Power distribution network operation state monitoring method and system based on multi-source data analysis

Abstract

The application discloses a power distribution network operation state monitoring method and system based on multi-source data analysis, and relates to the technical field of power distribution network operation state monitoring.The power distribution network operation state monitoring method based on multi-source data analysis comprises the following steps: time-frequency analysis effectiveness evaluation; blind source separation effectiveness evaluation; and power distribution network operation fault category identification.The application obtains a reflected wave signal through wavelet transform processing, evaluates the time-frequency analysis effectiveness according to a wavelet transform evaluation parameter, judges whether to perform wavelet transform optimization, then performs blind source separation processing to obtain a main reflected wave signal, evaluates the blind source separation effectiveness according to a blind source separation evaluation parameter, judges whether to perform blind source separation optimization, and finally identifies the power distribution network operation fault category according to the obtained time-frequency characteristics, thereby improving the time-frequency characteristic extraction efficiency in power distribution network operation state monitoring and solving the problem of low signal characteristic extraction efficiency in power distribution network operation state monitoring due to multi-source signal interference in the prior art.

Description

Technical Field

[0001] This invention relates to the field of power distribution network operation status monitoring technology, and in particular to a method and system for power distribution network operation status monitoring based on multi-source data analysis. Background Technology

[0002] Distribution networks are widely distributed and numerous. As the main carriers of electrical energy transmission, distribution lines are a crucial component of the power grid infrastructure. The safe operation of the distribution network is related to the stability of power supply and the normal functioning of social life. Distribution network faults or anomalies may lead to power outages or even more serious safety accidents. Therefore, ensuring the normal operation of the distribution network requires continuous, safe, and reliable monitoring. In monitoring the operational status of the distribution network, challenges such as signal overlap and complex interference necessitate combining traditional signal processing algorithms with modern machine learning techniques to achieve efficient and accurate signal separation and feature extraction.

[0003] Existing technologies integrate multi-dimensional data and employ state estimation techniques to estimate the real-time operating status of the distribution network. Finally, machine learning algorithms are used to analyze historical and real-time monitoring data of the distribution network to predict the possible location and timing of faults, thereby achieving an accurate assessment of the distribution network status.

[0004] For example, the invention patent announcement CN105515199B discloses a method for distributed fault detection in a smart distribution network, which includes: determining the protection range of each smart distribution terminal; each smart distribution terminal communicating with adjacent smart distribution terminals to exchange network topology information at their respective locations; the smart distribution terminal determining the fault location based on the fault information calculated by the adjacent smart distribution terminals and the regional longitudinal comparison protection principle; after determining the line where the fault location is located, determining the switch type corresponding to the line according to the network topology, and making corresponding fault isolation strategies according to the switch type.

[0005] For example, patent application CN116054415A discloses a distribution network monitoring system and distribution network, including: a main control module and at least one sub-control module, the main control module being connected to the sub-control module; the sub-control module includes a signal transmission unit, a Beidou positioning unit, a remote monitoring unit, and a fault detection unit; the signal transmission unit is connected to the main control module, the Beidou positioning unit is connected to both the signal transmission unit and the remote monitoring unit, and the remote monitoring unit is connected to the fault detection unit; the fault detection unit is used to obtain fault detection information of the distribution network object corresponding to the sub-control module; the signal transmission unit is used to feed back the fault detection information obtained by the remote monitoring unit to the main control module through the Beidou positioning unit, so that the main control module can monitor the distribution network object.

[0006] However, in the process of implementing the inventive technical solution in the embodiments of this application, it was found that the above-mentioned technology has at least the following technical problems: In existing technologies, the complex composition of multi-source signals or the presence of excessive interference signals can lead to incomplete separation by Independent Component Analysis (ICA), affecting the accuracy of main signal feature extraction. In multipath propagation environments, reflected waves and interference signals may be very close, resulting in overlap in the time and frequency domains, which can further complicate blind source separation and feature extraction. Even after ICA separation, subtle differences between signals may still lead to information loss or misjudgment, resulting in low signal feature extraction efficiency due to multi-source signal interference in power distribution network operation status monitoring. Summary of the Invention This application provides a method and system for monitoring the operation status of a distribution network based on multi-source data analysis. This solves the problem of low signal feature extraction efficiency caused by multi-source signal interference in the prior art for monitoring the operation status of a distribution network, and improves the efficiency of time-frequency feature extraction in the monitoring of the operation status of a distribution network.

[0007] This application provides a method for monitoring the operation status of a distribution network based on multi-source data analysis, including the following steps: S1, performing wavelet transform processing on the initial reflected wave signal obtained by the reflected wave signal receiving device of the distribution network, obtaining the reflected wave signal after wavelet transform processing, and evaluating the effectiveness of time-frequency analysis based on the wavelet transform evaluation parameters in the wavelet transform process to determine whether wavelet transform optimization should be performed; S2, if wavelet transform optimization is performed, performing blind source separation processing on the re-acquired reflected wave signal to obtain the main reflected wave signal in the distribution network; otherwise, directly performing blind source separation processing on the reflected wave signal to obtain the main reflected wave signal in the distribution network, and simultaneously evaluating the effectiveness of blind source separation based on the blind source separation evaluation parameters in the blind source separation process to determine whether blind source separation optimization should be performed; S3, if blind source separation optimization is performed, obtaining time-frequency features from the re-acquired main reflected wave signal; otherwise, directly obtaining time-frequency features from the acquired main reflected wave signal, and simultaneously identifying the type of operation fault in the distribution network based on the acquired time-frequency features.

[0008] This application provides a distribution network operation status monitoring system based on multi-source data analysis, including: a time-frequency analysis effectiveness evaluation module, a blind source separation effectiveness evaluation module, and a distribution network operation fault category identification module. The time-frequency analysis effectiveness evaluation module performs wavelet transform processing on the initial reflected wave signal obtained by the distribution network's reflected wave signal receiving device, obtains the wavelet-transformed reflected wave signal, and performs a time-frequency analysis effectiveness evaluation based on the wavelet transform evaluation parameters during the wavelet transform process to determine whether wavelet transform optimization should be performed. The blind source separation effectiveness evaluation module is used to determine whether wavelet transform optimization should be performed. If the reacquired reflected wave signal is processed by blind source separation, the main reflected wave signal in the distribution network is obtained. Otherwise, the reflected wave signal is directly processed by blind source separation to obtain the main reflected wave signal in the distribution network. At the same time, the effective evaluation of blind source separation is performed based on the blind source separation evaluation parameters in the blind source separation process to determine whether blind source separation optimization should be performed. The distribution network operation fault category identification module is used to obtain time-frequency features from the reacquired main reflected wave signal if blind source separation optimization is performed, otherwise the time-frequency features are directly obtained from the acquired main reflected wave signal. At the same time, the distribution network operation fault category is identified based on the acquired time-frequency features.

[0009] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages: 1. The reflected wave signal is obtained through wavelet transform processing, and the effectiveness of time-frequency analysis is evaluated to optimize the wavelet transform. Then, blind source separation processing is performed to obtain the main reflected wave signal, and the effectiveness of blind source separation is evaluated to optimize the blind source separation. Finally, the distribution network operation fault type is identified based on the obtained time-frequency features, thereby improving the separation quality of the main reflected wave signal and improving the efficiency of time-frequency feature extraction in distribution network operation status monitoring. This effectively solves the problem of low signal feature extraction efficiency caused by multi-source signal interference in the existing technology for distribution network operation status monitoring.

