention provides a partial discharge wireless co-frequency phase locking method and system that integrates big data intelligent analysis to solve the above-mentioned technical problems.

[0004] In a first aspect, the present invention provides a partial discharge wireless co-frequency phase locking method integrating big data intelligent analysis, the method comprising the following steps: When a mobile terminal establishes a wireless communication connection with a partial discharge detector, the standard phase angle standard vector is obtained from the cloud big data platform based on the device type of the partial discharge detector. The standard phase angle standard vector is a standard vector representing the absolute phase probability distribution generated by clustering historical partial discharge data. Discrete data points of the partial discharge phase-resolved spectrum containing the original discharge information are obtained from the partial discharge detector, and a two-dimensional partial discharge matrix of the discharge state is generated based on the phase interval and amplitude interval according to the discrete data points. The two-dimensional partial discharge matrix is ​​reduced in dimension by projection and accumulation along the amplitude dimension, generating the current phase angle density vector representing the current discharge density. Calculate the cross-power spectrum between the current phase angle density vector and the standard phase angle standard vector in the frequency domain space, and determine the original phase slip based on the peak value of the cross-power spectrum; The network delay variance during wireless communication is evaluated to determine the current network congestion status, and the target filtering parameters are matched from the preset filtering coefficient mapping table of the cloud big data platform based on the current network congestion status. The original phase shift is input into the tracking filter configured with the target filtering parameters to remove the phase shift caused by wireless network latency and output the predicted phase compensation step size. Based on the predicted phase compensation step size, a reverse cyclic translation operation is performed on the discrete data points in the two-dimensional partial discharge matrix to align and lock the offset discrete data points into the same frequency standard phase domain consistent with the power grid frequency, thus obtaining the same frequency locked discrete data points. Extract the current prediction residual of the tracking filter, adjust the scatter point overlap color threshold rule in the graphics rendering component according to the current prediction residual, and perform cumulative rendering display of the discrete data points after same frequency locking on the mobile terminal interface according to the adjusted scatter point overlap color threshold rule.

[0005] Optionally, the step of projecting and accumulating the two-dimensional partial discharge matrix along the amplitude dimension to reduce its dimensionality and generate the current phase angle density vector representing the current discharge density includes the following steps: Extract the amplitude data of all amplitude intervals corresponding to each phase interval in the two-dimensional partial discharge matrix; Calculate the first local variance of the amplitude data for all amplitude intervals, and construct a weighted projection function to suppress background noise based on the first local variance; The amplitude data of all amplitude intervals are input into the weighted projection function for weighted summation calculation to obtain the aggregate density value of the corresponding phase interval; The aggregated density values ​​of all phase intervals are concatenated into the current phase angle density vector according to the order of the phase intervals.

[0006] Optionally, constructing the weighted projection function for suppressing background noise based on the first local variance includes the following steps: Standard noise variance determined when acquiring historical blank environmental data collected by a partial discharge detector; Calculate the variance deviation between the first local variance and the standard noise variance. If the variance deviation is lower than the preset signal deviation threshold, the corresponding amplitude data is determined to be pure background noise, and the corresponding weighting coefficient is assigned to zero in the weighted projection function. If the variance deviation is not lower than the signal deviation threshold, the variance deviation is mapped to a continuous penalty coefficient using the exponential decay model. The continuous penalty coefficients are used as weighting coefficients for the corresponding amplitude data in the weighted projection function to construct the weighted projection function.

[0007] Optionally, the standard noise variance determined when acquiring historical blank environmental data collected by the partial discharge detector includes the following steps: When the partial discharge detector is not connected to the device under test, send a blank acquisition command to the partial discharge detector; The partial discharge detector continuously receives blank data points for a preset duration in response to the blank acquisition command, and generates a historical blank observation matrix based on the blank data points. Randomly sample the historical blank observation matrix according to the amplitude dimension to obtain multiple noise sample sequences; Calculate the independent variance of each noise sample sequence separately; Gaussian fitting is performed on all independent variance values ​​to obtain the expected value, and the expected value is determined as the standard noise variance.

[0008] Optionally, the step of calculating the cross-power spectrum of the current phase angle density vector and the standard phase angle standard vector in the frequency domain space, and determining the original phase slip based on the peak value of the cross-power spectrum, includes the following steps: Perform a Fast Discrete Fourier Transform on the current phase density vector to obtain the current complex sequence in the frequency domain; Perform a Fast Discrete Fourier Transform on the standard phase angle standard vector to obtain a standard frequency domain complex sequence; A conjugate sequence of complex numbers in the frequency domain is obtained by performing a conjugate operation on a standard frequency domain complex sequence. Multiply the current frequency domain complex sequence element-wise with the conjugate frequency domain complex sequence to generate a cross-frequency domain matrix; Normalizing the cross-frequency domain matrix yields a smooth cross-frequency domain matrix; Perform an inverse fast discrete Fourier transform on the smooth cross-frequency domain matrix to generate a cross-power spectrum sequence in the time domain, and obtain the original phase shift by peak finding in the cross-power spectrum sequence.

[0009] Optionally, obtaining the original phase shift by peak finding in the cross-power spectrum sequence includes the following steps: Calculate the global mean and global standard deviation of all elements in the cross power spectrum sequence, and use the global mean and global standard deviation to set the dynamic peak finding threshold; Select a set of candidate peak points whose values ​​are greater than the dynamic peak-finding threshold from the cross-power spectrum sequence; Calculate the local second derivative of each candidate peak point in the candidate peak point set, and select the candidate peak point with the largest local second derivative as the absolute main peak; Extract the array index subscript corresponding to the absolute main peak in the cross-power spectrum sequence, convert the array index subscript into the physical phase angle difference, and use the physical phase angle difference as the original phase shift.

[0010] Optionally, the step of evaluating the network delay variance during wireless communication to determine the current network congestion status, and matching the target filtering parameters from the preset filtering coefficient mapping table of the cloud big data platform based on the current network congestion status, includes the following steps: Record the local transmission time when sending a data acquisition request to the partial discharge detector, and record the local reception time when receiving the response data packet; Calculate the difference between the local reception time and the local transmission time to obtain the round-trip network delay. The single round-trip network latency within multiple consecutive polling cycles is stored in a sliding time window, and the latency variance data of all single round-trip network latencies within the sliding time window are calculated. The time delay variance data is input into a preset network state classifier to classify the corresponding current network congestion state, and the corresponding target filtering parameters are retrieved from the preset filtering coefficient mapping table of the cloud big data platform according to the current network congestion state.

[0011] Optionally, the step of inputting the original phase shift into a tracking filter configured with target filtering parameters, eliminating the phase shift caused by wireless network latency, and outputting the predicted phase compensation step size includes the following steps: The original phase shift is input into a tracking filter configured with target filtering parameters, and the target filtering parameters are resolved into observation gain coefficients for position updates and tracking gain coefficients for velocity updates. Calculate the position residual between the prior phase estimate and the original phase slip for the current prediction period; The posterior phase state of the tracking filter is updated using the observation gain coefficient and the position residual, and the slip velocity state of the tracking filter is updated using the tracking gain coefficient and the position residual. Based on the posterior phase state quantity and the updated slip velocity state quantity, the prior phase estimate for the next prediction period is calculated. The updated slip velocity state is extracted as the prediction phase compensation step size.

[0012] In a second aspect, the present invention also provides a partial discharge wireless co-frequency phase locking system integrating big data intelligent analysis, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the partial discharge wireless co-frequency phase locking method integrating big data intelligent analysis as described in any one of the first aspects.

[0013] Thirdly, the present invention also provides a computer-readable storage medium storing instructions that, when executed by a processor, configure the processor to perform the partial discharge wireless co-frequency phase locking method according to any one of the first aspects.

[0014] The beneficial effects of this invention are: Based on the device type of the partial discharge (PD) detector, a standard phase angle vector from a cloud-based big data platform is introduced, utilizing the probabilistic characteristics of historical data to provide a phase reference. Then, by projecting the two-dimensional PD matrix to reduce its dimensionality to a phase angle density vector and performing cross-power spectrum calculation with the standard vector in the frequency domain, the original phase shift caused by wireless transmission delay or device differences can be quickly identified, achieving initial phase alignment in the absence of a reference source. Next, addressing the instability of wireless network transmission, congestion status is determined by evaluating network delay variance, and filter parameters are dynamically matched. The original phase shift is smoothed by a tracking filter configured with target filter parameters, eliminating phase noise caused by network delay jitter and ensuring the prediction accuracy of the phase compensation step size. Using the predicted phase compensation step size, a reverse cyclic translation operation is performed on the two-dimensional PD matrix, aligning and locking the shifted points to the same-frequency standard phase domain, thereby eliminating the impact of wireless transmission delay on phase synchronization and significantly improving the accuracy of phase locking. Finally, the prediction residuals of the tracking filter are extracted to dynamically adjust the scatter point overlap color rendering threshold rules, and adaptive cumulative rendering display is achieved on the mobile terminal interface. This effectively overcomes image distortion caused by network fluctuations and ensures the clarity and diagnostic accuracy of partial discharge spectra in wireless environments. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating a partial discharge wireless phase-locking method integrating big data intelligent analysis in one embodiment of this application.

