Harmonic measurement method and system for power distribution network based on adaptive detection and parallel computation

Abstract

The application provides a power distribution network harmonic measurement method and system based on adaptive detection and parallel calculation, belongs to the technical field of dynamic harmonic measurement, and comprises the following steps: acquiring a power distribution network signal and performing discrete sampling in an observation time window to obtain a discrete signal sequence; calculating kurtosis, spectral entropy and residual gradient norm of the initial residual vector; dynamically adjusting an adaptive threshold parameter based on the kurtosis, spectral entropy and residual gradient norm; calculating a pulse noise detection threshold according to the adaptive threshold parameter; performing point-by-point comparison on the initial residual vector by using the pulse noise detection threshold, identifying a pulse noise occurrence position, and forming a pulse noise support set; constructing an enhanced basis matrix based on the pulse noise support set; decomposing a pseudo-inverse of the enhanced basis matrix into four sub-block matrices; performing parallel calculation on the four sub-block matrices; and finally obtaining amplitudes, phases, first derivatives and second derivatives of each harmonic.

Description

Technical Field

[0001] This invention belongs to the field of dynamic harmonic measurement technology, and particularly relates to a method and system for measuring harmonics in power distribution networks based on adaptive detection and parallel computing. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] With the large-scale integration of distributed photovoltaic power generation systems, electric vehicle charging facilities, energy storage converters, and other power electronic equipment into distribution networks, the operating environment of distribution networks exhibits significant nonlinear, weakly damped, and strong random disturbance characteristics. These power electronic devices introduce a large number of harmonics, electromagnetic interference, and transient impacts during operation, leading to a significant increase in the harmonic content of voltage and current signals in the distribution network, and making power quality problems increasingly prominent. Therefore, high-precision, real-time estimation of harmonic components in the distribution network is a crucial foundation for realizing online power quality monitoring, system protection, operational assessment, and fault diagnosis.

[0004] In practical engineering environments, in addition to being affected by background noise, power distribution network signals are often interfered with by random impulse noise caused by factors such as lightning strikes, equipment switching, transients in power electronic device switching, and communication interference. This type of impulse noise typically has characteristics such as random occurrence time, extremely short duration, and amplitude much larger than normal signals. Its statistical characteristics deviate significantly from the Gaussian distribution assumption, classifying it as typical non-Gaussian noise. This noise manifests as isolated or clustered outliers in sampled data, easily misidentified as valid signal components by traditional algorithms, thus creating a serious "bad data" problem and significantly reducing the accuracy of harmonic estimation and system robustness.

[0005] To address the aforementioned issues, existing research has proposed joint estimation methods based on Taylor-Fourier expansion, sparse representation, and enhanced basis matrices. These methods introduce impulse noise terms into the harmonic basis matrix to achieve simultaneous identification and separation of harmonic parameters and impulse noise. Such methods have improved harmonic estimation performance to some extent in non-Gaussian noise environments. However, existing joint identification methods still have the following shortcomings in practical applications: On the one hand, existing methods for detecting impulse noise (IPN) typically rely on fixed thresholds, empirical threshold parameters, or adaptive threshold values ​​under the Gaussian background noise assumption. These parameter settings lack adaptability to the statistical characteristics of the residual signal and the non-stationary changes in noise. When operating conditions change or noise intensity fluctuates, the background noise often fails to strictly satisfy the Gaussian distribution characteristics, exhibiting significant skewed distribution features. Fixed thresholds or adaptive threshold values ​​under the Gaussian background noise assumption struggle to balance detection sensitivity and false detection rate, easily leading to missed or false detections, thus affecting the accuracy and stability of the joint estimation results.

[0006] On the other hand, during the construction of the dynamically enhanced basis matrix, as the impulse noise pollution rate increases, it is necessary to continuously expand the dimension of the basis matrix and recursively calculate its pseudo-inverse matrix. Due to the high computational complexity of the pseudo-inverse operation, existing methods mostly use serial or recursive methods to solve it, resulting in the overall computation time increasing approximately linearly with the number of IPNs, which is difficult to meet the computational efficiency requirements of online monitoring and real-time analysis of distribution networks. Summary of the Invention

[0007] To overcome the shortcomings of the prior art, this invention provides a method and system for measuring harmonics in power distribution networks based on adaptive detection and parallel computing. It can achieve adaptive impulse noise detection and high-precision estimation of harmonic parameters in non-Gaussian and non-stationary noise environments, while taking into account both computational efficiency and real-time performance in harmonic analysis.

