Intelligent power distribution monitoring system and method

Abstract

The invention relates to the technical field of data monitoring, and discloses an intelligent power distribution monitoring system and method, and the method comprises the steps: collecting current and voltage signals of a preset point location, and building a signal data link; preprocessing the signal data chain based on a windowing function to obtain a windowed data chain, and determining an overlapping proportion according to the length of the windowed data chain to perform window division to obtain a plurality of frequency spectrums; the spectrums are compared in pairs, and whether abnormal early warning is carried out or not is judged according to the comparison result; when early warning is judged, processing each frequency spectrum based on Hilbert transform to obtain an instantaneous envelope and an instantaneous phase, and obtaining a phase difference value of two harmonics with the same frequency in all the frequency spectrums; and comparing the phase difference value and the spectrum similarity as feature data with a historical early warning data set for early warning. According to the method, the problems that a traditional method depends on a fixed threshold value and is slow in response and high in false alarm rate are solved, and the identification accuracy and response timeliness of dynamic harmonic abnormity are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data monitoring, and in particular, to an intelligent power distribution monitoring system and method. Background Art

[0002] With the access of new energy, the wide application of power electronic devices, and the diversification of user loads, the problem of harmonic pollution in the power distribution system has become increasingly prominent, seriously affecting the power quality, safe operation, and equipment life. To improve the intelligent operation and maintenance ability of the power distribution system, the power system urgently needs to build an intelligent power distribution monitoring method with real-time monitoring, abnormal identification, and fault warning capabilities.

[0003] However, the traditional FFT method has weak processing ability for non-stationary or disturbed signals and is difficult to accurately identify dynamic harmonic anomalies. Direct FFT processing of unwindowed signals is prone to spectral leakage, affecting the measurement accuracy of amplitude and phase. Moreover, the current focus is mostly on harmonic amplitude characteristics, and the evolution law of the phase difference between voltage and current signals has not been effectively explored, resulting in inaccurate determination of the harmonic source direction and propagation path. In addition, it relies on a fixed threshold judgment mechanism, lacks the ability to learn and adapt to historical data, and is prone to false alarms and missed alarms.

[0004] Therefore, it is necessary to design an intelligent power distribution monitoring system and method to solve the problems existing in the current technology. Summary of the Invention

[0005] In view of this, the present invention proposes an intelligent power distribution monitoring system and method, aiming to solve the problems of difficult capture of current time-varying characteristics, spectral leakage and aliasing, low utilization rate of phase information, and non-intelligent warning mechanism.

[0006] On the one hand, the present invention proposes an intelligent power distribution monitoring method, including:

[0007] Collecting current and voltage signals at preset points to establish a signal data chain, where the preset points include a distribution transformer and a user incoming line end, the signal data chain includes a plurality of acquisition time points and a plurality of current and voltage signals, and each acquisition time point corresponds to a current and voltage signal;

[0008] Preprocessing the signal data chain based on a window function to obtain a windowed data chain, determining an overlap ratio according to the length of the windowed data chain for window division, and obtaining a plurality of spectra;

[0009] Comparing the spectra pairwise, and judging whether to issue an abnormal warning according to the comparison result; when it is determined to issue a warning, processing each spectrum based on the Hilbert transform to obtain an instantaneous envelope and an instantaneous phase, and obtaining the phase difference between two harmonics with the same frequency in all the spectra;

[0010] Take the phase difference and the spectral similarity as feature data, compare the feature data with the historical early warning data set, and give an early warning according to the comparison result.

[0011] Further, when collecting the current and voltage signals at preset points to establish a signal data chain, it includes:

[0012] Perform wavelet transform on each of the current and voltage signals to extract high-frequency signals;

[0013] Use Daubechies 6 wavelet to perform discrete wavelet transform and decompose the current and voltage signals into 6 layers;

[0014] Use the soft threshold method to extract the detail signals, and set the threshold to 1.5 times the signal standard deviation;

[0015] Perform inverse wavelet transform on the extracted signals and then establish the signal data chain.

[0016] Further, when preprocessing the signal data chain based on a window function to obtain a windowed data chain, the window function includes a Hann window or a Hamming window.

[0017] Further, when determining the overlapping ratio for window division according to the length of the windowed data chain to obtain several spectra, it includes:

[0018] Compare the length of the windowed data chain with a preset length threshold, determine the overlapping ratio according to the comparison result, and the overlapping ratio is directly proportional to the length of the windowed data chain.