[0016] Figure 2 This is a schematic diagram of the resolution spectrum of the partial discharge phase in one embodiment of this application. Detailed Implementation

[0017] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0018] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, a first object can be one or more. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0019] Figure 1 This is a flowchart illustrating a partial discharge wireless phase-locking method integrating big data intelligent analysis in one embodiment. It should be understood that, although... Figure 1 The steps in the flowchart are shown sequentially as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order in which these steps are executed, and they can be performed in other orders. Figure 1 At least some steps in the process may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed alternately or in turn with other steps or at least a portion of the sub-steps or stages of other steps. For example Figure 1 As shown, the partial discharge wireless co-frequency phase locking method integrating big data intelligent analysis disclosed in this invention specifically includes the following steps: S101. When the mobile terminal establishes a wireless communication connection with the partial discharge detector, the standard phase angle standard vector is obtained from the cloud big data platform based on the device type of the partial discharge detector. The standard phase angle standard vector is a standard vector representing the absolute phase probability distribution generated by clustering historical partial discharge data.

[0020] The system automatically identifies and acquires the physical attributes of the partial discharge detector, such as its model, hardware version, and internal sensor configuration, to determine its specific device type. Then, using this identified device type as an index, a data retrieval request is sent via a wireless network to a pre-built cloud-based big data platform. This cloud-based big data platform pre-stores big data analysis results for various device types. These results are generated by collecting massive amounts of historical partial discharge data generated during the historical operation of each specific device type. This data undergoes feature cleaning, dimensionality reduction, and clustering analysis to eliminate random high-frequency noise interference, generating an absolute phase probability distribution that characterizes the mapping relationship between the probability of partial discharge signal occurrence and the absolute phase angle within a power frequency voltage cycle. This absolute phase probability distribution is ultimately stored as a one-dimensional numerical vector, defined as a standard phase angle vector. Each element in this standard phase angle vector corresponds to the expected probability of partial discharge occurring in a specific subdivided phase interval within a power frequency voltage cycle. This standard phase angle vector is then securely transmitted from the cloud-based big data platform and downloaded to the temporary storage medium of the mobile terminal.

[0021] S102. Obtain discrete data points of the partial discharge phase-resolved spectrum containing the original discharge information from the partial discharge detector, and generate a two-dimensional partial discharge matrix of the discharge state based on the phase interval and amplitude interval according to the discrete data points.

[0022] The partial discharge detector receives discrete data points containing the original discharge information, representing the phase-resolved spectrum of partial discharge, in real-time or periodically, via an established wireless communication connection. Each discrete data point contains at least two dimensions of feature information: the phase value at the time of discharge and the corresponding discharge amplitude. These discrete data points collectively reflect the transient characteristics of the partial discharge within the tested electrical equipment. To structure these disordered discrete data points, a complete cycle of the power frequency voltage (from 0 degrees to 360 degrees) is first divided into several equally wide phase intervals, for example, 360 phase intervals. Simultaneously, the measurement dynamic range of the partial discharge detector (from the minimum discharge amplitude to the maximum discharge amplitude) is divided into several equally wide amplitude intervals, for example, 100 amplitude intervals. Based on these division rules, a two-dimensional grid structure is constructed, with the number of rows corresponding to the number of amplitude intervals and the number of columns corresponding to the number of phase intervals. Then, all received discrete data points are traversed. Based on the actual phase value and actual amplitude carried by each discrete data point, the row and column indices corresponding to that discrete data point in the two-dimensional grid structure are determined, and the count values ​​corresponding to those indices are accumulated. After completing the classification and statistics of all discrete data points, the two-dimensional grid structure is transformed into a two-dimensional partial discharge matrix that characterizes the discharge state distribution.

[0023] S103. Project and accumulate the two-dimensional partial discharge matrix along the amplitude dimension to reduce its dimensionality, and generate the current phase angle density vector representing the current discharge density.

[0024] The process involves extracting amplitude data from each phase interval of a two-dimensional partial discharge matrix, specifically from all amplitude intervals corresponding to each column. For each column of amplitude data, a first local variance is calculated, which sensitively reflects the dispersion of the discharge signal within the current phase interval. Based on this first local variance, a weighted projection function is constructed to suppress background noise. This function dynamically allocates weights according to the discrete characteristics of the amplitude data, giving larger projection weights to amplitude data containing the actual discharge signal and extremely low or even zero weights to amplitude data containing random white noise or background noise. Subsequently, the amplitude data of all amplitude intervals corresponding to each phase interval are input into the weighted projection function for weighted summation, thereby compressing the multi-dimensional amplitude information of the phase interval into a single aggregated density value. Finally, the aggregated density values ​​calculated for all phase intervals are sequentially concatenated according to the ascending order of phase intervals from zero to 360 degrees, ultimately generating a current phase angle density vector representing the current discharge density and excluding noise interference.

[0025] S104. Calculate the cross-power spectrum of the current phase angle density vector and the standard phase angle standard vector in the frequency domain space, and determine the original phase slip based on the peak value of the cross-power spectrum.

[0026] First, a baseline standard noise variance is determined. This variance is obtained by sending a blank acquisition command to the partial discharge (PD) detector when it is in an unloaded state (no device connected to the test device). The PD detector responds to this command by continuously receiving a large number of blank data points within a preset time period and uses these data points to construct a historical blank observation matrix. Next, the historical blank observation matrix is ​​randomly sampled multiple times along the amplitude dimension to obtain multiple independent noise sample sequences, and the independent variance value of each noise sample sequence is calculated. These independent variance values ​​are then Gaussian fitted, and the expected value obtained after fitting is used as the standard noise variance. After obtaining the standard noise variance, the variance deviation between the first local variance of each phase interval obtained in the previous steps and the standard noise variance is calculated. If the variance deviation is lower than a preset signal deviation threshold, it indicates that the fluctuation in that phase interval is mainly caused by background noise. Therefore, the corresponding amplitude data is determined to be pure background noise, and the corresponding weighting coefficient in the weighted projection function is directly assigned to zero. Conversely, if the variance deviation is not lower than the signal deviation threshold, the variance deviation is mapped to a continuous penalty coefficient using an exponential decay model, and this continuous penalty coefficient is used as the corresponding weighting coefficient to construct the final weighted projection function.

[0027] S105. Evaluate the network delay variance during wireless communication to determine the current network congestion status, and match the target filtering parameters from the preset filtering coefficient mapping table of the cloud big data platform based on the current network congestion status.

[0028] The process involves obtaining the current phase density vector and the standard phase vector, then calculating their cross-power spectrum in the frequency domain to determine the original phase slip. First, a Fast Discrete Fourier Transform (FFT) is performed on the current phase density vector to obtain a current frequency domain complex sequence; simultaneously, a FFT is performed on the standard phase vector to obtain a standard frequency domain complex sequence. Then, each element in the standard frequency domain complex sequence is conjugated to obtain the corresponding conjugate frequency domain complex sequence. Next, the current frequency domain complex sequence and the conjugate frequency domain complex sequence are multiplied element-wise to generate a cross-frequency domain matrix containing phase difference information. To eliminate the impact of amplitude fluctuations on phase estimation accuracy, the cross-frequency domain matrix is ​​normalized to obtain a smoothed cross-frequency domain matrix. Finally, an Inverse Fast Discrete Fourier Transform (IFFT) is performed on the smoothed cross-frequency domain matrix to reconstruct the cross-power spectrum sequence in the time domain. To find the accurate peak in the cross-power spectrum sequence, the global mean and global standard deviation of all elements in the sequence are first calculated, and a dynamic peak-finding threshold is set using the sum of these two values. A set of candidate peak points with values ​​greater than this dynamic peak-finding threshold is then selected from the sequence, and the local second derivative of each candidate peak point is calculated. The candidate peak point with the largest local second derivative is selected as the absolute dominant peak, and its array index is extracted. Finally, this array index is converted into the physical phase angle difference.

[0029] S106. Input the original phase shift into the tracking filter configured with the target filtering parameters, remove the phase shift caused by the wireless network delay, and output the predicted phase compensation step size.

[0030] The process involves recording the local transmission timestamp when sending a data acquisition request to the partial discharge detector, and immediately recording the local reception timestamp upon receiving a response data packet from the detector. The difference between the local reception and transmission timestamps is used to calculate the round-trip network latency for the current interaction period. The calculated round-trip network latency over multiple consecutive polling periods is then sequentially stored in a sliding time window of fixed length, and the latency variance of all round-trip network latency data stored within this window is calculated. This latency variance accurately reflects the severity of network jitter. Next, this latency variance data is input into a pre-trained and deployed network state classifier. Classification calculations determine the current network congestion state, which can be categorized as, for example, uncongested, lightly congested, moderately congested, and severely congested. Finally, based on the identified current network congestion state, a matching filter coefficient mapping table pre-set on the cloud-based big data platform is retrieved via the wireless network to obtain the target filter parameters that best match the current network environment and balance filtering latency and tracking accuracy.