[0008] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions: The first aspect is a method for measuring harmonics in distribution networks based on adaptive detection and parallel computing, including: The distribution network signal is acquired and discretized within the observation time window to obtain a discrete signal sequence; Calculate the angular frequencies of each harmonic based on the fundamental frequency, and construct the Taylor-Fourier basis matrix. Establish a discrete signal model based on discrete signal sequences and Taylor-Fourier basis matrices; Calculate the initial harmonic parameter estimates in the discrete signal model and calculate the initial residual vector; Calculate the kurtosis, spectral entropy, and residual gradient norm of the initial residual vector; The adaptive threshold parameter is dynamically adjusted based on the kurtosis, spectral entropy, and residual gradient norm. Calculate the impulse noise detection threshold based on the adaptive threshold parameter; The initial residual vector is compared point by point using the impulse noise detection threshold to identify the location of impulse noise and form an impulse noise support set. Construct an enhancement basis matrix based on the impulse noise support set; The pseudo-inverse of the enhancement basis matrix is ​​decomposed into four sub-block matrices, and the four sub-block matrices are computed in parallel to obtain the amplitude, phase, first derivative and second derivative of each harmonic.

[0009] Secondly, a distribution network harmonic measurement system based on adaptive detection and parallel computing is disclosed, including: The signal acquisition module is configured to acquire power distribution network signals and perform discretization sampling within the observation time window to obtain a discrete signal sequence; The basis matrix construction module is configured to: calculate the angular frequencies of each harmonic based on the fundamental frequency, and construct the Taylor-Fourier basis matrix; Establish a discrete signal model based on discrete signal sequences and Taylor-Fourier basis matrices; The initial estimation module is configured to: calculate the initial harmonic parameter estimates in the discrete signal model and calculate the initial residual vector; The statistical feature extraction module is configured to: calculate the kurtosis, spectral entropy, and residual gradient norm of the initial residual vector; The adaptive threshold generation module is configured to dynamically adjust the adaptive threshold parameters based on the kurtosis, spectral entropy, and residual gradient norm. Calculate the impulse noise detection threshold based on the adaptive threshold parameter; The impulse noise detection module is configured to: apply the impulse noise detection threshold to perform point-by-point comparison of the initial residual vector, identify the location of impulse noise occurrence, and form an impulse noise support set; The enhanced basis matrix construction module is configured to: construct an enhanced basis matrix based on the impulse noise support set; The parallel computing and output module is configured to: decompose the pseudo-inverse of the enhanced basis matrix into four sub-block matrices, perform parallel computing on the four sub-block matrices, and finally obtain the amplitude, phase, first derivative and second derivative of each harmonic.

[0010] The above one or more technical solutions have the following beneficial effects: The technical solution of this invention can achieve adaptive impulse noise detection and high-precision estimation of harmonic parameters in non-Gaussian and non-stationary noise environments, and is a harmonic analysis method that takes into account both computational efficiency and real-time performance, so as to improve the reliability and engineering applicability of broadband oscillation measurement, power quality monitoring and analysis under complex operating conditions of power distribution networks.

[0011] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0012] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0013] Figure 1 This is Embodiment 1 of the present invention. Detailed Implementation It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0014] It should be noted that the terminology used herein is for the purpose of describing particular implementations only and is not intended to limit the exemplary implementations of the present invention.

[0015] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0016] Example 1 This embodiment discloses a method for measuring dynamic harmonics in distribution networks based on adaptive detection of strong impulse noise and parallel accelerated joint estimation, including: Step 1: Signal Acquisition and Preprocessing: At key nodes in the distribution network, three-phase voltage and / or current signals are acquired using power quality monitoring devices or miniature synchronous phasor measurement units, within the observation time window. Discretization sampling is performed within the internal sample, where For the observation center time, Window length. Sampling interval. This yields a discrete signal sequence:

[0017] in This represents the number of sampling points.

[0018] Representing discrete signals as vectors:

[0019] After obtaining the discrete sampled signal, the angular frequencies of each harmonic are calculated based on the sampling time and the known fundamental frequency, where the fundamental angular frequency is:

[0020] No. The angular frequency of a subharmonic is expressed as:

[0021] in, To determine the harmonic order, a second-order Taylor expansion is performed on the complex envelope of each harmonic at the observation center time. Substituting the expanded expression into the discrete sampling times, Taylor-Fourier basis functions containing zero-order, first-order, and second-order terms are constructed. These functions are then concatenated according to the harmonic order to obtain the Taylor-Fourier basis matrix. .