[0019] Further, when comparing the spectra pairwise and judging whether to give an abnormal early warning according to the comparison result, it includes:

[0020] Obtain the spectral similarity of each pair of spectra based on the Euclidean distance, compare the spectral similarity with a similarity threshold, and judge whether to give an abnormal early warning according to the comparison result;

[0021] When the spectral similarity is less than the similarity threshold, it is determined to give an abnormal early warning;

[0022] When the spectral similarity is greater than or equal to the similarity threshold, it is determined not to give an abnormal early warning.

[0023] Further, when processing each spectrum based on the Hilbert transform to obtain the instantaneous envelope and instantaneous phase, it includes:

[0024]

[0025] Among them, z(t) represents the analytic signal, and x(t) represents the original real signal, denotes the signal after Hilbert transform, which is the orthogonal component of x(t), A(t) denotes the instantaneous envelope, and φ(t) denotes the instantaneous phase.

[0026] Further, when comparing the feature data with the historical warning data set and giving a warning according to the comparison result, it includes:

[0027] The historical warning data set includes a number of historical feature data and a number of historical warning levels, and each of the historical feature data corresponds to a historical warning level;

[0028] Initialize K centroids among all the historical feature data, assign the feature data to the nearest centroid to form K clusters; repeatedly execute the assignment of centroids and the calculation of the centroids of each cluster after assignment until the centroids no longer change;

[0029] When there is historical feature data in the cluster containing the feature data, use the historical features in this cluster as a similar set, and determine the historical warning level corresponding to the maximum value of the feature similarity in the similar set to give a warning;

[0030] When there is no historical feature data in the cluster containing the feature data, give a warning based on the prediction result of the long short-term prediction model.

[0031] Further, when giving a warning based on the prediction result of the long short-term prediction model, it includes:

[0032] Sample the historical warning data set according to a preset ratio to obtain a training subset and a test subset;

[0033] Iteratively train the long short-term prediction model LSTM with the training subset, evaluate the iteratively trained long short-term prediction model LSTM according to the test subset, and obtain the final long short-term prediction model LSTM;

[0034] Input the feature data into the trained prediction model to obtain the prediction result.

[0035] Compared with the prior art, the beneficial effects of the present invention are as follows: By synchronously collecting current and voltage signals at preset key points such as the distribution transformer and the user incoming line end, a time-continuous signal data chain is established, enabling a comprehensive grasp of the propagation characteristics of harmonics in the distribution network; By introducing a windowing function and overlapping averaging preprocessing method, spectral leakage and boundary effects are effectively suppressed, improving the accuracy and stability of spectral analysis; The Hilbert transform is used to extract the instantaneous envelope and phase information of the signal, further realizing the accurate calculation of the phase difference between the same-frequency harmonic signals, so as to identify the harmonic source direction and propagation path; Taking spectral similarity and phase difference as the core features, combined with the historical warning data set for comparison, an adaptive anomaly recognition and warning mechanism is realized, overcoming the problems of traditional methods relying on fixed thresholds, slow response and high false alarm rate, and improving the recognition accuracy and response timeliness of dynamic harmonic anomalies.

[0036] On the other hand, the present application also provides an intelligent power distribution monitoring system for applying the above intelligent power distribution monitoring method, including:

[0037] An acquisition unit configured to collect current and voltage signals at preset points to establish a signal data chain, the preset points including a distribution transformer and a user incoming line end, the signal data chain including a plurality of acquisition time points and a plurality of current and voltage signals, and each acquisition time point corresponding to a current and voltage signal;

[0038] A processing unit configured to preprocess the signal data chain based on a windowing function to obtain a windowed data chain, determine an overlapping ratio according to the length of the windowed data chain for window division, and obtain a plurality of spectra;

[0039] A judgment unit configured to compare the spectra pairwise, and judge whether to issue an anomaly warning according to the comparison result; When it is determined to issue a warning, each spectrum is processed based on the Hilbert transform to obtain an instantaneous envelope and an instantaneous phase, and the phase difference between two same-frequency harmonics in all the spectra is obtained;

[0040] An early warning unit configured to use the phase difference and spectral similarity as characteristic data, compare the characteristic data with a historical warning data set, and issue a warning according to the comparison result.

[0041] Furthermore, it further includes a storage unit configured to store the characteristic data and prediction results.