[0031] S107. Based on the predicted phase compensation step size, perform a reverse cyclic translation operation on the discrete data points in the two-dimensional partial discharge matrix to align and lock the offset discrete data points into the same frequency standard phase domain consistent with the power grid frequency, and obtain the same frequency locked discrete data points.

[0032] In this process, after obtaining the target filtering parameters, the calculated original phase shift is input into a tracking filter configured with these parameters to eliminate spurious phase shifts caused by wireless network latency fluctuations. Specifically, the target filtering parameters are resolved into observation gain coefficients for position updates and tracking gain coefficients for velocity updates. Next, the position residual between the prior phase estimate for the current prediction period and the input original phase shift is calculated. The posterior phase state of the tracking filter is updated using the observation gain coefficients and the position residual, and simultaneously, the slip velocity state of the tracking filter is updated using the tracking gain coefficients and the position residual. Based on the updated posterior phase state and slip velocity state, the prior phase estimate for the next prediction period is calculated. The updated slip velocity state is then extracted as the prediction phase compensation step size. Subsequently, this prediction phase compensation step size is used to perform a reverse cyclic translation operation on each discrete data point in the aforementioned constructed two-dimensional partial discharge matrix. This translation operation is achieved by subtracting the predicted phase compensation step size from the phase coordinate value of each discrete data point and performing a modulo operation on the power frequency period. This allows the offset discrete data points to be precisely aligned and locked into the same frequency standard phase domain that is strictly consistent with the actual power grid frequency, thus obtaining the discrete data points after the same frequency locking is completed.

[0033] S108. Extract the current prediction residual of the tracking filter, adjust the scatter point overlap color threshold rule in the graphics rendering component according to the current prediction residual, and perform cumulative rendering display of the discrete data points after frequency locking on the mobile terminal interface according to the adjusted scatter point overlap color threshold rule.

[0034] After completing the same-frequency phase locking, to ensure that the image presented on the mobile terminal interface more realistically reflects the intensity of partial discharge source activity, dynamic optimization of the display effect is required. First, the current prediction residual generated by the tracking filter within the current prediction period is extracted. The magnitude of this current prediction residual directly reflects the accuracy and stability of the current phase tracking; that is, the smaller the residual, the more accurate the tracking and the higher the data reliability. Based on the value of this current prediction residual, the scattered point overlap color threshold rule in the mobile terminal's graphics rendering component is dynamically adjusted. Specifically, when the current prediction residual is small, it indicates extremely high phase alignment accuracy. In this case, lowering the scattered point overlap color threshold allows a smaller number of overlapping discrete data points to trigger high-brightness or high-saturation color display, thereby highlighting the concentrated area of ​​partial discharge. Conversely, when the current prediction residual is large, it indicates network fluctuations causing tracking jitter. In this case, increasing the scattered point overlap color threshold avoids visual divergence or misleading effects caused by incomplete phase alignment. Finally, the discrete data points processed by the same-frequency locking are input into the graphics rendering component according to the adjusted scattered point overlap color threshold rule, such as... Figure 2 As shown, the process involves frame-by-frame cumulative rendering on the user interface of a mobile terminal to create a clear, stable, and accurate two-dimensional partial discharge phase-resolved image that accurately indicates the severity of partial discharge.

[0035] In one implementation, the method of projecting and accumulating the two-dimensional partial discharge matrix along the amplitude dimension to reduce its dimensionality and generate a current phase angle density vector representing the current discharge density includes the following steps: Extract the amplitude data of all amplitude intervals corresponding to each phase interval in the two-dimensional partial discharge matrix; Calculate the first local variance of the amplitude data for all amplitude intervals, and construct a weighted projection function to suppress background noise based on the first local variance; The amplitude data of all amplitude intervals are input into the weighted projection function for weighted summation calculation to obtain the aggregate density value of the corresponding phase interval; The aggregated density values ​​of all phase intervals are concatenated into the current phase angle density vector according to the order of the phase intervals.

[0036] In this embodiment, during the specific extraction process, the constructed two-dimensional partial discharge matrix is ​​represented as a matrix. The number of rows in this matrix corresponds to the total number of amplitude intervals divided into regions, denoted as . The number of columns in this matrix corresponds to the total number of phase intervals divided, denoted as . In the actual extraction operation, the data is traversed sequentially from left to right, i.e., from the first column to the last column, according to the column index. For any given column... Each phase interval is read and copied from the two-dimensional partial discharge matrix. The Middle This involves extracting all elements from the column, thus fully retrieving the amplitude distribution data for the phase interval across all amplitude intervals. The extracted amplitude distribution data is mathematically represented as a one-dimensional column vector. The specific vector expression is as follows: In the formula, Indicates the first Amplitude data vector corresponding to each phase interval Represents the second partial discharge matrix in a two-dimensional matrix. line, number The matrix elements of the column, that is, the first column... The phase interval and in the phase interval The discharge point count value within each amplitude range, where the subscript The corresponding range is from 1 to Positive integers, subscript The corresponding range is from 1 to Positive integers. This extraction operation decouples continuous two-dimensional grid data into components that are independent of each other on the phase axis. A one-dimensional amplitude distribution sequence.

[0037] After extracting the amplitude data for each phase interval, a denoising model needs to be constructed by calculating local statistical features. In the specific calculation process, for each phase interval, the first local variance of the amplitude data for all corresponding amplitude intervals is calculated. This first local variance measures the fluctuation and dispersion of the discharge signal intensity at a specific phase. If a phase interval contains only uniformly distributed white noise, the calculated first local variance is usually small; if the phase interval contains real partial discharge pulse signals, the first local variance will significantly increase due to the instantaneous changes in discharge amplitude. The calculation of the... The first local variance of each phase interval The formula is as follows: In the formula, Indicates the first The first local variance corresponding to each phase interval Indicates the first The arithmetic mean of all amplitude data elements within each phase interval is calculated. After obtaining the first local variance for all phase intervals, a weighted projection function for suppressing background noise is constructed based on this first local variance. This weighted projection function assigns non-negative weighting coefficients to different amplitude data, ensuring that the amplitude containing the real discharge signal is preserved during projection, while the amplitude of the background noise component is attenuated. When constructing this weighted projection function, the first local variance is used as the independent variable to calculate the arithmetic mean of all amplitude data elements within each phase interval. The nth phase interval Weighting coefficients corresponding to each amplitude data point This leads to the formation of a weighted projection function with noise suppression capabilities, which greatly improves the overall anti-interference performance of subsequent phase alignment calculations.

[0038] After constructing a weighted projection function with background noise suppression capabilities, the amplitude data of all amplitude intervals corresponding to each phase interval are input into the corresponding weighted projection function. Dimensionality is compressed by performing weighted summation calculation. Since the behavior of partial discharge at a specific phase is determined by the discharge frequencies of multiple amplitude intervals, traditional direct summation projection easily accumulates the energy of background white noise, thus obscuring the true phase characteristics. Through weighted summation, the weighted projection function assigns extremely small weighting coefficients to amplitude regions determined as noise, and larger weighting coefficients to amplitude regions determined as true discharge, thereby compressing the multi-dimensional amplitude information of the phase interval into a single aggregate density value that truly represents the partial discharge intensity. The calculation of the... Polymer density values ​​for each phase interval The formula is as follows: In the formula, Indicates the first The polymerization density value was calculated from each phase interval. This indicates that the weighted projection function constructed in the preceding steps corresponds to the first... The phase interval, the first Weighting coefficients for each amplitude interval. By applying all... The amplitude data for each amplitude interval are multiplied element-wise with their corresponding weighting coefficients and then summed to obtain a highly physically representative value that eliminates random background noise interference. This weighted summation calculation is repeated for each column vector to obtain the aggregated density value for all phase intervals, thus achieving the goal of significantly reducing data dimensionality while preserving phase characteristics.

[0039] After calculating the aggregated density values ​​for all phase intervals sequentially, these discrete values ​​are reorganized into a one-dimensional vector to facilitate subsequent phase sliding estimation in the frequency domain. During the stitching process, the aggregated density values ​​of all phase intervals are stitched together sequentially, from zero degrees to 360 degrees within the power frequency voltage cycle, in ascending order. This stitching method, based on physical angle order, ensures that the resulting one-dimensional vector completely preserves the original partial discharge phase distribution in both the time and spatial domains. The current phase angle density vector obtained through stitching is represented as a one-dimensional vector. The specific vector concatenation formula is as follows: In the formula, This represents the generated current phase angle density vector, used to characterize the discharge density distribution of the partial discharge signal acquired by the current wireless transmission throughout the entire power frequency cycle; Indicates the first The polymerization density values ​​for each phase interval, and the subscripts The value range is from 1 to An increasing positive integer. The current phase angle density vector. The length is equal to the total number of phase intervals divided. By concatenating all aggregated density values ​​in phase order, the complex two-dimensional partial discharge matrix was successfully compressed into a one-dimensional current phase angle density vector that can accurately characterize the phase distribution probability. This not only greatly reduces the amount of data processing required for subsequent cross-power spectrum calculations, but also suppresses random background noise through the preceding weighted projection, significantly improving the stability of phase slip calculations.