[0010] 2. By processing the spectral width, a spectral width comparison coefficient is obtained. Then, the instantaneous frequency change coefficient is processed to obtain an instantaneous frequency change comparison coefficient. Next, the current waveform error coefficient is processed to obtain a current waveform error comparison coefficient. Finally, the spectral width comparison coefficient, instantaneous frequency change comparison coefficient, time-frequency focusing degree, and current waveform error comparison coefficient are processed to obtain a time-frequency analysis effectiveness evaluation index. This quantitatively evaluates the effectiveness of wavelet transform signal processing, thereby improving the separation quality of the main reflection wave.

[0011] 3. By processing the signal-to-noise ratio (SNR) to obtain the SNR comparison coefficient, then processing the current waveform error coefficient to obtain the corrected current waveform error comparison coefficient, then processing the separation mutual information to obtain the mutual information comparison coefficient, and simultaneously processing the blind source separation time to obtain the separation time comparison coefficient, finally processing the SNR comparison coefficient, the corrected current waveform error comparison coefficient, the mutual information comparison coefficient, and the separation time comparison coefficient to obtain the blind source separation effectiveness evaluation index, thereby quantitatively evaluating the effectiveness of blind source separation of the reflected wave signal, and thus improving the quality of time-frequency feature extraction. Attached Figure Description

[0012] Figure 1 A flowchart of a power distribution network operation status monitoring method based on multi-source data analysis provided in this application embodiment; Figure 2 A flowchart of the power distribution network operation status monitoring method provided in the embodiments of this application; Figure 3 A flowchart for identifying power distribution network operation fault categories provided in this application embodiment; Figure 4 A schematic diagram of the structure of a power distribution network operation status monitoring system based on multi-source data analysis provided in this application embodiment; Figure 5 An interface diagram of the time-frequency analysis effectiveness monitoring page of the power distribution network operation status monitoring system provided in this application embodiment; Figure 6 An interface diagram of the blind source separation effectiveness monitoring page of the power distribution network operation status monitoring system provided in this application embodiment; Figure 7 This is an interface diagram of the operation fault identification page of the power distribution network operation status monitoring system provided in this application embodiment. Detailed Implementation

[0013] This application provides a method and system for monitoring the operation status of a distribution network based on multi-source data analysis. This solves the problem of low signal feature extraction efficiency caused by multi-source signal interference in existing distribution network operation status monitoring technologies. The method obtains reflected wave signals through wavelet transform processing, and evaluates the effectiveness of time-frequency analysis based on wavelet transform evaluation parameters to determine whether wavelet transform optimization is necessary. Then, blind source separation processing is performed to obtain the main reflected wave signal. The effectiveness of blind source separation is evaluated based on blind source separation evaluation parameters to determine whether blind source separation optimization is necessary. Finally, the obtained time-frequency features are used to identify the type of distribution network operation faults, thus improving the efficiency of time-frequency feature extraction in distribution network operation status monitoring.

[0014] The technical solution in this application embodiment aims to address the problem of low signal feature extraction efficiency caused by multi-source signal interference in the above-mentioned power distribution network operation status monitoring. The overall approach is as follows: By acquiring the reflected wave signal and performing time-frequency analysis to evaluate its effectiveness for wavelet transform optimization, then acquiring the main reflected wave signal and performing blind source separation to evaluate its effectiveness for blind source separation optimization, and finally identifying the distribution network operation fault type based on the time-frequency characteristics, the efficiency of time-frequency feature extraction in distribution network operation status monitoring is improved.

[0015] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0016] like Figure 1 The diagram shown is a flowchart of a power distribution network operation status monitoring method based on multi-source data analysis provided in this application embodiment. The method includes the following steps: S1, Time-Frequency Analysis Effectiveness Evaluation: The initial reflected wave signal obtained by the reflected wave signal receiving device of the distribution network is subjected to wavelet transform processing, and the reflected wave signal after wavelet transform processing is obtained. The time-frequency analysis effectiveness is evaluated according to the wavelet transform evaluation parameters in the wavelet transform process to determine whether wavelet transform optimization is required. The reflected wave signal is re-obtained after wavelet transform optimization. The signals generated by multiple distributed signal generators are injected into the distribution network line, and the initial reflected wave signal is detected by the reflected wave signal receiving device. Wavelet transform processing can convert the initial reflected wave signal from the time domain to the time-frequency domain.

[0017] S2, Effective evaluation of blind source separation: If wavelet transform optimization is performed, the reacquired reflected wave signal is processed by blind source separation to obtain the main reflected wave signal in the distribution network; otherwise, the reflected wave signal is directly processed by blind source separation to obtain the main reflected wave signal in the distribution network. At the same time, the effective evaluation of blind source separation is performed based on the blind source separation evaluation parameters in the blind source separation process to determine whether blind source separation optimization is performed, and the main reflected wave signal is reacquired after blind source separation optimization. Blind source separation processing is used to separate the main reflected wave signal from the reflected wave signal.

[0018] S3, Distribution Network Operation Fault Category Identification: If blind source separation optimization is performed, the time-frequency characteristics are obtained from the reacquired main reflected wave signal; otherwise, the time-frequency characteristics are directly obtained from the acquired main reflected wave signal. At the same time, the distribution network operation fault category is identified based on the acquired time-frequency characteristics. The time-frequency characteristics include, but are not limited to, the instantaneous frequency and amplitude of the main reflected wave signal.

[0019] Before designing a distribution network operation status monitoring method based on multi-source data analysis, a database is established to store various set data. The database includes, but is not limited to, preset blind source separation duration, preset instantaneous frequency change coefficient, analysis and evaluation compensation value, and cutoff frequency mapping set. Various values ​​are directly set by technical personnel. The preset blind source separation duration can be set by preset personnel. For example, the preset blind source separation duration can be represented by the average value of historical blind source separation durations within the historical time period in the database. In addition, various values ​​in the database can be set and fine-tuned by technical personnel according to actual debugging.

[0020] In this embodiment, as Figure 2 The flowchart shown is a flowchart of the power distribution network operation status monitoring method provided in this application embodiment. The specific logic is as follows: an initial reflected wave signal is acquired, wavelet transform processing is performed to acquire the reflected wave signal, and then the time-frequency analysis effectiveness evaluation of the wavelet transform processing is performed. It is determined whether the time-frequency analysis effectiveness evaluation index is less than the preset analysis effectiveness evaluation threshold. If so, wavelet transform optimization is performed, and blind source separation processing is performed on the reacquired reflected wave signal to obtain the main reflected wave signal. Otherwise, blind source separation processing is performed on the reflected wave signal to obtain the main reflected wave signal. At the same time, blind source separation effectiveness evaluation is performed on the blind source separation processing. It is determined whether the blind source separation effectiveness evaluation index is less than the preset separation effectiveness evaluation threshold. If so, blind source separation optimization is performed, and time-frequency features are obtained from the reacquired main reflected wave signal. Otherwise, time-frequency features are obtained from the acquired main reflected wave signal, and the power distribution network operation fault category is identified based on the acquired time-frequency features.

[0021] The complexity of multi-source signals or the presence of excessive interference can lead to incomplete separation by Independent Component Analysis (ICA), affecting the accuracy of main signal feature extraction. In multipath propagation environments, reflected waves and interference signals may be very close, resulting in overlap in the time and frequency domains, which can further complicate blind source separation and feature extraction. Even after ICA separation, subtle differences between signals may still lead to information loss or misjudgment. The effectiveness of wavelet transform is highly dependent on the chosen wavelet basis function; different wavelet basis functions may have significantly different effects on different types of signals, and incorrect wavelet basis functions may reduce the accuracy of time-frequency analysis. Within certain frequency ranges, the time-frequency resolution may be uneven, potentially causing blurring of low-frequency components.