[0022] Using second-order Taylor expansion ( K=2) The time-varying characteristics of the approximate harmonic parameters, and the established discrete signal model are as follows: In the formula The Taylor-Fourier basis matrix is ​​obtained by multiplying the Taylor expansion coefficient matrix and the frequency domain twitch factor matrix:

[0023] in, for Identity matrix, exponent term Indicates the first Subharmonics in time shift Phase rotation factor at that point For the highest harmonic order under consideration, This represents the Kronecker product, used to construct an extended basis matrix with a block diagonal structure.

[0024] This is a harmonic parameter vector, containing the phasors of each harmonic and their first and second derivatives:

[0025] in, Let h be the phasor of the h-th harmonic. Let h be the first derivative of the phasor of the h-th harmonic. The phasor second derivative of the h-th harmonic is... This is a noise vector containing background noise and impulse noise.

[0026] Step 2: Adaptive IPN detection.

[0027] Since the support set of IPN is unknown in the initial stage, initialization... (Empty set), degenerates into a traditional TFT model. Initial estimates are calculated using the least squares method:

[0028] in, The initial harmonic parameter estimation vector contains the estimates of the complex phasors of each harmonic and their time-varying first and second derivatives.

[0029] Then calculate the initial residual vector:

[0030] in, This is the initial residual vector. The time-domain statistical characteristics of the residual vector are calculated: the mean is obtained. Solve for the standard deviation Solve for skewness Solving for kurtosis .

[0031] In this implementation example, kurtosis is used to measure the concentration of extreme amplitude samples in the residual signal, and it is highly sensitive to a small number of high-amplitude, transient impulse noises. Compared with second-order statistics such as variance, kurtosis can effectively distinguish between stationary harmonic disturbances and sudden impulse noise, avoiding false detections under strong harmonic conditions. Introducing kurtosis into the adaptive threshold adjustment process can dynamically reflect changes in impulse noise intensity in non-Gaussian, non-stationary noise environments, improving the robustness and accuracy of impulse noise detection.

[0032] Perform a Fast Fourier Transform on the residual vector:

[0033] in, The residual signal at the 1st Frequency domain representation at each frequency point; Complex exponential basis functions are used to implement the mapping from the time domain to the frequency domain; The length of the residual sequence participating in the transformation, i.e., the number of sampling points; Let be the residual signal sample value at the nth discrete time.

[0034] Calculate the spectral entropy:

[0035] in .

[0036] The aforementioned spectral entropy is used to measure the dispersion and disorder of the spectral energy distribution of the residual signal. When the residual is mainly composed of random impulse noise or broadband perturbations, its spectral energy distribution is more dispersed, and the spectral entropy increases significantly. Conversely, when the residual is mainly composed of harmonics or narrowband components, the spectral energy is concentrated at a few frequency points, and the spectral entropy is lower. Introducing spectral entropy as a feature quantity can effectively distinguish between structured signals and random noise components, providing a frequency domain basis for adaptive threshold adjustment, thereby improving the robustness and stability of impulse noise detection under non-stationary conditions.

[0037] Calculate the spectral centroid:

[0038] Based on residual statistical characteristics, the detection threshold parameter is dynamically adjusted to calculate an adaptive value. parameter:

[0039] in: Based on 4 Initial value selection for the principle, weighting coefficients , , Reference eigenvalues , .

[0040] Calculate the residual gradient vector:

[0041] Gradient norm .

[0042] The residual gradient vector is used to express the instantaneous changes in the residual signal between adjacent sampling points. Impulse noise typically exhibits abrupt amplitude changes and a steep rise time, causing a significant increase in the difference between adjacent samples in the time domain, while the changes in stationary harmonics and background noise are relatively gradual. By introducing the residual gradient vector, the sensitivity to sudden anomalies can be enhanced, the interference of slowly changing components can be effectively suppressed, and additional time-domain criteria can be provided for impulse noise detection, thereby improving detection accuracy and reducing the probability of false detection.

[0043] Formula for adjusting the β parameter:

[0044] in , , .

[0045] This adaptive adjustment formula for the β parameter achieves a joint expression of impulse noise intensity and abrupt changes by introducing a nonlinear mapping between kurtosis and the residual gradient norm. Specifically, the hyperbolic tangent function smooths and compresses the kurtosis, preventing drastic parameter fluctuations caused by extreme kurtosis values; the exponential decay term constrains the residual gradient norm, making β more sensitive to sudden, sharp changes and less responsive to steady-state disturbances. This design improves the stability and robustness of parameter adjustment while maintaining detection sensitivity, making it suitable for non-Gaussian and non-stationary noise environments.