[0040] In one implementation, constructing a weighted projection function for suppressing background noise based on a first local variance includes the following steps: Standard noise variance determined when acquiring historical blank environmental data collected by a partial discharge detector; Calculate the variance deviation between the first local variance and the standard noise variance. If the variance deviation is lower than the preset signal deviation threshold, the corresponding amplitude data is determined to be pure background noise, and the corresponding weighting coefficient is assigned to zero in the weighted projection function. If the variance deviation is not lower than the signal deviation threshold, the variance deviation is mapped to a continuous penalty coefficient using the exponential decay model. The continuous penalty coefficients are used as weighting coefficients for the corresponding amplitude data in the weighted projection function to construct the weighted projection function.

[0041] In this embodiment, when the partial discharge detector is not connected to any charged device under test and is in a purely external electromagnetic blank environment, continuous data acquisition is performed for a preset duration to obtain a blank environment point dataset that does not contain any real partial discharge characteristics. The blank environment point dataset is then gridded statistically to construct a historical blank observation matrix. Subsequently, multiple random samples are performed along the dimension of the amplitude in this historical blank observation matrix. Each sample extracts a fixed-length noise sample sequence, and the variance value corresponding to each noise sample sequence is calculated, thus obtaining a set of independent variances reflecting the fluctuation amplitude of the blank environment noise. To eliminate the statistical randomness error caused by a single random sampling, the obtained set of independent variances is fitted with a Gaussian probability density function. The center position corresponding to this variance set, i.e., the expected value, is calculated through Gaussian fitting. Finally, the calculated expected value is determined as the standard noise variance of the partial discharge detector operating in the current environment. In the mathematical formula, this standard noise variance is expressed as a constant. .

[0042] Next, the variance deviation between the first local variance and the standard noise variance is calculated. In specific implementation, for the first... For each phase interval, subtract the standard noise variance from the first local variance, and then divide this difference by the standard noise variance. Calculate the... variance deviation of each phase interval The formula is as follows: In the formula, Indicates the first The variance deviation of each phase interval This represents the first local variance. This represents the standard noise variance. After calculating the deviation from this variance, the deviation is compared with a preset signal deviation threshold. Let this signal deviation threshold be a constant. If the variance deviation The signal deviates from the threshold. This indicates that the signal fluctuation within this phase interval is extremely weak and cannot be effectively distinguished from background noise, thus it is determined to be pure background noise. To completely eliminate background noise interference, when constructing the weighted projection function for noise suppression, all corresponding weighting coefficients are... Forced constraint and assignment of zero. This hard thresholding method can directly filter out noise regions without partial discharge characteristics and reduce them to zero.

[0043] If the calculated variance deviation is not lower than the aforementioned preset signal deviation threshold, it indicates the presence of a significant discharge signal characteristic exceeding the range of electromagnetic background noise fluctuations within that phase interval. To adaptively weight discharge signals of different intensities in subsequent dimensionality reduction projection, an exponential decay model is needed to nonlinearly map the variance deviation to continuous penalty coefficients. The design principle of this exponential decay model is that as the variance deviation increases, the rate of change of the mapping coefficients gradually slows down and tends to stabilize, thereby protecting the strong partial discharge characteristics while suppressing abrupt distortion caused by occasional ultra-large interference pulses. Specifically, in the mapping calculation, for the... For each phase interval, the continuous penalty coefficient corresponding to that phase interval is calculated using the exponential decay model. The formula is as follows: In the formula, Represents the calculated first... The continuous penalty coefficients corresponding to each phase interval, with constant 1 serving as the upper limit benchmark for the weighting coefficients. Indicates the first The variance deviation of each phase interval This represents a pre-configured decay adjustment factor used to control the steepness of the exponential decay curve. This nonlinear mapping allows variance deviations exceeding a threshold to be normalized into a continuous penalty coefficient within the range of zero to one.

[0044] For the first [unclear] with a valid partial discharge signal Each phase interval, all The weighting coefficients corresponding to the amplitude intervals are assigned values ​​equal to the continuous penalty coefficients. In other words, during the final projection calculation, the weighting of each amplitude data point within that phase interval is positively correlated with the current variance deviation. Through this design, it is mathematically possible to construct a weighting coefficient for the th amplitude interval in a two-dimensional partial discharge matrix. The phase interval, the first The final weighting coefficient for each amplitude range of data The formula for segmented construction is expressed as follows: In the formula, Indicates the first The phase interval, the first The final weighting coefficients for each amplitude range, Indicates the degree of variance deviation. This indicates that the signal deviates from the threshold. This represents the calculated continuous penalty coefficient. This piecewise formula constitutes the complete operating logic of the weighted projection function, achieving the suppression effect of zeroing the weights in the pure noise interval and adaptively continuously weighting the signal interval, which can greatly improve the data quality after projection dimensionality reduction.

[0045] In one embodiment, the standard noise variance determined when acquiring historical blank environmental data collected by the partial discharge detector includes the following steps: When the partial discharge detector is not connected to the device under test, send a blank acquisition command to the partial discharge detector; The partial discharge detector continuously receives blank data points for a preset duration in response to the blank acquisition command, and generates a historical blank observation matrix based on the blank data points. Randomly sample the historical blank observation matrix according to the amplitude dimension to obtain multiple noise sample sequences; Calculate the independent variance of each noise sample sequence separately; Gaussian fitting is performed on all independent variance values ​​to obtain the expected value, and the expected value is determined as the standard noise variance.

[0046] In this embodiment, during the partial discharge detection calibration initialization phase, it is first ensured that the signal input terminals of the partial discharge detector are completely unconnected or not connected to any operating power equipment, in order to shield all actual partial discharge signals from power equipment. After confirming that the physical connection is empty, a pre-encapsulated blank acquisition command is sent to the partial discharge detector via the mobile terminal's wireless communication link. The blank acquisition command contains specific function control codes, device address codes, checksums, and command length identifiers in its data frame format, specifically used to trigger the high-bandwidth analog signal conditioning circuit and high-speed analog-to-digital conversion circuit inside the partial discharge detector to enter the background noise acquisition mode. Through the triggering mechanism in a purely physical blank state, it can be ensured that all subsequently acquired data is completely free from any actual partial discharge pulse waveforms, thus purely reflecting the electrical noise level of the partial discharge detector in the current spatial electromagnetic environment. The process of sending blank acquisition commands, as the first step in parameter calibration, enables the partial discharge detector to accurately capture the most basic hardware thermal noise, shot noise, and inherent background high-frequency radiation interference of the surrounding environment without being affected by the polarization interference of the high-voltage alternating electric field of the power equipment. This provides a reference point without signal pollution for the noise suppression of the entire partial discharge analysis scheme.

[0047] Upon receiving a blank data acquisition command, the partial discharge detector immediately activates its high-speed analog-to-digital conversion channel and maintains continuous high-frequency signal acquisition for a preset duration, converting the acquired analog electrical signals into continuous digital blank data points. The preset duration is typically set to include several complete power grid voltage cycles, such as ten seconds, to ensure sufficient coverage of any potential power frequency-related periodic environmental electromagnetic radiation interference. All blank data points acquired within the preset duration are sequentially transmitted to the mobile terminal via a wireless channel. Upon receiving the blank data points representing the environmental background noise, the mobile terminal extracts the amplitude and phase information corresponding to each blank data point. Then, using the same phase and amplitude interval division rules as those used in constructing the two-dimensional partial discharge matrix, the phase is also divided into... The amplitude is divided into intervals. For each interval, a two-dimensional grid structure is established. All received blank data points are traversed, and each blank data point is assigned to its corresponding grid, with the statistical count incremented by one to generate a historical blank observation matrix characterizing the background noise distribution. Let the generated historical blank observation matrix be represented as... Any element in the matrix is ​​denoted as , indicating the first The amplitude range and the first The cumulative number of blank data points within each phase interval. The generated historical blank observation matrix. As a static two-dimensional statistical grid, it fully records the frequency of environmental noise occurrence in the cross regions of different phases and amplitudes under the condition of no partial discharge.

[0048] To extract statistically representative and independent noise features from the historical blank observation matrix, it is necessary to randomly sample the historical blank observation matrix according to its amplitude dimension. In the specific sampling process, since the distribution of background noise across different amplitude intervals is random and non-uniform, directly using the entire data segment can easily introduce localized interference. Therefore, randomness is introduced to eliminate statistical bias. The total number of sample sequences is defined as follows: The first. For the first A sequence of noise samples needs to be constructed from the historical blank observation matrix. of From each row vector (i.e., the magnitude dimension), a predetermined number of matrix elements are randomly selected and combined. Let the number of elements selected each time be... During sampling, a uniform random sampling method with replacement is used to ensure that noise data in each amplitude range has an equal probability of being selected. Through this sampling method, the first... The acquired noise sample sequence can be represented as a one-dimensional sequence. The specific sampling formula is as follows: In the formula, Indicates the first A sequence of noise samples, index This is the index of the sample sequence, with values ​​ranging from 1 to... Positive integers; This represents a matrix element randomly selected from the historical blank observation matrix, where... Indicates the first The sequence number is... The random row index corresponding to the next extraction. Indicates the corresponding random column index. The value range is from 1 to A positive integer. Repeat the above random sampling process for a total of... In one step, multiple independent noise sample sequences can be obtained.