[0022] Since reflected wave signals from distribution networks are typically non-stationary, containing rich information about transient changes and frequency components, wavelet transform is chosen. Wavelet transform performs multi-scale analysis of reflected wave signals in the time and frequency domains, simultaneously capturing both the time and frequency characteristics of the signals. It is suitable for analyzing non-stationary signals and capturing transient changes. For example, in distribution networks, reflected wave signals may contain transient changes caused by faults, which wavelet transform can effectively capture. Furthermore, since reflected wave signals in distribution networks often propagate through multiple paths, the received signal is a superposition of multiple reflected waves. Independent component analysis (ICA) is a blind source separation algorithm that can separate independent source signals from the superposition of multiple reflected waves without knowing the source signals or the mixing process.

[0023] In this application, wavelet transform processing can effectively decompose the initial reflected wave signal and capture its time-frequency characteristics, providing support for subsequent initial reflected wave signal separation and fault diagnosis. Blind source separation enables the recovery of the main reflected wave signal from the reflected wave signal, ensuring signal accuracy. The effective evaluation process for blind source separation ensures the quality of the separation, preventing erroneous separation results from affecting subsequent analysis. The time-frequency characteristics provide detailed information about the main reflected wave signal, helping to identify fault types in the distribution network (such as short circuits and overloads), thereby enabling timely operation and maintenance decisions. This ultimately improves the efficiency of time-frequency feature extraction in distribution network operation status monitoring.

[0024] Furthermore, the effectiveness of time-frequency analysis is evaluated based on the wavelet transform evaluation parameters during the wavelet transform process. The specific method is as follows: The instantaneous frequency variation coefficient is obtained by comparing the standard deviation and the average value of the instantaneous frequency; the comparison process involved in this application represents a division operation; the specific limiting expression for the instantaneous frequency variation coefficient is as follows: ; In the formula, SPB represents the instantaneous frequency change coefficient in the wavelet transform process, PBZ represents the standard deviation of the instantaneous frequency, and PJZ represents the average value of the instantaneous frequency. The instantaneous frequency is calculated through Hilbert transform.

[0025] The current waveform in the reflected wave signal is similar to a preset current waveform to obtain the current waveform error coefficient; the preset current waveform is set by a designated person; the specific limiting expression for the current waveform error coefficient is as follows: ; In the formula, t represents the number of the preset sampling point within the preset sampling time period. T represents the total number of preset sampling points, and IBZ represents the current waveform error coefficient during the wavelet transform process. This represents the current value at the t-th preset sampling point in the reflected wave signal. This represents the current value at the t-th preset sampling point of the preset current waveform.

[0026] The spectral width obtained during the wavelet transform process is compared with the preset spectral width obtained from the preset database to obtain the spectral width comparison coefficient, i.e. In the formula, PKD represents the spectral width during the wavelet transform process. The time-frequency graph is obtained through wavelet transform (such as Morlet wavelet), and then the spectral width is obtained by calculating the standard deviation of the spectrum of the time-frequency graph. This indicates the preset spectral width, which is set by a preset user, for example, by representing 10% of the frequency range of the initial reflected wave signal.

[0027] The instantaneous frequency change coefficients during the wavelet transform process are compared with preset instantaneous frequency change coefficients obtained from a preset database to obtain the instantaneous frequency change comparison coefficients, i.e. In the formula, SPB represents the instantaneous frequency change coefficient during the wavelet transform process. This indicates the preset instantaneous frequency change coefficient, which is set by the preset personnel.

[0028] The relative difference in time-frequency focus during wavelet transform is processed to obtain the time-frequency focus deviation coefficient, i.e. In the formula, SPJ represents the time-frequency focus during the wavelet transform process, which is obtained by calculating the area of ​​the time-frequency plot.

[0029] The current waveform error comparison coefficient is obtained by comparing the error coefficients of the current waveform obtained during the wavelet transform process with the preset current waveform error coefficients obtained from the preset database. In the formula, IBZ represents the current waveform error coefficient during the wavelet transform process. This indicates the preset current waveform error coefficient, which is set by the preset personnel.

[0030] The effective evaluation compensation value is introduced to assign values ​​to the spectrum width comparison coefficient, instantaneous frequency change comparison coefficient, time-frequency focus deviation coefficient, and current waveform error comparison coefficient to obtain the effective coefficient of time-frequency analysis. The effective evaluation compensation value includes the first effective evaluation compensation value, the second effective evaluation compensation value, the third effective evaluation compensation value, and the fourth effective evaluation compensation value.

[0031] The time-frequency analysis effectiveness evaluation index is obtained by performing an inverse proportional operation on the effective coefficients of the time-frequency analysis. The time-frequency analysis effectiveness evaluation index is used to quantitatively evaluate the effectiveness of wavelet transform signal processing.

[0032] The specific constraint expression for the effective evaluation index of time-frequency analysis is as follows: ; In the formula, SP represents the effective evaluation index of time-frequency analysis in the wavelet transform process. This indicates that the first analysis effectively assessed the compensation value. This indicates that the second analysis effectively evaluates the compensation value. This indicates that the third analysis effectively assesses the compensation value. This indicates that the fourth analysis effectively evaluates the compensation value.

[0033] The effective evaluation compensation values ​​involved are obtained from a preset database. The first effective evaluation compensation value represents the influence of spectral width on the effective evaluation index of time-frequency analysis; the second effective evaluation compensation value represents the influence of instantaneous frequency change coefficient on the effective evaluation index of time-frequency analysis; the third effective evaluation compensation value represents the influence of key rotation frequency on the effective evaluation index of time-frequency analysis; and the fourth effective evaluation compensation value represents the influence of current waveform error coefficient on the effective evaluation index of time-frequency analysis. The sum of these four values ​​is 1. For example, the spectral width and the preset first effective evaluation compensation value form a spectral width mapping set. Inputting the real-time spectral width into the spectral width mapping set yields the corresponding first effective evaluation compensation value. Instantaneous frequency change... The coefficients and preset second effective evaluation compensation values ​​form an instantaneous frequency mapping set. The real-time instantaneous frequency change coefficients are input into the instantaneous frequency mapping set to obtain the corresponding second effective evaluation compensation value. The time-frequency focus and preset third effective evaluation compensation value form a time-frequency focus mapping set. The real-time time-frequency focus is input into the time-frequency focus mapping set to obtain the corresponding third effective evaluation compensation value. The current waveform error coefficient and preset fourth effective evaluation compensation value form a current waveform error coefficient mapping set. The real-time current waveform error coefficient is input into the current waveform error coefficient mapping set to obtain the corresponding fourth effective evaluation compensation value. The mapping relationship can be one-to-one or many-to-one.

[0034] In this embodiment, a larger spectral width generally means that the reflected wave signal has a higher rate of change in the frequency domain, which may lead to more drastic changes in instantaneous frequency, thereby increasing the instantaneous frequency change coefficient; it may also lead to a worse degree of time-frequency localization of the signal, resulting in a decrease in time-frequency focus; a larger instantaneous frequency change coefficient may lead to insufficient time-frequency focus of the signal, thereby reducing time-frequency focus; the error of the current waveform may be caused by changes in instantaneous frequency. If the instantaneous frequency of the signal changes drastically, it may lead to an increase in the error coefficient of the current waveform; a larger spectral width may mean that the signal contains more high-frequency components, which may increase the error coefficient of the current waveform because the high-frequency components of the signal may not be able to accurately track the ideal waveform; a lower time-frequency focus generally indicates that the signal is more diffuse in the time-frequency domain, making it difficult to accurately restore to the ideal waveform, which may lead to an increase in the error coefficient of the current waveform.