[0042] It can be understood that the above intelligent power distribution monitoring system and method have the same beneficial effects, which will not be elaborated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] By reading the following detailed description of the preferred embodiments, various other advantages and benefits will become clear to those of ordinary skill in the art. The drawings are only for the purpose of showing the preferred embodiments and are not considered to be a limitation of the present invention. Moreover, throughout the drawings, the same reference symbols are used to represent the same components. In the drawings:

[0044] Figure 1 is a flowchart of the intelligent power distribution monitoring method provided by an embodiment of the present invention;

[0045] Figure 2 is a functional block diagram of the intelligent power distribution monitoring system provided by an embodiment of the present invention. Specific Embodiments

[0046] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the drawings. Although the exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be fully conveyed to those skilled in the art. It should be noted that, without conflict, the embodiments in the present invention and the features in the embodiments can be combined with each other. The present invention will be described in detail below with reference to the drawings and in combination with the embodiments.

[0047] In some embodiments of the present application, referring to Figure 1 as shown, an intelligent power distribution monitoring method includes:

[0048] S100: Collect current and voltage signals at preset points to establish a signal data chain. The preset points include a distribution transformer and a user incoming line end. The signal data chain includes a number of acquisition time points and a number of current and voltage signals, and each acquisition time point corresponds to a current and voltage signal.

[0049] S200: Preprocess the signal data chain based on a window function to obtain a windowed data chain, determine an overlap ratio according to the length of the windowed data chain for window division, and obtain a number of spectra.

[0050] S300: Compare the spectra pairwise, and determine whether to issue an abnormal warning according to the comparison result. When it is determined to issue a warning, process each spectrum based on the Hilbert transform to obtain an instantaneous envelope and an instantaneous phase, and obtain the phase difference between two harmonics with the same frequency in all spectra.

[0051] S400: Use the phase difference and the spectrum similarity as feature data, compare the feature data with a historical warning data set, and issue a warning according to the comparison result.

[0052] Specifically, first, sampling points are set at key nodes such as distribution transformers and user incoming line ends to collect current and voltage signals at each moment, and a signal data chain containing time series is constructed to obtain the propagation characteristics of medium harmonics in space and time. The signal data chain is preprocessed using window functions (such as Hamming window, Hanning window, etc.) to reduce spectral leakage while maintaining frequency resolution. The windowed data chain is divided into windows by setting an overlapping ratio to further improve spectral continuity and accuracy. After performing spectral analysis on the signals within each window, a pairwise spectral comparison operation is carried out. Through the comparison of spectral structure and energy distribution, potential anomalies are initially identified. When an abnormal trend is judged to exist, a Hilbert transform is further applied to the spectral signal to extract the instantaneous envelope and instantaneous phase of each frequency component, and then the phase difference between key frequency harmonics is calculated to infer the source of harmonics and the abnormal influence path. The obtained phase difference and spectral similarity are used as a combined feature vector to perform a similarity comparison with the historical early warning data set, realizing an adaptive early warning judgment mechanism that integrates historical experience, thereby improving the accuracy and response ability of early warning.

[0053] It can be understood that a highly robust intelligent power distribution monitoring mechanism is constructed through multi-dimensional fusion of time domain and frequency domain, amplitude and phase, and current and historical data. Compared with the traditional passive recognition method that only relies on amplitude thresholds, the windowed spectral processing and phase difference discrimination mechanism are introduced to improve the perception ability of the dynamic evolution characteristics of harmonics, and through a data-driven adaptive comparison algorithm, accurate identification and timely early warning of abnormal events are realized.

[0054] In some embodiments of the present application, when collecting current and voltage signals at preset points to establish a signal data chain, it includes: performing wavelet transform on each current and voltage signal to extract high-frequency signals. Using Daubechies6 wavelet for discrete wavelet transform, the current and voltage signals are decomposed into 6 layers of wavelets. Using the soft threshold method, the detail signals are extracted, and the threshold is set to 1.5 times the standard deviation of the signal. After performing inverse wavelet transform on the extracted signals, a signal data chain is established.