[0049] The independent variance (IV) values ​​of each noise sample sequence are calculated to quantitatively assess the range of numerical fluctuations within each sequence. Variance is a core statistical indicator for measuring the dispersion of a set of data. In this step, calculating the variance reflects the dispersion of each specific noise sample sequence from its arithmetic mean, thus characterizing the micro-fluctuation characteristics of the noise. Regarding the first set of noise samples obtained in the preceding steps... A sequence of noise samples First, calculate all of the samples in the sequence. The arithmetic mean of the n elements is calculated, and then the sum of squares of the differences between each element in the sequence and the mean is calculated using the mean, and then divided equally. The nth element is calculated... Independent variance of each noisy sample sequence The calculation formula is as follows: In the formula, Indicates the first The independent variance values ​​calculated from each noise sample sequence Indicates the first The arithmetic mean of a sequence of noisy samples. For all By repeatedly performing the above variance calculation process on a sequence of noisy samples, a set of independent variance values ​​can be obtained. This set can be mathematically represented as a set containing... A one-dimensional vector of independent variance values. These independent variance values, derived from random sampling of different dimensions of the historical blank observation matrix, can comprehensively cover the fluctuations in background noise caused by factors such as environmental abrupt changes and temperature drift of the hardware itself.

[0050] After calculating the independent variances of all noise sample sequences, a Gaussian probability distribution is fitted to all independent variances. Gaussian fitting is typically based on the maximum likelihood estimation principle. By using the obtained independent variances as observed samples, it approximates a classic continuous normal distribution curve, thereby identifying the two core parameters of the Gaussian distribution: the expected value and the variance. The probability density function formula for Gaussian fitting is expressed as follows: In the formula, This represents the independent variance value used as the independent variable. This represents the expected value obtained through Gaussian fitting. The standard deviation of the Gaussian distribution is used to characterize the dispersion of the independent variance values. Gaussian fitting can effectively filter out abnormal variance values ​​caused by occasional large noise pulses. After completing the fitting calculation and obtaining the expected value of the Gaussian distribution... Next, the expected value will be calculated. The standard noise variance of the partial discharge detector is determined. Based on the definition above, this expected value is... Assigned to standard noise variance , that is to say The standard noise variance serves as the boundary between background noise and actual partial discharge signal in subsequent algorithms. Due to the layers of statistical smoothing through random sampling, independent variance calculation, and Gaussian fitting, it has extremely high robustness and can ensure that no misjudgment occurs in complex wireless monitoring environments.

[0051] In one implementation, calculating the cross-power spectrum of the current phase angle density vector and the standard phase angle standard vector in the frequency domain space, and determining the original phase slip based on the peak value of the cross-power spectrum includes the following steps: Perform a Fast Discrete Fourier Transform on the current phase density vector to obtain the current complex sequence in the frequency domain; Perform a Fast Discrete Fourier Transform on the standard phase angle standard vector to obtain a standard frequency domain complex sequence; A conjugate sequence of complex numbers in the frequency domain is obtained by performing a conjugate operation on a standard frequency domain complex sequence. Multiply the current frequency domain complex sequence element-wise with the conjugate frequency domain complex sequence to generate a cross-frequency domain matrix; Normalizing the cross-frequency domain matrix yields a smooth cross-frequency domain matrix; Perform an inverse fast discrete Fourier transform on the smooth cross-frequency domain matrix to generate a cross-power spectrum sequence in the time domain, and obtain the original phase shift by peak finding in the cross-power spectrum sequence.

[0052] In this embodiment, a Fast Discrete Fourier Transform (FFT) is performed on the generated current phase density vector to obtain the corresponding current frequency domain complex sequence. The FFT is an efficient algorithm for implementing the Discrete Fourier Transform, capable of decomposing the discharge density characteristics of the time-domain distribution into sine and cosine harmonic components of different frequencies, and representing them in complex form. Each complex number contains both amplitude and phase information. Let the current phase density vector be represented as the aforementioned one-dimensional vector. The length of this vector is equal to the total number of phase intervals divided. After performing a transformation on the vector, the generated current frequency domain complex sequence is the first... Complex elements at frequency indices The calculation formula is as follows: In the formula, Represents the current complex sequence in the frequency domain. Complex elements at frequency indices Represents the current phase angle density vector The first in One element, The length of the current phase angle density vector is represented by the independent variable. For time-domain indexing, This represents a frequency index, and the value of this frequency index ranges from zero to... Integers between -1 and -1.

[0053] Similar to the frequency domain transformation performed on the current phase angle density vector, the same Fast Discrete Fourier Transform (FFT) is performed on the standard phase angle standard vector downloaded from the cloud big data platform to obtain the corresponding standard frequency domain complex sequence. The standard phase angle standard vector is generated by clustering historical partial discharge data and represents the absolute phase probability distribution under ideal conditions. Transforming it to the frequency domain allows extraction of the standard spectral distribution pattern representing the characteristics of this device type. In implementation, the standard phase angle standard vector obtained from the cloud is represented as a one-dimensional vector. The length of this vector is also equal to the total number of phase intervals divided. After performing a Fast Discrete Fourier Transform on the vector, the first number in the generated standard frequency domain complex sequence is... Complex elements at frequency indices The calculation formula is as follows: In the formula, Represents the first complex number in the standard frequency domain sequence. Complex elements at frequency indices Standard phase angle standard vector The first in Each element.

[0054] Next, a complex conjugate operation with dispersion compensation needs to be performed on the standard frequency domain complex sequence to obtain a conjugate frequency domain complex sequence. In practice, this conjugate operation is not simply a matter of inverting the imaginary part; instead, a nonlinear dispersion compensation mechanism is introduced to correct the phase distortion caused by hardware filtering. This conjugate operation is implemented using the following nonlinear dispersion-compensated conjugate formula: .

[0055] In the formula, Represents the first conjugate standard frequency domain complex sequence. Complex elements at frequency indices; Represents the first complex number in the standard frequency domain sequence. Amplitude components at each frequency index; Represents the first complex number in the standard frequency domain sequence. The original phase angle at each frequency index; The global reference variance represents the amplitude components at all frequency indices in the standard frequency domain complex sequence, and is used as a quantization noise floor benchmark; This represents the preset phase curvature compensation constant, used to correct the nonlinear group delay defocusing generated by the internal hardware filter circuit of the partial discharge detector. Represents the original phase angle at the th Discrete second-order partial derivatives at each frequency index. This formula enables hardware dispersion correction during conjugate operations, resulting in a high-fidelity conjugate frequency domain complex sequence.

[0056] The relative phase difference information between two signals is extracted using mathematical superposition. In practice, the current frequency domain complex sequence is multiplied element-wise with its conjugate frequency domain complex sequence to generate a cross-frequency domain matrix representing their frequency domain correlation. Since multiplying two complex numbers in the frequency domain is equivalent to multiplying their magnitudes and subtracting their phase angles, this multiplication operation can directly extract the phase difference distribution of the current signal relative to the standard reference signal. In the specific calculation, for each frequency index, complex multiplication is performed on the corresponding two complex elements. Let the generated cross-frequency domain matrix contain the i-th... The complex elements at each frequency index are represented as The specific formula for element-by-element multiplication is as follows: In the formula, the element-wise multiplication operation merges two independent frequency domain complex sequences into a single cross-frequency domain moment that contains complete relative phase shift characteristics.

[0057] To eliminate amplitude inhomogeneities in partial discharge signals caused by drastic fluctuations in discharge intensity or transient large pulses, thus ensuring that phase slip estimation depends entirely on the phase itself rather than the discharge amplitude, the cross-frequency domain matrix must be normalized to obtain a smooth cross-frequency domain matrix. Specifically, the normalization process employs a phase transform weighting method, retaining the phase angle information of each complex element in the cross-frequency domain matrix while forcibly normalizing all its amplitudes to a unit length. This process significantly enhances the algorithm's adaptability to different discharge intensities and effectively suppresses interference from high-frequency burst noise on phase alignment. Let the obtained smooth cross-frequency domain matrix contain the... The normalized complex elements at each frequency index are represented as follows: The specific normalization calculation formula is as follows: In the formula, Represents the first element in the smooth cross-frequency domain matrix. Normalized complex elements at frequency indices Represents the first in the cross-frequency domain matrix The modulus of the complex number at each frequency index. This represents a preset minimum positive number used to prevent the denominator from being zero. Through this normalization calculation, each frequency domain element in the cross-frequency domain matrix is ​​smoothed into a complex number containing only pure phase difference components and having a magnitude of one, thereby generating a high-quality smooth cross-frequency domain matrix, which greatly improves the stability and resolution of the final peak finding calculation.