[0046] After each residual update, the kurtosis γ² and the norm of the residual gradient vector are calculated based on the current residual. R∥, substitute into the above formula to update β in real time. adaptive The updated β parameter, together with the residual standard deviation, is used to construct the impulse noise detection threshold, thereby dynamically adjusting the detection sensitivity and guiding the identification and updating of the subsequent IPN support set.

[0047] The final detection threshold is calculated as follows:

[0048] Among them, adjustment factor Using a modified Sigmoid function:

[0049] in This is the sensitivity coefficient. The calculation method can dynamically adjust the threshold parameters according to the statistical characteristics of the residual signal, such as kurtosis, skewness, and spectral entropy, thereby adapting to the non-Gaussian and non-stationary changes in the statistical characteristics of background noise.

[0050] Apply an adaptive threshold to the residual vector to detect the location of IPN occurrence:

[0051] Record Support Set Size That is, the number of IPNs detected.

[0052] By applying an adaptive threshold directly to the residual vector, the impulse noise detection problem can be transformed into an anomaly location discrimination problem. Since IPNs (Instantaneous Prone to Noise) exhibit bursts in the time domain with amplitudes significantly higher than background noise and modeling errors, the adaptive threshold can reliably distinguish anomalies while comprehensively considering changes in residual statistical characteristics. This approach eliminates the need for a pre-defined fixed noise model, effectively adapts to non-Gaussian and non-stationary conditions, reduces the probability of false positives and false negatives, and provides explicit noise location constraints for subsequent joint estimation.

[0053] After obtaining the current residual vector, an adaptive threshold τ is used. new For each residual sample, perform point-by-point comparisons to satisfy |r n ∣>τ new The sampling locations are identified as impulse noise points and collected to form a support set Φ. The size of the support set |Φ| is the number of detected IPNs. This result is used to construct or update the enhancement basis matrix and participates in the subsequent joint estimation process of harmonic parameters and impulse noise amplitude.

[0054] Step 3: Parallel acceleration of joint estimation solution.

[0055] Based on the detected IPN support set Constructing an enhanced basis matrix :

[0056] in It is a subset of the columns of the identity matrix, and the column indices are determined by... specified.

[0057] The corresponding joint estimation vector is:

[0058] in It is a non-zero IPN magnitude vector. The joint estimation model is: , It is the background noise vector.

[0059] Calculate the enhancement basis matrix The pseudo-inverse is divided into four sub-blocks that can be executed in parallel: Calculate sub-block 1 :

[0060] Among them, the intermediate matrix

[0061] Calculate sub-block 2 :

[0062] Calculate sub-block 3 :

[0063] Calculate sub-block 4 :

[0064] One thread block is allocated for each sub-block calculation. Number of thread blocks:

[0065] in The total number of threads. For the number of processors. Memory access optimization can be achieved through the following methods: using shared memory caching for frequently accessed small matrices, such as... It optimizes global memory access by adopting a merged access mode and utilizes texture memory to cache constant matrices.

[0066] In this implementation example, by dividing the computation process of the enhanced basis matrix pseudoinverse into multiple independent sub-blocks and solving each sub-block in parallel, the structural characteristics of matrix operations can be fully utilized. The highly complex overall pseudoinverse operation is transformed into several low-dimensional matrix multiplications and inverse operations, significantly reducing the computational burden. Each sub-block is used to construct the intermediate matrix, compensate for cross terms, and fuse the results, providing the necessary components for the accurate reconstruction of the final pseudoinverse matrix. This block-parallel strategy not only avoids the problem of a sharp increase in computation time as the number of IPNs increases, but also improves data throughput efficiency through memory access optimization. This allows joint harmonic and impulse noise estimation to meet the requirements of online real-time computation while maintaining accuracy, demonstrating good engineering practicality.

[0067] Combine the four sub-blocks into a complete pseudo-inverse matrix:

[0068] The joint estimate is solved using the pseudo-inverse matrix computed in parallel:

[0069] Extracting harmonic parameter estimation (forward (element) and IPN magnitude estimation (back (elements).

[0070] Step 4: Iterative optimization and convergence.

[0071] Calculate the residual for the current i-th iteration:

[0072] in, Let be the residual vector of the i-th iteration. and Let be the enhanced basis matrix and the joint estimation vector for the i-th iteration, respectively.