[0035] By using the spectral width comparison coefficient, abnormal changes in the signal spectrum are identified, thereby improving the accuracy of feature extraction. The instantaneous frequency change comparison coefficient quantifies the frequency fluctuation anomalies of the initial reflected wave signal, thus improving the accuracy of feature extraction. The current waveform error comparison coefficient measures the difference between the reflected wave signal and the preset standard waveform, thereby improving the accuracy of feature extraction. The time-frequency focus deviation coefficient reflects the degree of focus of the initial reflected wave signal in time and frequency, thus improving the efficiency of time-frequency feature extraction. By comprehensively considering the spectral width comparison coefficient, instantaneous frequency change comparison coefficient, time-frequency focus deviation coefficient, and current waveform error comparison coefficient, the effectiveness of wavelet transform signal processing is quantitatively evaluated. This helps reduce noise and interference in the initial reflected wave signal, effectively improving the separation quality of the main reflected wave, ensuring the accuracy and reliability of the extracted time-frequency features, and ultimately improving the efficiency of time-frequency feature extraction in power distribution network operation status monitoring.

[0036] Further, the process for determining whether to perform wavelet transform optimization is as follows: If the effective evaluation index of time-frequency analysis is less than the preset effective evaluation threshold obtained from the preset database, wavelet transform optimization is performed; otherwise, wavelet transform optimization is not performed. The preset effective evaluation threshold is represented by the average value of the effective evaluation index of time-frequency analysis within the historical time period.

[0037] Wavelet transform optimization includes signal filtering optimization, wavelet basis function optimization, and wavelet scaling adjustment.

[0038] If the effective evaluation index of the time-frequency analysis after signal filtering optimization is less than the preset effective evaluation threshold obtained from the preset database, then wavelet basis function optimization is performed and the reflected wave signal is reacquired; otherwise, the current wavelet basis function is used and the wavelet transform optimization ends.

[0039] If the effective evaluation index of time-frequency analysis after wavelet basis function optimization is less than the preset effective evaluation threshold obtained from the preset database, wavelet scale adjustment is performed and the reflected wave signal is reacquired; otherwise, wavelet transform optimization ends.

[0040] If the effective evaluation index of time-frequency analysis after wavelet scale adjustment is less than the preset effective evaluation threshold obtained from the preset database, feedback is provided; otherwise, wavelet transform optimization ends.

[0041] Specifically, signal filtering optimization refers to processing the reflected wave signal using a low-pass filter that allows signal components below the cutoff frequency to pass through.

[0042] The cutoff frequency is obtained by inputting the maximum frequency of the reflected wave signal and the effective evaluation index of time-frequency analysis into the cutoff frequency mapping set. The cutoff frequency mapping set is a set obtained from a preset database that represents the mapping relationship between the maximum frequency of the reflected wave signal, the effective evaluation index of time-frequency analysis, and the cutoff frequency. The maximum frequency represents the highest frequency point in the spectrum of the reflected wave signal obtained by performing wavelet transform on the reflected wave signal.

[0043] Wavelet basis function optimization means replacing the currently used wavelet basis function with an alternative wavelet basis function, namely the Daubechies wavelet. The Daubechies wavelet has a finite support interval, meaning that the wavelet function and the scaling function are non-zero only within a finite interval. This gives the wavelet transform local analysis capabilities, effectively capturing the local features of the signal. It is suitable for analyzing non-stationary and transient signals. Furthermore, the Daubechies wavelet has a vanishing moment, meaning that the inner product of the wavelet function and the polynomial signal is zero. The higher the vanishing moment, the stronger the wavelet function's ability to approximate the polynomial signal, and the more effectively it can compress the smooth part of the initial reflected wave signal and highlight the abrupt changes in the initial reflected wave signal.

[0044] Wavelet scaling adjustment refers to the local analysis of the initial reflected wave signal after wavelet basis function optimization through a sliding window and step size. Within the sliding window for local analysis, the wavelet scale of the wavelet basis function during wavelet transform processing is adjusted to a modified wavelet scale.

[0045] The modified wavelet scale is obtained by inputting the instantaneous frequency, sliding window, and time-frequency analysis effective evaluation index of the initial reflected wave signal after wavelet basis function optimization into the wavelet scale mapping set. The wavelet scale mapping set is a collection obtained from a preset database that represents the mapping relationship between the instantaneous frequency, sliding window, time-frequency analysis effective evaluation index, and wavelet scale of the initial reflected wave signal after wavelet basis function optimization.

[0046] The sliding window is obtained by inputting the instantaneous frequency change coefficient and the effective evaluation index of time-frequency analysis into the sliding window mapping set. The sliding window mapping set is a collection obtained from a preset database that represents the mapping relationship between the instantaneous frequency change coefficient, the effective evaluation index of time-frequency analysis, and the sliding window.

[0047] The step size is obtained by inputting the sliding window and the effective evaluation index of time-frequency analysis into the step size mapping set. The step size mapping set is a collection obtained from a preset database that represents the mapping relationship between the sliding window, the effective evaluation index of time-frequency analysis, and the step size.

[0048] In this embodiment, signal filtering optimization removes high-frequency signals (signal components above the cutoff frequency) from the initial reflected wave signal, making the initial reflected wave signal smoother, reducing noise interference to subsequent blind source separation processing, and improving the quality of the reflected wave signal. Wavelet basis function optimization selects the Daubechies wavelet, which is more suitable for the characteristics of the reflected wave signal, as the wavelet basis function, which can improve the wavelet transform's ability to extract time-frequency features and better capture the time-frequency characteristics in the reflected wave signal. Wavelet scaling adjustment enables more accurate extraction of the time-frequency features of the reflected wave signal in different sliding windows.

[0049] Noise removal can reduce erroneous time-frequency features caused by noise in subsequent time-frequency feature extraction, improving the accuracy and efficiency of time-frequency feature extraction. Setting the cutoff frequency can make the filtered signal more suitable for subsequent time-frequency analysis, avoiding increased computation due to retaining too many useless frequency components, thus improving the efficiency of time-frequency feature extraction. The selection of wavelet basis functions can more effectively decompose the initial reflected wave signal, reducing unnecessary calculations and improving the efficiency and quality of time-frequency feature extraction. Local analysis and wavelet scaling can reduce unnecessary computation while ensuring the quality of time-frequency feature extraction, thus improving its efficiency. Setting the sliding window can balance computational efficiency and the accuracy of time-frequency feature extraction, improving its efficiency. Setting the step size can avoid repeated calculations and omissions of key time-frequency features, thus improving the efficiency of time-frequency feature extraction.

[0050] Furthermore, the effectiveness of blind source separation is evaluated based on the blind source separation evaluation parameters during the blind source separation process. The specific method is as follows: The joint probability distribution of the main reflected wave signal and the rejection signal, the marginal probability distribution of the main reflected wave signal, and the marginal probability distribution of the rejection signal in the distribution network are subjected to mutual information processing to obtain the separation mutual information; the specific calculation formula for the mutual information is as follows: ,in, This represents the joint probability distribution of the main reflected wave signal and the discarded signal. This represents the marginal probability distribution of the primary reflected wave signal. This represents the marginal probability distribution of the removed signal.

[0051] The joint probability distribution is obtained by comparing the number of occurrences of each pair of signals with the total number of preset sampling points. The marginal probability distribution of the main reflected wave signal is obtained by coupling the joint probability distribution along the dimension of the removed signal. The marginal probability distribution of the removed signal is obtained by coupling the joint probability distribution along the dimension of the main reflected wave signal.