[0055] Specifically, the Daubechies 6 (db6) wavelet function with good local characteristics and stability is selected as the mother wavelet, and 6-layer wavelet decomposition is performed on each sampling signal. The original signal is decomposed into a low-frequency approximation component (A6) and multiple high-frequency detail components (D1 to D6). The high-frequency part is more sensitive to abnormal information such as harmonics and pulse interference. On this basis, in order to suppress noise and retain high-frequency abnormal features, a soft-threshold denoising method is used to process the detail signals. A threshold is set, which is equal to 1.5 times the standard deviation of the detail signal (i.e., λ = 1.5σ). For the detail components with amplitudes lower than the threshold, they are directly weakened to zero, and for those higher than the threshold, they are smoothed and reduced, retaining their main energy and reducing sharp transitions. The soft-threshold denoising strategy takes into account both noise reduction and signal integrity, and is superior to the hard threshold in terms of continuity and smoothness. After processing, the wavelet inverse transform is performed on the denoised signal to reconstruct the denoised voltage and current signals. The finally formed signal data chain has a higher signal-to-noise ratio and dynamic abnormal response ability, providing a cleaner and more sensitive signal input for subsequent processing steps such as windowing, spectral analysis, and Hilbert transform.

[0056] It can be understood that by utilizing the multi-scale characteristics of wavelets, the accurate extraction of high-frequency abnormal components and noise suppression of power signals are realized, effectively enhancing the information expression ability related to harmonic anomalies in the original signal. Through the advantages of the db6 wavelet in edge detection and harmonic capture, and the smoothness guarantee of the soft threshold in the signal reconstruction process, the established signal data chain not only retains the key feature information, but also significantly reduces the interference caused by high-frequency noise to subsequent analysis, improving the stability and accuracy of the entire intelligent power distribution monitoring system.

[0057] In some embodiments of the present application, when preprocessing the signal data chain based on a window function to obtain the windowed data chain, the window function includes a Hanning window or a Hamming window.

[0058] In some embodiments of the present application, when determining the overlapping ratio for window division according to the length of the windowed data chain to obtain several spectra, it includes: comparing the length of the windowed data chain with a preset length threshold, and determining the overlapping ratio according to the comparison result. The overlapping ratio is directly proportional to the length of the windowed data chain.

[0059] Specifically, the Hanning window: The Hamming window: n represents the sampling point index within the window, and N represents the window length. The window function attenuates the signal at both ends by weighting, effectively eliminating the mutations caused by truncation, thereby improving the accuracy of the spectral energy distribution. The windowed data chain is further divided into multiple analysis windows for spectral transformation. During the window division process, a strategy of adaptively adjusting the overlap ratio based on the data chain length is introduced, that is, comparing the length of the currently windowed data chain with a preset length threshold. If the data chain is long, the overlap ratio is increased to enhance the time continuity and spectral smoothness of the analysis; if the data chain is short, the overlap ratio is appropriately reduced to ensure the frequency resolution and computational efficiency. A dynamic adjustment mechanism in which the overlap ratio is proportional to the data chain length is implemented, which not only maintains the accuracy of spectral estimation but also avoids redundant calculations caused by excessive overlap.

[0060] It can be understood that by selecting the window function, the influence of spectral leakage is weakened, and at the same time, the flexibility and adaptive ability of spectral analysis are enhanced by dynamically setting the overlap ratio in combination with the signal length. Compared with the traditional fixed window length and overlap ratio processing method, the processing parameters can be intelligently adjusted according to the data volume of the collected signal under different operating scenarios, making the spectral extraction process more robust and efficient.

[0061] In some embodiments of the present application, when comparing spectra pairwise and determining whether to issue an abnormal warning based on the comparison results, it includes: obtaining the spectral similarity of each pair of spectra based on the Euclidean distance, comparing the spectral similarity with a similarity threshold, and determining whether to issue an abnormal warning according to the comparison results. When the spectral similarity is less than the similarity threshold, it is determined that an abnormal warning is to be issued. When the spectral similarity is greater than or equal to the similarity threshold, it is determined that no abnormal warning is to be issued.

[0062] It can be understood that then the spectral similarity is deduced from the distance magnitude. The smaller the distance, the closer the two spectral shapes are, and the higher the similarity. Conversely, it indicates that the spectral structure has changed, and there may be abnormal power quality or harmonic disturbances. During the comparison process, the obtained spectral similarity is compared with a preset similarity threshold: if the similarity is lower than the threshold (i.e., the spectral difference is large), it is judged that the current spectrum has an abnormal trend that significantly deviates from the historical or normal state, triggering a warning; if the similarity is greater than or equal to the threshold (the spectral structure has not changed significantly), it is considered that the operation is normal and no warning is issued. The stability of the system is judged by the relative change degree of the global characteristics of the spectral shape, avoiding the misjudgment risk of simply relying on the amplitude of a single frequency point or instantaneous jump.