[0058] Next, the frequency domain information is converted back to the time domain or the spatial physical phase domain to intuitively extract the specific phase offset. In practice, an inverse fast discrete Fourier transform is performed on the smoothed cross-frequency domain matrix to reconstruct a cross-power spectrum sequence representing the correlation between the two signals in the time-space domain. The inverse fast discrete Fourier transform is a fast implementation of the inverse discrete Fourier transform, which can convert the phase difference information in the frequency domain into the impulse response in the time-space domain. Let the generated cross-power spectrum sequence contain the... The real elements at each time-domain index are represented as follows: The specific inverse transform formula is as follows: .

[0059] In the formula, The first digit represents the cross-power spectrum sequence in the time domain. The real value at each time-domain index, the independent variable The value range is from zero to... Integers between -1 and 1. After calculating the cross-power spectrum sequence, the position with the largest value in the sequence corresponds to the relative offset of the two phase angle density vectors with the highest degree of overlap. By performing a peak-finding search algorithm in the generated cross-power spectrum sequence, the array index subscript corresponding to the maximum value, i.e., the absolute main peak, is located. This array index subscript reflects the physical deviation between the two waveforms and is identified as the original phase slip caused by the current wireless transmission.

[0060] In one embodiment, obtaining the original phase shift by peak finding in the cross-power spectrum sequence includes the following steps: Calculate the global mean and global standard deviation of all elements in the cross power spectrum sequence, and use the global mean and global standard deviation to set the dynamic peak finding threshold; Select a set of candidate peak points whose values ​​are greater than the dynamic peak-finding threshold from the cross-power spectrum sequence; Calculate the local second derivative of each candidate peak point in the candidate peak point set, and select the candidate peak point with the largest local second derivative as the absolute main peak; Extract the array index subscript corresponding to the absolute main peak in the cross-power spectrum sequence, convert the array index subscript into the physical phase angle difference, and use the physical phase angle difference as the original phase shift.

[0061] In this embodiment, after obtaining the cross-power spectrum sequence in the time domain, in order to accurately identify the peaks representing the true phase deviation in spectral lines containing background noise and occasional perturbations, it is necessary to perform global statistical feature calculations on the cross-power spectrum sequence to set a dynamic peak-finding threshold that can adaptively adjust according to the current signal quality. In the specific calculation process, firstly, all real elements in the cross-power spectrum sequence are traversed, and the global average value of all elements in the sequence and the global standard deviation used to characterize the amplitude of data discrete fluctuations are calculated. These two statistical measures are used to set the dynamic peak-finding threshold, thereby adaptively raising or lowering the threshold according to the background energy fluctuations of the current cross-power spectrum. The calculation of the... The global average of the cross-power spectrum sequence composed of real elements at each time-domain index. Global standard deviation and dynamic peak finding threshold The specific calculation formula is as follows: .

[0062] In the formula, Represents the global average value of the cross-power spectrum sequence; The first digit represents the cross-power spectrum sequence in the time domain. The real value at each time-domain index; This indicates the length of the cross-power spectrum sequence, i.e., the total number of phase intervals divided. This represents the global standard deviation of the cross-power spectrum sequence; This represents the calculated dynamic peak-finding threshold. This represents the preset threshold adjustment coefficient, used to adjust the tolerance for noise amplitude suppression.

[0063] After successfully setting the dynamic peak-finding threshold, the cross-power spectrum sequence is filtered for local maxima to extract all candidate peaks that may correspond to the actual physical phase angle difference. During the filtering process, due to potential local jitter caused by residual high-frequency noise in the cross-power spectrum sequence, simply searching for the global maximum value can easily lead to spurious peaks. Therefore, a dual constraint is used to traverse and filter the cross-power spectrum sequence. The first constraint requires that the element value at the current index position must be strictly greater than the calculated dynamic peak-finding threshold; the second constraint requires that the current element must be a local maximum within the adjacent range. Considering the 360-degree cyclic physical characteristic of the partial discharge phase distribution, cyclic modulo operation is used when determining the adjacency relationship of boundary elements. Let the set of candidate peak points satisfying the above filtering conditions be represented as... The formula for selecting elements from the candidate peak point set is expressed as follows: .

[0064] In the formula, This represents the set of candidate peak points selected. This represents the array index subscript indicating the location of the candidate peak, with values ​​ranging from zero to... Between one and... This indicates that the index is... Real values ​​of the cross-power spectrum at the location; Indicates the dynamic peak finding threshold; This indicates the length of the cross-power spectrum sequence, i.e., the total number of phase intervals divided. This represents the mathematical modulo operation. Using this formula, when traversing the entire sequence, points that do not meet the threshold condition or are not local maxima are all removed, thus filtering out a large amount of spectral noise. Only potential peaks that meet physical characteristics are indexed into the candidate peak point set, significantly reducing the search space for subsequent calculations.

[0065] From the selected set of candidate peak points, it is necessary to further identify the absolute main peak that represents the strongest correlation phase alignment information. In the specific identification process, simply selecting the point with the largest amplitude may introduce severe alignment deviations when facing complex conditions with strong network noise or multiple partial discharge signal sources. The spectral peaks that truly achieve phase alignment at the same frequency should possess an extremely steep geometric profile in their physical structure. This geometric steepness can be accurately characterized in discrete space by calculating the discrete local second derivative. For each candidate peak point in the set, a second-order difference operation is performed using the forward and backward adjacent points to scientifically quantify the sharpness of the peak. The calculation of the... Local second derivative values ​​of each candidate peak point The specific formula is as follows: .

[0066] In the formula, Indicates the first The local second derivative values ​​calculated from each candidate peak point are used to quantify the physical sharpness of the peak. Indicates the corresponding index subscript The real values ​​of the cross power spectrum at the given location, and Belongs to the set of candidate peak points ; This represents the length of the cross-power spectrum sequence, i.e., the total number of phase intervals divided. After sequentially calculating the local second derivative values ​​corresponding to all elements in the candidate peak point set, a comparative search is performed to select the local second derivative value. The largest candidate peak is selected as the absolute dominant peak. This peak curvature-based screening mechanism can completely eliminate smooth stray signals and lock in the phase-aligned peak with the highest physical reliability.

[0067] The storage location of the absolute dominant peak in the discrete array space is converted into a phase angle difference parameter with practical electrical and physical significance. In the specific extraction and conversion process, since the cross-power spectrum sequence consists of two sequences of length... The phase density vector is reconstructed after performing Fourier transform and inverse transform. Therefore, each array index in this cross-power spectrum sequence linearly maps to and represents a phase deflection within the power frequency voltage cycle. First, the array index corresponding to the element located as the absolute main peak in the waveform correlation calculation is extracted. Let this array index be represented as a positive integer. Subsequently, by multiplying by the total angle value of the power frequency cycle and dividing by the linear proportionality of the total number of discrete phase intervals, the discrete array index subscripts are converted into continuous phase difference values ​​in physical angle units, which are then used as the original phase shift output. The original phase shift is then calculated. The physical conversion formula is as follows: In the formula, This represents the final calculated and output raw phase shift, with the physical unit defined as degrees in degrees. This represents the array index subscript corresponding to the extracted absolute main peak in the cross-power spectrum sequence. The value of this subscript ranges from zero to... Between one and... This represents the length of the cross-power spectrum sequence, i.e., the total number of phase intervals divided. This conversion formula successfully and efficiently restores the discrete digital signal analysis results to a specific physical phase offset parameter, which accurately reflects the absolute phase deviation between the currently wirelessly acquired partial discharge waveform and the power grid standard frequency reference.

[0068] In one implementation, evaluating the network delay variance during wireless communication to determine the current network congestion status, and matching the target filtering parameters from a preset filtering coefficient mapping table on a cloud-based big data platform based on the current network congestion status, includes the following steps: Record the local transmission time when sending a data acquisition request to the partial discharge detector, and record the local reception time when receiving the response data packet; Calculate the difference between the local reception time and the local transmission time to obtain the round-trip network delay. The single round-trip network latency within multiple consecutive polling cycles is stored in a sliding time window, and the latency variance data of all single round-trip network latencies within the sliding time window are calculated. The time delay variance data is input into a preset network state classifier to classify the corresponding current network congestion state, and the corresponding target filtering parameters are retrieved from the preset filtering coefficient mapping table of the cloud big data platform according to the current network congestion state.

[0069] In this embodiment, when a specific data acquisition request containing data control characteristics is sent to the partial discharge detector, at the critical moment when the data packet containing the specific data acquisition request is about to be transmitted through the wireless transmission physical layer, a high-precision system-level timer is immediately invoked to capture and save the current local high-precision nanosecond-level timestamp as the local transmission time. The partial discharge detector receives the specific data acquisition request in the wireless network channel, and after completing internal data reading and data format encapsulation, generates a wireless response data packet containing discrete data points of the partial discharge phase-resolved spectrum, and transmits it in reverse through the wireless channel. At the specific moment when the wireless response data packet is received and a network reception interrupt at the underlying communication protocol stack is triggered, the aforementioned high-precision system-level timer is invoked again to immediately capture and save the current local high-precision nanosecond-level timestamp as the local reception time.