[0052] The signal-to-noise ratio (SNR) during blind source separation is compared with a preset SNR obtained from a preset database to obtain the SNR comparison coefficient, i.e. In the formula, XZB represents the signal-to-noise ratio during the blind source separation process. The signal power and noise power are obtained by summing the squares of the main reflected wave signal and the discarded signal at all preset sampling points. Where XP represents signal power and ZP represents noise probability; This indicates the preset signal-to-noise ratio, which is set by a preset user. For example, it can be set to 20dB.

[0053] The error coefficient of the current waveform in the reflected wave signal is compared with the preset current waveform error coefficient obtained from the preset database to obtain the corrected current waveform error comparison coefficient, i.e. In the formula, This represents the error coefficient of the current waveform in the reflected wave signal. This indicates the preset current waveform error coefficient.

[0054] The separated mutual information is compared with the preset mutual information obtained from the preset database to obtain the mutual information comparison coefficient, i.e. In the formula, HXX represents the separation mutual information in the blind source separation process. This represents the preset mutual information, which is set by a preset user. It can be set to a value close to 0, for example, it can be set to 0.1.

[0055] When the blind source separation time is not greater than the preset blind source separation time obtained from the preset database, the blind source separation time is compared with the preset blind source separation time obtained from the preset database to obtain the separation time comparison coefficient; otherwise, the separation time comparison coefficient is recorded as 1. The specific constraint expression for the separation time comparison coefficient is as follows: ; In the formula, FTX represents the separation time comparison coefficient, FLT represents the blind source separation time, which is usually the running time of the blind source separation algorithm, and represents the time from the start of the input reflected wave signal to the completion of blind source separation; This indicates the preset blind source separation duration, which is set by preset personnel, for example, by the maximum value of the blind source separation duration within a historical event segment.

[0056] By introducing a separation effective evaluation compensation value, the corrected current waveform error comparison coefficient, mutual information comparison coefficient, and separation time comparison coefficient are assigned and coupled, and then an inverse proportional operation is performed to obtain the blind source separation effective coefficient. The separation effective evaluation compensation value includes a second separation effective evaluation compensation value, a third separation effective evaluation compensation value, and a fourth separation effective evaluation compensation value. After introducing the first separation effective evaluation compensation value to assign a value to the signal-to-noise ratio comparison coefficient, it is coupled with the blind source separation effective coefficient to obtain the blind source separation effective evaluation index. The blind source separation effective evaluation index is used to quantitatively evaluate the effectiveness of blind source separation of reflected wave signals.

[0057] The specific limiting expression for the effective evaluation index of blind source separation is as follows: ; In the formula, MY represents the effective evaluation index for blind source separation. This indicates the effective compensation value for the first separation assessment. This indicates the effective assessment compensation value for the second separation. This indicates the effective assessment compensation value for the third separation. This indicates the effective compensation value for the fourth separation.

[0058] The separation effective evaluation compensation values ​​involved are obtained from a preset database. The first separation effective evaluation compensation value represents the influence of the signal-to-noise ratio (SNR) on the blind source separation effective evaluation index; the second separation effective evaluation compensation value represents the influence of the current waveform error coefficient on the blind source separation effective evaluation index; the third separation effective evaluation compensation value represents the influence of the separation mutual information on the blind source separation effective evaluation index; and the fourth separation effective evaluation compensation value represents the influence of the blind source separation duration on the blind source separation effective evaluation index. The sum of these four values ​​is 1. For example, the signal-to-noise ratio (SNR) and the preset first separation effective evaluation compensation value form an SNR mapping set. Inputting the real-time signal-to-noise ratio into the SNR mapping set yields the corresponding first separation effective evaluation compensation value. The current waveform error coefficient and the preset second separation effective evaluation compensation value form a separation current waveform error coefficient mapping set. The real-time current waveform error coefficient is input into the separation current waveform error coefficient mapping set to obtain the corresponding second separation effective evaluation compensation value. The separation mutual information and the preset third separation effective evaluation compensation value form a separation mutual information mapping set. The real-time separation mutual information is input into the separation mutual information mapping set to obtain the corresponding third separation effective evaluation compensation value. The blind source separation duration and the preset fourth separation effective evaluation compensation value form a blind source separation duration mapping set. The real-time blind source separation duration is input into the blind source separation duration mapping set to obtain the corresponding fourth separation effective evaluation compensation value. The mapping relationship can be one-to-one or many-to-one.

[0059] In this embodiment, a higher signal-to-noise ratio (SNR) generally indicates higher quality of the reflected wave signal, resulting in a reduced current waveform error coefficient. Furthermore, the blind source separation algorithm can complete blind source separation more quickly and accurately, thereby shortening the required blind source separation time. A higher SNR also helps increase the mutual information between the separated discarded signal and the main reflected wave signal, because better reflected wave signal quality leads to stronger independence and separability. Higher current waveform accuracy reflects less information loss during the separation process, potentially increasing separation mutual information and thus reducing blind source separation time. Larger separation mutual information generally indicates better signal separation, which may lead to a shorter blind source separation time.

[0060] The separation effect of the reflected wave signal was evaluated using the mutual information comparison coefficient, thereby improving the efficiency of time-frequency feature extraction. The signal quality of the main reflected wave signal after blind source separation was evaluated using the signal-to-noise ratio comparison coefficient, thereby improving the quality of time-frequency feature extraction. The difference between the current waveform of the main reflected wave signal and the preset current waveform was evaluated using the corrected current waveform error comparison coefficient, thereby improving the quality of time-frequency feature extraction. The reasonableness of the time required for the blind source separation process was evaluated using the blind source separation duration comparison coefficient, thereby improving the efficiency of time-frequency feature extraction. By comprehensively considering the mutual information comparison coefficient, signal-to-noise ratio comparison coefficient, corrected current waveform error comparison coefficient, and blind source separation duration comparison coefficient, the effectiveness of blind source separation of the reflected wave signal was quantitatively evaluated, avoiding extraction errors caused by signal aliasing or noise in the reflected wave signal, thus improving the efficiency of time-frequency feature extraction.

[0061] Further, the process for determining whether to perform blind source separation optimization is as follows: If the effective evaluation index for blind source separation is less than the preset effective evaluation threshold obtained from the preset database, the blind source separation algorithm is optimized and the main reflected wave signal is reacquired; otherwise, blind source separation optimization is not performed. The preset effective evaluation threshold is represented by the average value of the effective evaluation index for blind source separation over a historical time period.

[0062] If the effective evaluation index of blind source separation after optimization is less than the preset effective evaluation threshold obtained from the preset database, then the reflected wave signal is reduced in dimension and the main reflected wave signal is reacquired; otherwise, the blind source separation optimization ends.

[0063] If the effective evaluation index of blind source separation after dimensionality reduction of the reflected wave signal is less than the preset effective evaluation threshold obtained from the preset database, then blind source separation parallel processing is performed; otherwise, blind source separation optimization is terminated.

[0064] If the effective evaluation index of blind source separation after parallel processing is less than the preset effective evaluation threshold obtained from the preset database, feedback is given; otherwise, the blind source separation optimization ends.

[0065] In summary, blind source separation optimization includes blind source separation algorithm optimization, dimensionality reduction of reflected wave signals, and parallel processing of blind source separation.

[0066] Specifically, the blind source separation algorithm optimization means that in the entropy-based independent component analysis method, the non-Gaussian characteristics of the reflected wave signal are enhanced to reduce the entropy value of the reflected wave signal, thereby achieving the separation of the reflected wave signal; the entropy value is estimated by calculating the probability density function of the reflected wave signal, which is obtained by the kernel density estimation method.