[0063] In some embodiments of the present application, when processing each spectrum based on the Hilbert transform to obtain the instantaneous envelope and instantaneous phase, it includes:

[0064]

[0065] Among them, \(z(t)\) represents the analytic signal, and \(x(t)\) represents the original real signal. represents the signal after Hilbert transform, which is the orthogonal component of \(x(t)\). \(A(t)\) represents the instantaneous envelope, and \(\varphi(t)\) represents the instantaneous phase.

[0066] It can be understood that \(j\) represents the imaginary unit. By using the Hilbert transform to extract the instantaneous envelope and instantaneous phase of the corresponding frequency harmonics from each spectrum, the phase difference of the same frequency harmonics between different monitoring points can be accurately calculated subsequently, which is crucial for analyzing the harmonic source direction, identifying system disturbances, and determining the abnormal trend of signals. By constructing an analytic signal with the Hilbert transform and extracting two types of instantaneous features, the envelope and phase of the signal, compared with static frequency domain features (such as amplitude and energy), it can reflect the dynamic evolution process of the signal on the time axis, especially for abnormalities such as transient events and non-stationary disturbances, with higher recognition sensitivity and time positioning accuracy, further improving the analytic ability and early warning accuracy of the intelligent distribution monitoring system for harmonic dynamic behavior.

[0067] In some embodiments of the present application, when comparing the feature data with the historical early warning data set and giving an early warning according to the comparison result, it includes: The historical early warning data set includes several historical feature data and several historical early warning levels, and each historical feature data corresponds to a historical early warning level. Initialize \(K\) centroids among all historical feature data, assign the feature data to the nearest centroid to form \(K\) clusters. Recalculate the centroid of each cluster. Repeat the operations of assigning centroids and calculating the centroids of the assigned clusters until the centroids no longer change.

[0068] Specifically, when there is historical feature data in the cluster containing the feature data, use the historical features in this cluster as a similarity set, and determine the historical early warning level corresponding to the maximum feature similarity value in the similarity set for early warning. When there is no historical feature data in the cluster containing the feature data, give an early warning based on the prediction result of the long short-term prediction model.

[0069] In some embodiments of the present application, when giving an early warning based on the prediction result of the long short-term prediction model, it includes: Sampling the historical early warning data set according to a preset ratio to obtain a training subset and a test subset. Iteratively train the long short-term prediction model LSTM with the training subset, evaluate the iteratively trained long short-term prediction model LSTM according to the test subset, and obtain the final long short-term prediction model LSTM. Input the feature data into the trained prediction model to obtain the prediction result.

[0070] Specifically, there are several groups of historical feature data, each group consisting of multidimensional values such as "phase difference + spectral similarity"; the corresponding historical warning levels (such as normal, slightly abnormal, severely abnormal, etc.) serve as label values. K-means mines the internal distribution structure in historical data through unsupervised learning to achieve automatic clustering and similarity matching without thresholds; LSTM, based on its time series learning ability, can capture the evolution trend of feature values over time for dynamic prediction and completion. This not only ensures the immediacy of early warnings but also improves the inference ability for unseen scenarios.

[0071] It can be understood that by combining the two machine learning methods of K-means and LSTM, high-dimensional intelligent discrimination of spectral features and phase features in voltage and current signals is achieved, without relying on fixed thresholds and being adaptable to various working conditions and equipment states. At the same time, historical experience data is introduced for analog enhancement, effectively improving the accuracy and stability of early warning judgments and enhancing the monitoring and early warning capabilities of the distribution system for complex non-linear harmonic disturbances.

[0072] In the above embodiments, by synchronously collecting current and voltage signals at preset key points such as distribution transformers and user incoming line ends to establish a time-continuous signal data chain, the propagation characteristics of harmonics in the distribution network can be comprehensively grasped. By introducing preprocessing methods such as windowing functions and overlapping averaging, spectral leakage and boundary effects are effectively suppressed, improving the accuracy and stability of spectral analysis. Hilbert transform is used to extract the instantaneous envelope and phase information of the signal, further enabling accurate calculation of the phase difference between same-frequency harmonic signals, thereby identifying the harmonic source direction and propagation path. With spectral similarity and phase difference as the core features, comparison with the historical early warning data set is carried out to achieve an adaptive abnormal identification and early warning mechanism, overcoming the problems of traditional methods relying on fixed thresholds, slow response, and high false alarm rates, and improving the identification accuracy and response timeliness for dynamic harmonic anomalies.