[0070] Due to uncertain physical factors such as multipath fading, multi-user contention, and retransmission mechanisms in wireless networks, each data round trip consumes a different amount of time, which is a fundamental indicator for determining network congestion. The specific calculation principle is as follows: subtract the local transmission time from the local reception time recorded in the preceding steps. This difference operation eliminates the static overhead caused by hardware queuing and channel contention at the transmitting end, thereby extracting the single round-trip network delay, which includes uplink network delay, partial discharge detector internal data processing delay, and downlink network delay. The calculation of the... Network latency per round trip within the next polling cycle The formula is as follows: In the formula, Indicates the first The network latency per round trip corresponding to each polling cycle; Indicates the first The local reception time recorded in the next polling cycle; Indicates the first The local sending time recorded in the next polling cycle, where the independent variable is a positive integer representing the polling period.

[0071] Set the length of the sliding time window to This length corresponds to the number of consecutive polling cycles collected. After each round-trip network latency calculation, the latest calculated single-round-trip network latency is pushed to the end of the sliding time window, while the oldest latency data is removed from the beginning of the window to maintain a constant total amount of data within the sliding time window. Subsequently, the latency variance data for all single-round-trip network latencies stored within the sliding time window is calculated. Variance quantifies the severity of latency data fluctuations and is an excellent statistical parameter reflecting network jitter. The corresponding latency variance data is then calculated. The formula is as follows: In the formula, This represents the calculated time delay variance data; Indicates the length of the preset sliding time window; Indicates the number of times within the current sliding time window. A stored single round-trip network latency; This represents the arithmetic mean of all saved single-round-trip network delays within the current sliding time window.

[0072] The pre-defined network state classifier can be constructed using classification algorithms such as Support Vector Machines, Decision Trees, or Multilayer Perceptrons, and the classification boundaries can be determined in advance through offline learning. In specific implementation, the network state classifier, based on the magnitude of the input delay variance data, calculates the current network congestion state within several predefined congestion level intervals through forward inference. The corresponding current network congestion state can intuitively represent the severity of interference and congestion on the current wireless transmission channel, for example, it can be divided into four levels: smooth, lightly congested, moderately congested, and severely congested. Then, using the identified current network congestion state as the retrieval key, a search and matching operation is performed in a pre-defined filter coefficient mapping table in the cloud big data platform via the wireless communication link. This filter coefficient mapping table has a pre-established optimal correspondence between different congestion states and filter parameters. Through the retrieval operation, the target filter parameters that best match the current network condition are extracted and downloaded. These target filter parameters include observation gain coefficients and tracking gain coefficients, which are used to dynamically adjust the filter's response speed and noise smoothing capability, ensuring excellent phase correction accuracy even in any unstable network environment.

[0073] In one embodiment, inputting the original phase shift into a tracking filter configured with target filtering parameters, eliminating the phase shift caused by wireless network latency, and outputting the predicted phase compensation step size includes the following steps: The original phase shift is input into a tracking filter configured with target filtering parameters, and the target filtering parameters are resolved into observation gain coefficients for position updates and tracking gain coefficients for velocity updates. Calculate the position residual between the prior phase estimate and the original phase slip for the current prediction period; The posterior phase state of the tracking filter is updated using the observation gain coefficient and the position residual, and the slip velocity state of the tracking filter is updated using the tracking gain coefficient and the position residual. Based on the posterior phase state quantity and the updated slip velocity state quantity, the prior phase estimate for the next prediction period is calculated. The updated slip velocity state is extracted as the prediction phase compensation step size.

[0074] In this embodiment, the original phase slip reflecting the current wireless transmission phase deviation is obtained. The original phase was shifted The data is sequentially input into the tracking filter configured with the target filtering parameters. During the initialization data parsing phase of the tracking filter, the target filtering parameters are decoupled and specifically parsed into two key control gain coefficients: the observation gain coefficient for position state updates and the tracking gain coefficient for velocity state updates. These two gain coefficients are obtained by retrieving and matching data from the cloud-based big data platform based on the network congestion status in the preceding steps, and possess adaptive network fluctuation characteristics. Let the parsed observation gain coefficient be expressed as a floating-point constant. The tracking gain coefficient obtained after analysis is expressed as a floating-point constant. The observed gain coefficient The tracking gain coefficient is used to adjust the correction ratio of the actual observations in the current period to the filter state position estimate. It is used to adjust the correction sensitivity of the phase sliding speed caused by the offset of wireless channel transmission delay.

[0075] When the tracking filter reaches the iteration time of the current prediction period, the position residual between the prior phase estimate and the actual observed phase slip is first calculated. The prior phase estimate is a forward prediction of the phase offset at the current moment based on multiple historical filter states, while the actual phase slip is the actual observed value obtained at the current moment through cross-power spectrum peak finding, containing random interference from network delay. Due to the unavoidable random queuing, channel preemption, and retransmission mechanisms in wireless communication, the current actual observed value is mixed with high-frequency noise jitter. By calculating the physical deviation between these two values, the position residual between the current predicted position and the actual observed position can be accurately obtained. The calculated position residual is denoted as [symbol missing]. The mathematical formula for calculating the residual at this location is expressed as follows: , This represents the positional residual for the current forecast period; This represents the original phase shift of the input; This represents the prior phase estimate for the current prediction period, indicated by the subscript in the formula. Represents a priori.

[0076] The analytically obtained observation gain coefficient is multiplied by the position residual, and this product is added to the prior phase estimate for the current prediction period. This updates the posterior phase state of the filter, eliminating random delay noise and achieving smooth phase reconstruction. Simultaneously, to dynamically track the phase slip velocity caused by power frequency offset or long-term delay accumulation, the analytically obtained tracking gain coefficient is multiplied by the position residual, and this product is added to the slip velocity state stored in the previous period, updating the slip velocity state. The corresponding state correction calculation formula is expressed as follows: .

[0077] In the formula, The subscript in the formula represents the updated posterior phase state. Representing the posterior; The updated slip velocity state variable is indicated by its subscript in the formula. Represents speed; This represents the slip velocity state value stored by the tracking filter at the end of the previous prediction cycle, indicated by the subscript in the formula. This represents historical data. Through this decoupled parallel update of the two states, an optimal dynamic balance between noise suppression and trend following can be achieved, effectively enhancing the phase tracking robustness in complex power grid environments.

[0078] Next, a forward prediction calculation is performed on the phase shift trend within the next prediction period. This forward prediction calculation demonstrates the adaptive extrapolation capability of the tracking filter. The physical principle of extrapolation is that, within an extremely short grid cycle interval, the phase slip caused by network delay offset increases linearly with the updated slip velocity state variable as the increment. In the specific calculation implementation, the updated posterior phase state variable and the updated slip velocity state variable are summed to obtain the prior phase estimate for the next prediction period. The prior phase estimate for the next prediction period is then calculated. The formula is as follows: , This represents the calculated prior phase estimate for the next prediction period, and this prior phase estimate will replace the current prior phase estimate during the iteration of the next period. This forward-step prediction calculation, by physically fusing the current phase position state with the velocity trend, can predict the phase shift caused by network latency in advance, giving the entire phase alignment and correction process proactive compensation and fundamentally eliminating the in-frequency alignment hysteresis error caused by network feedback delay.

[0079] After completing the above iterative updates, in order to apply the denoising results obtained from the filtering calculations to the actual two-dimensional partial discharge (PD) matrix correction, it is necessary to extract the physical slip increment in the latest state. In the specific implementation steps, the updated slip velocity state variable is extracted from the memory data structure of the tracking filter and directly assigned as the predicted phase compensation step size. This slip velocity state variable is the best estimate of the true phase offset slope of the PD waveform after network delay jitter smoothing filtering, representing the absolute phase cumulative offset rate caused by small fluctuations in the physical power frequency voltage. Let this predicted phase compensation step size be expressed as... The specific formulas for extraction and assignment are as follows: Through this extraction and transformation process, the high-dimensional virtual state variables inside the tracking filter are successfully converted into phase displacement control parameters that can directly affect the physical data space alignment. This predicted phase compensation step size completely eliminates various non-steady-state random delay errors introduced by the wireless network during transmission, accurately reflecting the phase offset step size caused by the non-coordinated clocks between the mobile terminal and the partial discharge detector.

[0080] This invention also discloses a partial discharge wireless co-frequency phase locking system integrating big data intelligent analysis, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the partial discharge wireless co-frequency phase locking method integrating big data intelligent analysis as described above.

[0081] The processor can be a central processing unit (CPU). Of course, depending on the actual use, it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), off-the-shelf programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc., and this application does not limit it.

[0082] The memory can be an internal storage unit of a computer device, such as a hard disk or RAM, or an external storage device, such as a plug-in hard disk, smart memory card (SMC), secure digital card (SD), or flash memory card (FC) provided on the computer device. Furthermore, the memory can be a combination of internal storage units and external storage devices of a computer device. The memory is used to store computer programs and other programs and data required by the computer device. The memory can also be used to temporarily store data that has been output or will be output. This application does not limit this.