[0073] Detect new IPN location:

[0074] in, Let i be the set of new IPN locations detected in the i-th iteration. Let n be the component of the residual vector at the nth sampling point in the i-th iteration. The adaptive detection threshold is calculated in the i-th iteration.

[0075] Updated support set:

[0076] in, This is the IPN support set after the (i+1)th iteration update; Convergence criteria are based on the rate of change of the residual norm:

[0077] in This is the convergence threshold. Simultaneously, changes in the support set are checked. The iteration terminates when both conditions are met, and the final estimate is output.

[0078] The method of this invention outputs the following set of parameters within each observation time window:

[0079] in, Indicates at time The effective amplitude of the h-th harmonic. This indicates that the h-th harmonic occurs at time [time]. Phase estimation, This indicates that the amplitude of the h-th harmonic is within the range of... This represents a trend of change within a short timescale centered on [the subject / entity]. This indicates that the h-th harmonic occurs at time [time]. Frequency estimation, This indicates that the amplitude of the h-th harmonic is within the range of... It is a secondary change trend within a short timescale centered on the subject.

[0080] In this implementation example, harmonics are estimated by establishing a Taylor-Fourier parametric model from the voltage or current sampling signals of the distribution network, and then performing joint parameter estimation of the signals under strong impulse noise interference. Specifically, Taylor-Fourier basis functions for each harmonic are constructed using the known fundamental frequency. The complex phasors of each harmonic at the center of the observation window, along with their first and second-order time-varying parameters, are obtained through least squares and enhanced modeling methods. The final estimation results include the amplitude and phase of each harmonic, as well as the dynamic changes in harmonic amplitude and phase. This enables high-precision, dynamic measurement of the harmonic level and broadband oscillation characteristics of the distribution network, providing a reliable basis for power quality monitoring and analysis.

[0081] Example 2 The purpose of this embodiment is to provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the above-described method.

[0082] Example 3 The purpose of this embodiment is to provide a computer-readable storage medium.

[0083] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the steps of the above-described method.

[0084] Example 4 The purpose of this embodiment is to provide a distribution network harmonic measurement system based on adaptive detection and parallel computing, including: The signal acquisition module is configured to acquire power distribution network signals and perform discretization sampling within the observation time window to obtain a discrete signal sequence; The basis matrix construction module is configured to: calculate the angular frequencies of each harmonic based on the fundamental frequency, and construct the Taylor-Fourier basis matrix; Establish a discrete signal model based on discrete signal sequences and Taylor-Fourier basis matrices; The initial estimation module is configured to: calculate the initial harmonic parameter estimates in the discrete signal model and calculate the initial residual vector; The statistical feature extraction module is configured to: calculate the kurtosis, spectral entropy, and residual gradient norm of the initial residual vector; The adaptive threshold generation module is configured to dynamically adjust the adaptive threshold parameters based on the kurtosis, spectral entropy, and residual gradient norm. Calculate the impulse noise detection threshold based on the adaptive threshold parameter; The impulse noise detection module is configured to: apply the impulse noise detection threshold to perform point-by-point comparison of the initial residual vector, identify the location of impulse noise occurrence, and form an impulse noise support set; The enhanced basis matrix construction module is configured to: construct an enhanced basis matrix based on the impulse noise support set; The parallel computing and output module is configured to: decompose the pseudo-inverse of the enhanced basis matrix into four sub-block matrices, perform parallel computing on the four sub-block matrices, and finally obtain the amplitude, phase, first derivative and second derivative of each harmonic.

[0085] Example 5 The purpose of this embodiment is to provide a computer program product containing instructions that, when run on a computer, cause the computer to perform the methods and functions involved in any of the above embodiments.

[0086] The steps and methods involved in the apparatus of the above embodiments correspond to those in Embodiment 1. For specific implementation details, please refer to the relevant description section of Embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood as including any medium capable of storing, encoding, or carrying an instruction set for execution by a processor and enabling the processor to perform any of the methods in this invention.

[0087] Those skilled in the art will understand that the modules or steps of the present invention described above can be implemented using general-purpose computer devices. Optionally, they can be implemented using computer-executable program code, thereby allowing them to be stored in a storage device for execution by a computer device, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. The present invention is not limited to any particular combination of hardware and software.