[0067] Dimensionality reduction of reflected wave signals means reducing the dimension of the reflected wave signals according to the number of principal components, and then performing blind source separation processing on the dimension-reduced reflected wave signals to obtain the main reflected wave signals in the power distribution network.

[0068] The number of principal components is obtained by inputting the maximum variance of the reflected wave signal and the effective evaluation index for blind source separation into the principal component number mapping set. The principal component number mapping set is a collection obtained from a preset database that represents the mapping relationship between the maximum variance of the reflected wave signal, the effective evaluation index for blind source separation, and the number of principal components. The specific method for obtaining the maximum variance is as follows: First, the reflected wave signal is sampled at a preset sampling frequency to obtain a signal sequence. Then, the signal sequence is arranged in chronological order to form a signal matrix. The covariance of the signal matrix is ​​calculated to obtain the covariance matrix of the signal matrix. Next, the covariance matrix is ​​decomposed into eigenvalues ​​to obtain eigenvalues ​​and corresponding eigenvectors. Finally, the variance corresponding to the largest eigenvalue is recorded as the maximum variance. The preset sampling frequency is set by preset personnel.

[0069] Parallel processing of blind source separation means decomposing the blind source separation process into a preset number of blind source separation subtasks, and using a distributed computing framework to parallelize the blind source separation subtasks. The blind source separation subtasks include matrix factorization and gradient descent; the preset number is set by preset personnel.

[0070] In this embodiment, the signal separation algorithm is optimized using independent component analysis (ICA). The non-Gaussianity of the reflected wave signal enhances its independence, thereby reducing the entropy of the reflected wave signal and improving the clarity and independence of the main reflected wave signal separation. Dimensionality reduction of the reflected wave signal reduces noise and redundancy, improving the accuracy of blind source separation and time-frequency feature extraction. Parallel processing of blind source separation accelerates the blind source separation process, thus improving its efficiency. Through these steps, the reflected wave signal is accurately separated, providing a clearer and more accurate main reflected wave signal for time-frequency feature extraction, thereby improving the efficiency of time-frequency feature extraction and ultimately enhancing the overall processing efficiency of power distribution network operation status monitoring.

[0071] Furthermore, the specific method for obtaining time-frequency characteristics from the reacquired main reflected wave signal is as follows: Determine whether the variance of the newly acquired primary reflected wave signal is greater than the preset variance of the primary reflected wave obtained from the preset database: If the re-acquired variance is greater than the preset primary reflection wave variance obtained from the preset database, the first extraction scheme is used to obtain the time-frequency characteristics in the re-acquired primary reflection wave signal; otherwise, the second extraction scheme is used to obtain the time-frequency characteristics in the re-acquired primary reflection wave signal. The preset primary reflection wave variance is set by preset personnel, for example, represented by the average value of the variance of the primary reflection wave signal within a historical time period.

[0072] The first extraction scheme involves obtaining the time-frequency characteristics of the main reflected wave signal through an adaptive filter; the second extraction scheme involves obtaining the time-frequency characteristics of the main reflected wave signal through wavelet transform. The adaptive filter can change the filter coefficients in real time according to the statistical characteristics of the main reflected wave signal to achieve the best filtering effect, thereby obtaining the time-frequency characteristics of the main reflected wave signal.

[0073] In this embodiment, the characteristics of the reacquired main reflected wave signal can be initially determined by comparing the variance of the main reflected wave signal, thereby selecting the corresponding extraction scheme and improving the accuracy and effectiveness of subsequent time-frequency feature extraction. For signals with larger variances, there may be more noise or complex frequency components, making it easier to extract time-frequency features. For signals with smaller variances, wavelet transform can more accurately analyze the local characteristics of the signal in the time and frequency domains, thus accurately extracting time-frequency features. Through adaptive filtering, important information of the main reflected wave signal at different times and frequencies can be extracted, such as changes in frequency components and transient characteristics of the signal, providing strong support for subsequent time-frequency feature extraction. Through wavelet transform, subtle changes and features in the main reflected wave signal can be accurately captured, such as the frequency component distribution of the main reflected wave signal, effectively decomposing the main reflected wave signal into different frequency sub-bands, facilitating more in-depth analysis and processing of the main reflected wave signal, and improving the accuracy and reliability of time-frequency feature extraction.

[0074] Furthermore, based on the acquired time-frequency characteristics, the distribution network operation fault categories are identified, and the specific methods are as follows: Obtain the Euclidean distance between the main reflected wave and all reflected wave samples in the preset database; the Euclidean distance is used to measure the similarity between two vectors.

[0075] The Euclidean distances are sorted in ascending order to obtain the Euclidean distance sequence. The reflected wave samples with the smallest preset sample number in the Euclidean distance sequence are selected; the preset sample number is set by preset personnel.

[0076] Obtain the distribution network operation fault category corresponding to a preset number of reflected wave samples.

[0077] The operational fault type that appears most frequently in the reflected wave samples of a preset number of samples is set as the operational fault type of the main reflected wave.

[0078] In this embodiment, as Figure 2 The diagram shows a flowchart for identifying distribution network operation fault categories provided in this application embodiment. The specific logic is as follows: reacquire the main reflected wave signal, determine whether the variance of the main reflected wave signal is greater than the preset main reflected wave variance, if so, adopt the first extraction scheme to obtain time-frequency features, otherwise adopt the second extraction scheme to obtain time-frequency features, obtain the Euclidean distance between the main reflected wave and the reflected wave sample, sort to obtain the Euclidean distance sequence, select the reflected wave sample with the smallest preset sample number in the Euclidean distance sequence, obtain the distribution network operation fault category corresponding to the preset sample number of reflected wave samples, and set the operation fault type with the highest frequency of occurrence as the operation fault type of the main reflected wave.

[0079] By calculating the Euclidean distance, the similarity between the main reflected wave and each reflected wave sample in the preset database can be quantified. The smaller the Euclidean distance, the more similar the main reflected wave is to the sample, providing a basis for subsequent selection of similar samples. The sorting operation can clearly show the similarity relationship between the main reflected wave and each reflected wave sample. Selecting the preset number of reflected wave samples with the smallest Euclidean distance can filter out the samples most similar to the main reflected wave, improving the accuracy of fault category identification. By statistically analyzing the most frequent fault types and adopting the principle of majority rule, the operational fault type of the main reflected wave can be determined more accurately, reducing the probability of misjudgment due to the particularity of individual samples or noise interference, and improving the reliability and accuracy of fault category identification.

[0080] like Figure 4The diagram shown is a structural schematic of a distribution network operation status monitoring system based on multi-source data analysis provided in this application embodiment. The system includes: a time-frequency analysis effectiveness evaluation module, a blind source separation effectiveness evaluation module, and a distribution network operation fault category identification module. The time-frequency analysis effectiveness evaluation module performs wavelet transform processing on the initial reflected wave signal obtained by the distribution network's reflected wave signal receiving device, obtains the wavelet-transformed reflected wave signal, and performs time-frequency analysis effectiveness evaluation based on the wavelet transform evaluation parameters during the wavelet transform process to determine whether wavelet transform optimization is required. (Blind source...) The effective evaluation module is used to perform blind source separation processing on the reacquired reflected wave signal to obtain the main reflected wave signal in the distribution network if wavelet transform optimization is performed; otherwise, blind source separation processing is directly performed on the reflected wave signal to obtain the main reflected wave signal in the distribution network. At the same time, the effective evaluation of blind source separation is performed based on the blind source separation evaluation parameters in the blind source separation process to determine whether blind source separation optimization should be performed. The distribution network operation fault category identification module is used to obtain time-frequency features from the reacquired main reflected wave signal if blind source separation optimization is performed; otherwise, time-frequency features are directly obtained from the acquired main reflected wave signal. At the same time, the distribution network operation fault category is identified based on the acquired time-frequency features.