[0083] The present invention also discloses a computer-readable storage medium storing instructions that, when executed by a processor, configure the processor to perform the partial discharge wireless co-frequency phase locking method integrating big data intelligent analysis described in any of the above embodiments.

[0084] The computer program can be stored in a machine-readable medium. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or certain middleware. The machine-readable medium includes any entity or device capable of carrying computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the machine-readable medium includes, but is not limited to, the above-mentioned components.

[0085] The partial discharge wireless phase-locking method integrating big data intelligent analysis described in the above embodiments is stored in the computer-readable storage medium and loaded and executed on the processor to facilitate the storage and application of the above method.

[0086] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of protection of this application is limited to these examples; within the framework of this application, the technical features of the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of different aspects of one or more embodiments of this application as described above, which are not provided in detail for the sake of brevity.

[0087] One or more embodiments in this application are intended to cover all such substitutions, modifications, and variations that fall within the broad scope of this application. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of one or more embodiments in this application should be included within the protection scope of this application.

Claims

1. A partial discharge wireless co-frequency phase locking method integrating big data intelligent analysis, characterized in that, The steps include the following: When a mobile terminal establishes a wireless communication connection with a partial discharge detector, the standard phase angle standard vector is obtained from the cloud big data platform based on the device type of the partial discharge detector. The standard phase angle standard vector is a standard vector representing the absolute phase probability distribution generated by clustering historical partial discharge data. Discrete data points of the partial discharge phase-resolved spectrum containing the original discharge information are obtained from the partial discharge detector, and a two-dimensional partial discharge matrix of the discharge state is generated based on the phase interval and amplitude interval according to the discrete data points. The two-dimensional partial discharge matrix is ​​reduced in dimension by projection and accumulation along the amplitude dimension, generating the current phase angle density vector representing the current discharge density. Calculate the cross-power spectrum between the current phase angle density vector and the standard phase angle standard vector in the frequency domain space, and determine the original phase slip based on the peak value of the cross-power spectrum; The network delay variance during wireless communication is evaluated to determine the current network congestion status, and the target filtering parameters are matched from the preset filtering coefficient mapping table of the cloud big data platform based on the current network congestion status. The original phase shift is input into the tracking filter configured with the target filtering parameters to remove the phase shift caused by wireless network latency and output the predicted phase compensation step size. Based on the predicted phase compensation step size, a reverse cyclic translation operation is performed on the discrete data points in the two-dimensional partial discharge matrix to align and lock the offset discrete data points into the same frequency standard phase domain consistent with the power grid frequency, thus obtaining the same frequency locked discrete data points. Extract the current prediction residual of the tracking filter, adjust the scatter point overlap color threshold rule in the graphics rendering component according to the current prediction residual, and perform cumulative rendering display of the discrete data points after same frequency locking on the mobile terminal interface according to the adjusted scatter point overlap color threshold rule.

2. The partial discharge wireless co-frequency phase locking method integrating big data intelligent analysis according to claim 1, characterized in that, The step of projecting and accumulating the two-dimensional partial discharge matrix along the amplitude dimension to reduce its dimensionality and generate the current phase angle density vector representing the current discharge density includes the following steps: Extract the amplitude data of all amplitude intervals corresponding to each phase interval in the two-dimensional partial discharge matrix; Calculate the first local variance of the amplitude data for all amplitude intervals, and construct a weighted projection function to suppress background noise based on the first local variance; The amplitude data of all amplitude intervals are input into the weighted projection function for weighted summation calculation to obtain the aggregate density value of the corresponding phase interval; The aggregated density values ​​of all phase intervals are concatenated into the current phase angle density vector according to the order of the phase intervals.

3. The partial discharge wireless co-frequency phase locking method integrating big data intelligent analysis according to claim 2, characterized in that, The construction of the weighted projection function for suppressing background noise based on the first local variance includes the following steps: Standard noise variance determined when acquiring historical blank environmental data collected by a partial discharge detector; Calculate the variance deviation between the first local variance and the standard noise variance. If the variance deviation is lower than the preset signal deviation threshold, the corresponding amplitude data is determined to be pure background noise, and the corresponding weighting coefficient is assigned to zero in the weighted projection function. If the variance deviation is not lower than the signal deviation threshold, the variance deviation is mapped to a continuous penalty coefficient using the exponential decay model. The continuous penalty coefficients are used as weighting coefficients for the corresponding amplitude data in the weighted projection function to construct the weighted projection function.

4. The partial discharge wireless co-frequency phase locking method integrating big data intelligent analysis according to claim 3, characterized in that, The standard noise variance determined when acquiring historical blank environmental data collected by the partial discharge detector includes the following steps: When the partial discharge detector is not connected to the device under test, send a blank acquisition command to the partial discharge detector; The partial discharge detector continuously receives blank data points for a preset duration in response to the blank acquisition command, and generates a historical blank observation matrix based on the blank data points. Randomly sample the historical blank observation matrix according to the amplitude dimension to obtain multiple noise sample sequences; Calculate the independent variance of each noise sample sequence separately; Gaussian fitting is performed on all independent variance values ​​to obtain the expected value, and the expected value is determined as the standard noise variance.

5. The partial discharge wireless phase-locking method integrating big data intelligent analysis according to claim 1, characterized in that, The step of calculating the cross-power spectrum of the current phase angle density vector and the standard phase angle standard vector in the frequency domain space, and determining the original phase slip based on the peak value of the cross-power spectrum, includes the following steps: Perform a Fast Discrete Fourier Transform on the current phase density vector to obtain the current complex sequence in the frequency domain; Perform a Fast Discrete Fourier Transform on the standard phase angle standard vector to obtain a standard frequency domain complex sequence; A conjugate sequence of complex numbers in the frequency domain is obtained by performing a conjugate operation on a standard frequency domain complex sequence. Multiply the current frequency domain complex sequence element-wise with the conjugate frequency domain complex sequence to generate a cross-frequency domain matrix; Normalizing the cross-frequency domain matrix yields a smooth cross-frequency domain matrix; Perform an inverse fast discrete Fourier transform on the smooth cross-frequency domain matrix to generate a cross-power spectrum sequence in the time domain, and obtain the original phase shift by peak finding in the cross-power spectrum sequence.

6. The partial discharge wireless co-frequency phase locking method integrating big data intelligent analysis according to claim 5, characterized in that, The process of obtaining the original phase shift by peak finding in the cross-power spectrum sequence includes the following steps: Calculate the global mean and global standard deviation of all elements in the cross-power spectrum sequence, and use the global mean and global standard deviation to set the dynamic peak-finding threshold; Select a set of candidate peak points whose values ​​are greater than the dynamic peak-finding threshold from the cross-power spectrum sequence; Calculate the local second derivative of each candidate peak point in the candidate peak point set, and select the candidate peak point with the largest local second derivative as the absolute main peak; Extract the array index subscript corresponding to the absolute main peak in the cross-power spectrum sequence, convert the array index subscript into the physical phase angle difference, and use the physical phase angle difference as the original phase shift.

7. The partial discharge wireless phase-locking method integrating big data intelligent analysis according to claim 1, characterized in that, The process of evaluating network delay variance during wireless communication to determine the current network congestion status, and matching target filtering parameters from a preset filtering coefficient mapping table on the cloud big data platform based on the current network congestion status, includes the following steps: Record the local transmission time when sending a data acquisition request to the partial discharge detector, and record the local reception time when receiving the response data packet; Calculate the difference between the local reception time and the local transmission time to obtain the round-trip network delay. The single round-trip network latency within multiple consecutive polling cycles is stored in a sliding time window, and the latency variance data of all single round-trip network latencies within the sliding time window are calculated. The time delay variance data is input into a preset network state classifier to classify the corresponding current network congestion state, and the corresponding target filtering parameters are retrieved from the preset filtering coefficient mapping table of the cloud big data platform according to the current network congestion state.

8. The partial discharge wireless co-frequency phase locking method integrating big data intelligent analysis according to claim 1, characterized in that, The step of inputting the original phase shift into a tracking filter configured with target filtering parameters, eliminating the phase shift caused by wireless network latency, and outputting the predicted phase compensation step size includes the following steps: The original phase shift is input into a tracking filter configured with target filtering parameters, and the target filtering parameters are resolved into observation gain coefficients for position updates and tracking gain coefficients for velocity updates. Calculate the position residual between the prior phase estimate and the original phase slip for the current prediction period; The posterior phase state of the tracking filter is updated using the observation gain coefficient and the position residual, and the slip velocity state of the tracking filter is updated using the tracking gain coefficient and the position residual. Based on the posterior phase state quantity and the updated slip velocity state quantity, the prior phase estimate for the next prediction period is calculated. The updated slip velocity state is extracted as the prediction phase compensation step size.

9. A partial discharge wireless phase-locking system integrating big data intelligent analysis, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the partial discharge wireless phase-locking method that integrates big data intelligent analysis as described in any one of claims 1 to 8.

10. A computer-readable storage medium storing instructions thereon, characterized in that, When executed by the processor, the instruction causes the processor to be configured to perform the partial discharge wireless co-frequency phase locking method according to any one of claims 1 to 8, which integrates big data intelligent analysis.

Images