[0088] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A method for measuring harmonics in distribution networks based on adaptive detection and parallel computing, characterized in that, include: The distribution network signal is acquired and discretized within the observation time window to obtain a discrete signal sequence; Calculate the angular frequencies of each harmonic based on the fundamental frequency, and construct the Taylor-Fourier basis matrix. Establish a discrete signal model based on discrete signal sequences and Taylor-Fourier basis matrices; Calculate the initial harmonic parameter estimates in the discrete signal model and calculate the initial residual vector; Calculate the kurtosis, spectral entropy, and residual gradient norm of the initial residual vector; The adaptive threshold parameter is dynamically adjusted based on the kurtosis, spectral entropy, and residual gradient norm. Calculate the impulse noise detection threshold based on the adaptive threshold parameter; The initial residual vector is compared point by point using the impulse noise detection threshold to identify the location of impulse noise and form an impulse noise support set. Construct an enhancement basis matrix based on the impulse noise support set; The pseudo-inverse of the enhancement basis matrix is ​​decomposed into four sub-block matrices, and the four sub-block matrices are computed in parallel to obtain the amplitude, phase, first derivative and second derivative of each harmonic.

2. The method for measuring harmonics in a distribution network based on adaptive detection and parallel computing as described in claim 1, characterized in that, After performing parallel computation on the four sub-block matrices, the process also includes: The four sub-block matrices obtained by parallel computing are combined into a complete pseudo-inverse matrix. The joint estimate is then solved using the complete pseudo-inverse matrix to extract the harmonic parameter estimate and the impulse noise amplitude estimate. Calculate the current iteration residual vector, detect new impulse noise locations, and update the impulse noise support set; The convergence is determined based on the norm change rate of the current iterative residual vector. When the convergence condition is met, the amplitude, phase, first derivative, and second derivative of each harmonic are output.

3. The method for measuring harmonics in a distribution network based on adaptive detection and parallel computing as described in claim 1, characterized in that, The kurtosis is calculated using the fourth-order central moment normalization formula, the spectral entropy is calculated based on the fast Fourier transform result of the initial residual vector, and the residual gradient norm is calculated using the first-order difference between adjacent residual samples.

4. The method for measuring harmonics in a distribution network based on adaptive detection and parallel computing as described in claim 1, characterized in that, The calculation steps for dynamically adjusting the adaptive threshold parameter include: Based on the statistical characteristics of the residuals, the detection threshold parameters are dynamically adjusted, and the first adaptive parameter is calculated. The second adaptive parameter is calculated by introducing a nonlinear mapping between kurtosis and residual gradient norm.

5. The method for measuring harmonics in a distribution network based on adaptive detection and parallel computing as described in claim 1, characterized in that, The formula for calculating the impulse noise detection threshold is: in, This is an adjustment factor.

6. The method for measuring harmonics in a distribution network based on adaptive detection and parallel computing as described in claim 1, characterized in that, The enhancement basis matrix is ​​a concatenation matrix of a Taylor-Fourier basis matrix and a column subset of the identity matrix, wherein the column index of the column subset of the identity matrix is ​​specified by the impulse noise support set.

7. A method for measuring harmonics in distribution networks based on adaptive detection and parallel computing, characterized in that, include: The signal acquisition module is configured to acquire power distribution network signals and perform discretization sampling within the observation time window to obtain a discrete signal sequence; The basis matrix construction module is configured to: calculate the angular frequencies of each harmonic based on the fundamental frequency, and construct the Taylor-Fourier basis matrix; Establish a discrete signal model based on discrete signal sequences and Taylor-Fourier basis matrices; The initial estimation module is configured to: calculate the initial harmonic parameter estimates in the discrete signal model and calculate the initial residual vector; The statistical feature extraction module is configured to: calculate the kurtosis, spectral entropy, and residual gradient norm of the initial residual vector; The adaptive threshold generation module is configured to dynamically adjust the adaptive threshold parameters based on the kurtosis, spectral entropy, and residual gradient norm. Calculate the impulse noise detection threshold based on the adaptive threshold parameter; The impulse noise detection module is configured to: apply the impulse noise detection threshold to perform point-by-point comparison of the initial residual vector, identify the location of impulse noise occurrence, and form an impulse noise support set; The enhanced basis matrix construction module is configured to: construct an enhanced basis matrix based on the impulse noise support set; The parallel computing and output module is configured to: decompose the pseudo-inverse of the enhanced basis matrix into four sub-block matrices, perform parallel computing on the four sub-block matrices, and finally obtain the amplitude, phase, first derivative and second derivative of each harmonic.

8. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 6.

9. A computer device 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 program, it implements the steps of the method described in any one of claims 1-6.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it performs the steps of the method described in any one of claims 1-6 above.

Images