[0081] In this embodiment, by evaluating the effectiveness of time-frequency analysis and blind source separation, problems in the reflected wave signal processing process can be identified and optimized in a timely manner, thereby improving the accuracy of signal processing and helping to extract the main reflected wave signal in the distribution network more accurately, providing a reliable data foundation for subsequent time-frequency feature extraction.

[0082] By obtaining time-frequency characteristics from the main reflected wave signal and identifying the fault type, the operating status and fault type of the distribution network can be determined more accurately, which helps to detect and handle faults in the distribution network in a timely manner and improve the reliability and security of the distribution network.

[0083] By dynamically adjusting the signal processing procedure (such as wavelet transform optimization and blind source separation optimization), the performance of time-frequency feature extraction can be optimized, including improving the efficiency of reflected wave signal processing, reducing computational complexity, and reducing resource consumption, thereby enabling the power distribution network operation status monitoring system to operate more efficiently and stably.

[0084] It should be added that, such as Figure 5 The image shown is an interface diagram of the time-frequency analysis effectiveness monitoring page of the power distribution network operation status monitoring system provided in this application embodiment. Figure 6 The image shown is an interface diagram of the blind source separation effectiveness monitoring page of the distribution network operation status monitoring system provided in this application embodiment. Figure 7The image shown is an interface diagram of the operation fault identification page of the power distribution network operation status monitoring system provided in this application embodiment; by Figure 5 , 6 As can be seen from section 7, the navigation bar of the power distribution network operation status monitoring system provided in this application embodiment includes a time-frequency analysis effectiveness monitoring bar, a blind source separation effectiveness monitoring bar, an operation fault identification bar, and a system configuration management bar.

[0085] Depend on Figure 5 It can be seen that when the current spectral width is 12.45Hz, the current instantaneous frequency is 50.23Hz, the current time-spectrum focus is 87.6%, and the current current waveform error is 7.8%, the time-frequency analysis status is displayed as an abnormal time-frequency validity status, the status indicator shows a red alarm light, and the wavelet transform optimization is in the second stage: wavelet basis function optimization.

[0086] Depend on Figure 6 It can be seen that when the current mutual information is 0.32 bits, the current blind source separation time is 2.45 seconds, the current signal-to-noise ratio is 18.2 dB, and the current waveform error is 7.8%, the blind source separation status is displayed as an abnormal state of blind source separation effectiveness. The status indicator light shows a red alarm light, and the wavelet transform optimization is the first stage: blind source separation algorithm optimization.

[0087] Depend on Figure 7 It can be seen that the current variance of the main reflected wave signal is 2.45V. 2 The preset normal range threshold for the variance of the primary reflected wave is 0.8-1.2V. 2 The corresponding distribution network operation fault identification status is ground fault. In the fault identification results, the probability of ground fault is 93.7%, the probability of poor grounding is 27.3%, and the probability of short circuit fault is 12.5%. The time-frequency feature extraction scheme is the second extraction scheme, which obtains the time-frequency features in the main reflected wave signal through wavelet transform.

[0088] In summary, the embodiments of this application obtain the reflected wave signal through wavelet transform processing, perform time-frequency analysis effectiveness evaluation to optimize the wavelet transform, then perform blind source separation processing to obtain the main reflected wave signal, and perform blind source separation effectiveness evaluation to optimize the blind source separation. Finally, the distribution network operation fault category is identified based on the obtained time-frequency features, thereby improving the separation quality of the main reflected wave signal and improving the time-frequency feature extraction efficiency in distribution network operation status monitoring. This effectively solves the problem of low signal feature extraction efficiency caused by multi-source signal interference in distribution network operation status monitoring in the prior art.

[0089] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied 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.

[0090] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0091] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0092] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0093] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0094] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for monitoring the operating status of a distribution network based on multi-source data analysis, characterized in that, Includes the following steps: S1. Perform wavelet transform processing on the initial reflected wave signal obtained by the reflected wave signal receiving device of the distribution network, obtain the reflected wave signal after wavelet transform processing, and evaluate the effectiveness of time-frequency analysis based on the wavelet transform evaluation parameters in the wavelet transform process to determine whether wavelet transform optimization should be performed. S2, if wavelet transform optimization is performed, the reacquired reflected wave signal is processed by blind source separation to obtain the main reflected wave signal in the distribution network; otherwise, the reflected wave signal is directly processed by blind source separation to obtain the main reflected wave signal in the distribution network. At the same time, the effective evaluation of blind source separation is performed based on the blind source separation evaluation parameters in the blind source separation process to determine whether blind source separation optimization is to be performed. S3. If blind source separation optimization is performed, the time-frequency characteristics are obtained from the reacquired main reflected wave signal; otherwise, the time-frequency characteristics are obtained directly from the acquired main reflected wave signal. At the same time, the distribution network operation fault category is identified based on the acquired time-frequency characteristics.

2. The method for monitoring the operation status of a distribution network based on multi-source data analysis as described in claim 1, characterized in that: The method for evaluating the effectiveness of time-frequency analysis based on wavelet transform evaluation parameters during the wavelet transform process is as follows: The instantaneous frequency variation coefficient is obtained by comparing the standard deviation of the instantaneous frequency with the average value of the instantaneous frequency. The current waveform in the reflected wave signal is similar to the preset current waveform to obtain the current waveform error coefficient. The spectral width obtained during the wavelet transform process is compared with the preset spectral width obtained from the preset database to obtain the spectral width comparison coefficient. The instantaneous frequency change coefficients in the wavelet transform process are compared with the preset instantaneous frequency change coefficients obtained from the preset database to obtain the instantaneous frequency change comparison coefficients. The relative difference of time-frequency focus during wavelet transform is processed to obtain the time-frequency focus deviation coefficient. The current waveform error comparison coefficient is obtained by comparing the current waveform error coefficient obtained in the wavelet transform process with the preset current waveform error coefficient obtained from the preset database. The effective evaluation compensation value is introduced to assign values ​​to the spectral width comparison coefficient, instantaneous frequency change comparison coefficient, time-frequency focus deviation coefficient, and current waveform error comparison coefficient to obtain the effective coefficient of time-frequency analysis. The effective evaluation compensation value includes a first effective evaluation compensation value, a second effective evaluation compensation value, a third effective evaluation compensation value, and a fourth effective evaluation compensation value. The time-frequency analysis effectiveness evaluation index is obtained by performing an inverse proportional operation on the effective coefficients of the time-frequency analysis. The time-frequency analysis effectiveness evaluation index is used to quantitatively evaluate the effectiveness of wavelet transform signal processing.

3. The method for monitoring the operation status of a distribution network based on multi-source data analysis as described in claim 2, characterized in that: The specific process for determining whether to perform wavelet transform optimization is as follows: If the effective evaluation index of time-frequency analysis is less than the preset effective evaluation threshold obtained from the preset database, wavelet transform optimization is performed; otherwise, wavelet transform optimization is not performed. The wavelet transform optimization includes signal filtering optimization, wavelet basis function optimization, and wavelet scaling adjustment; If the effective evaluation index of time-frequency analysis after signal filtering optimization is less than the preset effective evaluation threshold obtained from the preset database, then wavelet basis function optimization is performed and the reflected wave signal is re-acquired; otherwise, the current wavelet basis function is used and the wavelet transform optimization ends. If the effective evaluation index of time-frequency analysis after wavelet basis function optimization is less than the preset effective evaluation threshold obtained from the preset database, then wavelet scale adjustment is performed and the reflected wave signal is reacquired; otherwise, wavelet transform optimization ends. If the effective evaluation index of time-frequency analysis after wavelet scaling is less than the preset effective evaluation threshold obtained from the preset database, feedback is performed; otherwise, wavelet transform optimization ends.