[0073] In another preferred embodiment based on the above embodiments, refer to Figure 2 As shown, this embodiment provides an intelligent distribution monitoring system for applying the above intelligent distribution monitoring method, including:

[0074] An acquisition unit configured to acquire current and voltage signals at preset points to establish a signal data chain. The preset points include distribution transformers and user incoming line ends. The signal data chain includes several acquisition time points and several current and voltage signals, and each acquisition time point corresponds to a current and voltage signal.

[0075] A processing unit configured to preprocess the signal data chain based on a windowing function to obtain a windowed data chain, determine an overlapping ratio according to the length of the windowed data chain for window division, and obtain several spectra.

[0076] A judgment unit configured to compare spectra pairwise and determine whether to give an abnormal warning according to the comparison results. When it is determined to give a warning, each spectrum is processed based on Hilbert transform to obtain the instantaneous envelope and instantaneous phase, and the phase difference between two harmonics with the same frequency in all spectra is obtained.

[0077] A warning unit configured to use the phase difference and spectrum similarity as characteristic data, compare the characteristic data with a historical warning data set, and give a warning according to the comparison results.

[0078] Further, it further includes a storage unit configured to store the characteristic data and prediction results.

[0079] It can be understood that by synchronously collecting current and voltage signals at preset key points such as distribution transformers and user incoming line ends to establish a time-continuous signal data chain, the propagation characteristics of harmonics in the distribution network can be comprehensively grasped. By introducing the preprocessing methods of window function and overlapping averaging, spectrum leakage and boundary effects are effectively suppressed, and the accuracy and stability of spectrum analysis are improved. Hilbert transform is used to extract the instantaneous envelope and phase information of the signal, further realizing the accurate calculation of the phase difference between harmonics with the same frequency, so as to identify the harmonic source direction and propagation path. With spectrum similarity and phase difference as the core features, combined with comparison with the historical warning data set, an adaptive abnormal recognition and warning mechanism is realized, overcoming the problems of traditional methods relying on fixed thresholds, slow response and high false alarm rate, and improving the recognition accuracy and response timeliness of dynamic harmonic anomalies.

[0080] Those skilled in the art should understand that the embodiments of the present application can be provided as methods, systems or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0081] The present application is described with reference to the flowcharts and / or block diagrams of methods, devices (systems) and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram can be realized by computer program instructions, and the combination of processes and / or blocks in the flowchart and / or block diagram can also be realized by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate for realizing in the process Figure 1 each process or multiple processes and / or blocks Figure 1a device for the functions specified in one or more boxes

[0082] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured article including an instruction device that implements the operations in the process Figure 1 one process or more processes and / or boxes Figure 1 the functions specified in one box or more boxes

[0083] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to produce a computer-implemented process. Thus, the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in the process Figure 1 one process or more processes and / or boxes Figure 1 the functions specified in one box or more boxes

[0084] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: it is still possible to modify the specific implementation manners of the present invention or make equivalent replacements. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention shall be covered by the protection scope of the claims of the present invention.

Claims

1. An intelligent power distribution monitoring method, characterized in that, Including: Collecting current and voltage signals at preset points to establish a signal data chain. The preset points include distribution transformers and user incoming line terminals. The signal data chain includes several acquisition time points and several current and voltage signals, and each acquisition time point corresponds to a current and voltage signal; Preprocessing the signal data chain based on a window function to obtain a windowed data chain, determining an overlapping ratio according to the length of the windowed data chain for window division to obtain several spectra; Comparing the spectra pairwise and judging whether to give an abnormal warning according to the comparison result; When it is determined to give a warning, processing each spectrum based on the Hilbert transform to obtain the instantaneous envelope and instantaneous phase, and obtaining the phase difference between two harmonics with the same frequency in all the spectra; Using the phase difference and spectrum similarity as characteristic data, comparing the characteristic data with a historical warning data set, and giving a warning according to the comparison result.

2. The intelligent power distribution monitoring method according to claim 1, wherein When collecting current and voltage signals at preset points to establish a signal data chain, it includes: Performing wavelet transform on each current and voltage signal to extract high-frequency signals; Using Daubechies 6 wavelet to perform discrete wavelet transform and decomposing the current and voltage signal into 6 layers of wavelets; Using the soft threshold method to extract the detail signals, and setting the threshold to 1.5 times the signal standard deviation; Performing inverse wavelet transform on the extracted signals and then establishing the signal data chain.

3. The intelligent power distribution monitoring method according to claim 1, characterized in that, When preprocessing the signal data chain based on a window function to obtain a windowed data chain, the window function includes a Hanning window or a Hamming window.

4. The intelligent power distribution monitoring method according to claim 3, wherein, When determining the overlapping ratio according to the length of the windowed data chain for window division to obtain several spectra, it includes: Comparing the length of the windowed data chain with a preset length threshold, determining the overlapping ratio according to the comparison result, and the overlapping ratio is in a direct proportional relationship with the length of the windowed data chain.

5. The intelligent power distribution monitoring method according to claim 1, characterized in that, When comparing the spectra pairwise and judging whether to give an abnormal warning according to the comparison result, it includes: Obtaining the spectrum similarity of every two spectra based on the Euclidean distance, comparing the spectrum similarity with a similarity threshold, and judging whether to give an abnormal warning according to the comparison result; When the spectrum similarity is less than the similarity threshold, it is determined to give an abnormal warning; When the spectrum similarity is greater than or equal to the similarity threshold, it is determined not to give an abnormal warning.

6. The intelligent power distribution monitoring method according to claim 1, characterized in that, When processing each spectrum based on the Hilbert transform to obtain the instantaneous envelope and instantaneous phase, it includes: where \(z(t)\) represents the analytic signal, and \(x(t)\) represents the original real signal. represents the signal after Hilbert transform, which is the quadrature component of \(x(t)\), \(A(t)\) represents the instantaneous envelope, and \(\varphi(t)\) represents the instantaneous phase.

7. The intelligent power distribution monitoring method according to claim 6, wherein, When comparing the characteristic data with a historical warning data set and giving a warning according to the comparison result, it includes: The historical warning data set includes several historical characteristic data and several historical warning levels, and each historical characteristic data corresponds to a historical warning level; Initializing K centroids among all the historical characteristic data, allocating the characteristic data to the nearest centroid to form K clusters; recalculating the centroid of each cluster; repeating the allocation of centroids and the calculation of the centroids of each allocated cluster until the centroids no longer change; When there is historical feature data in the cluster containing the feature data, the historical features in the cluster are used as a similarity set, and an early warning is issued based on the historical warning level corresponding to the maximum feature similarity value in the similarity set; When there is no historical feature data in the cluster containing the feature data, an early warning is issued based on the prediction result of the long short-term prediction model.

8. The intelligent power distribution monitoring method according to claim 7, characterized in that, When issuing an early warning based on the prediction result of the long short-term prediction model, it includes: Sampling the historical warning data set according to a preset ratio to obtain a training subset and a test subset; Iteratively training the long short-term prediction model LSTM with the training subset, evaluating the iteratively trained long short-term prediction model LSTM according to the test subset, and obtaining the final long short-term prediction model LSTM; Inputting the feature data into the trained prediction model to obtain the prediction result.

9. An intelligent power distribution monitoring system for applying the intelligent power distribution monitoring method according to any one of claims 1-8, characterized in that It includes: An acquisition unit configured to acquire current and voltage signals at preset points to establish a signal data chain, where the preset points include a distribution transformer and a user incoming line end, the signal data chain includes a number of acquisition time points and a number of current and voltage signals, and each acquisition time point corresponds to a current and voltage signal; A processing unit configured to preprocess the signal data chain based on a window function to obtain a windowed data chain, determine an overlap ratio according to the length of the windowed data chain for window division, and obtain a number of spectra; A judgment unit configured to compare the spectra pairwise and judge whether to issue an abnormal early warning according to the comparison result; When it is determined to issue an early warning, each spectrum is processed based on the Hilbert transform to obtain an instantaneous envelope and an instantaneous phase, and the phase difference between two harmonics with the same frequency in all the spectra is obtained; An early warning unit configured to use the phase difference and the spectrum similarity as feature data, compare the feature data with the historical warning data set, and issue an early warning according to the comparison result.

10. The intelligent power distribution monitoring system according to claim 9, wherein, It further includes: A storage unit configured to store the feature data and the prediction result.

Images