4. The method for monitoring the operation status of a distribution network based on multi-source data analysis as described in claim 3, characterized in that: The signal filtering optimization refers to processing the reflected wave signal using a low-pass filter that allows signal components below the cutoff frequency to pass through; The cutoff frequency is obtained by inputting the maximum frequency of the reflected wave signal and the effective evaluation index of time-frequency analysis into the cutoff frequency mapping set. The cutoff frequency mapping set is a set obtained from a preset database that represents the mapping relationship between the maximum frequency of the reflected wave signal, the effective evaluation index of time-frequency analysis, and the cutoff frequency. The wavelet basis function optimization means replacing the currently used wavelet basis function with an alternative wavelet basis function; The wavelet scale adjustment refers to the local analysis of the initial reflected wave signal after wavelet basis function optimization through a sliding window and step size. Within the sliding window for local analysis, the wavelet scale of the wavelet basis function during wavelet transform processing is adjusted to a modified wavelet scale. The modified wavelet scale is obtained by inputting the instantaneous frequency, sliding window, and time-frequency analysis effective evaluation index of the initial reflected wave signal after wavelet basis function optimization into the wavelet scale mapping set. The wavelet scale mapping set is a set obtained from a preset database that represents the mapping relationship between the instantaneous frequency, sliding window, time-frequency analysis effective evaluation index, and wavelet scale of the initial reflected wave signal after wavelet basis function optimization. The sliding window is obtained by inputting the instantaneous frequency change coefficient and the effective evaluation index of time-frequency analysis into the sliding window mapping set. The sliding window mapping set is a set obtained from a preset database that represents the mapping relationship between the instantaneous frequency change coefficient, the effective evaluation index of time-frequency analysis, and the sliding window. The step size is obtained by inputting the sliding window and the effective evaluation index of time-frequency analysis into the step size mapping set. The step size mapping set is a collection obtained from a preset database that represents the mapping relationship between the sliding window, the effective evaluation index of time-frequency analysis, and the step size.

5. The method for monitoring the operation status of a distribution network based on multi-source data analysis as described in claim 1, characterized in that: The method for evaluating the effectiveness of blind source separation based on the blind source separation evaluation parameters during the blind source separation process is as follows: The joint probability distribution of the main reflected wave signal and the rejection signal, the marginal probability distribution of the main reflected wave signal, and the marginal probability distribution of the rejection signal in the distribution network are processed with mutual information to obtain the separation mutual information. The joint probability distribution is obtained by comparing the number of occurrences of each pair of signals with the total number of preset sampling points. The marginal probability distribution of the main reflected wave signal is obtained by coupling the joint probability distribution along the dimension of the removed signal. The marginal probability distribution of the removed signal is obtained by coupling the joint probability distribution along the dimension of the main reflected wave signal. The signal-to-noise ratio (SNR) during the blind source separation process is compared with the preset SNR obtained from the preset database to obtain the SNR comparison coefficient. The error coefficient of the current waveform in the reflected wave signal is compared with the preset current waveform error coefficient obtained from the preset database to obtain the corrected current waveform error comparison coefficient. The separated mutual information is compared with the preset mutual information obtained from the preset database to obtain the mutual information comparison coefficient. When the blind source separation time is not greater than the preset blind source separation time obtained from the preset database, the blind source separation time is compared with the preset blind source separation time obtained from the preset database to obtain the separation time comparison coefficient; otherwise, the separation time comparison coefficient is recorded as 1. By introducing a separation effective evaluation compensation value, after assigning and coupling the correction current waveform error comparison coefficient, mutual information comparison coefficient and separation time comparison coefficient, the blind source separation effective coefficient is obtained by performing an inverse proportional operation. The separation effective evaluation compensation value includes a second separation effective evaluation compensation value, a third separation effective evaluation compensation value and a fourth separation effective evaluation compensation value. After the signal-to-noise ratio comparison coefficient is assigned a value by introducing a first separation effective evaluation compensation value, it is coupled with the blind source separation effective coefficient to obtain the blind source separation effective evaluation index. The blind source separation effective evaluation index is used to quantitatively evaluate the effectiveness of blind source separation of reflected wave signals.

6. The method for monitoring the operation status of a distribution network based on multi-source data analysis as described in claim 5, characterized in that: The specific process for determining whether to perform blind source separation optimization is as follows: If the effective evaluation index of blind source separation is less than the preset effective evaluation threshold obtained from the preset database, then the blind source separation algorithm is optimized and the main reflected wave signal is reacquired; otherwise, blind source separation optimization is not performed. If the effective evaluation index of blind source separation after optimization by the blind source separation algorithm is less than the preset effective evaluation threshold obtained from the preset database, then the dimension reduction of the reflected wave signal is performed and the main reflected wave signal is re-acquired; otherwise, the blind source separation optimization ends. If the effective evaluation index of blind source separation after dimensionality reduction of the reflected wave signal is less than the preset effective evaluation threshold obtained from the preset database, then blind source separation parallel processing is performed; otherwise, blind source separation optimization is terminated. If the effective evaluation index of blind source separation after parallel processing is less than the preset effective evaluation threshold obtained from the preset database, feedback is given; otherwise, the blind source separation optimization ends. The blind source separation optimization includes blind source separation algorithm optimization, dimensionality reduction of reflected wave signals, and parallel processing of blind source separation.

7. The method for monitoring the operation status of a distribution network based on multi-source data analysis as described in claim 6, characterized in that: The optimized representation of the blind source separation algorithm is to reduce the entropy value of the reflected wave signal by enhancing the non-Gaussian characteristics of the reflected wave signal in the entropy-based independent component analysis method. The dimension reduction of the reflected wave signal means reducing the dimension of the reflected wave signal according to the number of principal components, and then performing blind source separation processing on the dimension-reduced reflected wave signal to obtain the main reflected wave signal in the power distribution network. The principal component number is obtained by inputting the maximum variance of the reflected wave signal and the effective evaluation index of blind source separation into the principal component number mapping set. The principal component number mapping set is a set obtained from a preset database that represents the mapping relationship between the maximum variance of the reflected wave signal, the effective evaluation index of blind source separation, and the principal component number. The parallel processing of blind source separation means decomposing the blind source separation process into a preset number of blind source separation subtasks, and using a distributed computing framework to process the blind source separation subtasks in parallel. The blind source separation subtasks include matrix factorization and gradient descent.

8. The method for monitoring the operation status of a distribution network based on multi-source data analysis as described in claim 1, characterized in that: The specific method for obtaining time-frequency features from the reacquired main reflected wave signal is as follows: Determine whether the variance of the newly acquired primary reflected wave signal is greater than the preset variance of the primary reflected wave obtained from the preset database: If the variance of the reacquired main reflection wave signal is greater than the preset main reflection wave variance obtained from the preset database, the first extraction scheme is used to obtain the time-frequency characteristics in the reacquired main reflection wave signal; otherwise, the second extraction scheme is used to obtain the time-frequency characteristics in the reacquired main reflection wave signal. The first extraction scheme involves obtaining the time-frequency characteristics of the main reflected wave signal using an adaptive filter; The second extraction scheme involves obtaining the time-frequency characteristics of the main reflected wave signal through wavelet transform.

9. The method for monitoring the operation status of a distribution network based on multi-source data analysis as described in claim 1, characterized in that: The specific method for identifying the type of power distribution network operation faults based on the acquired time-frequency characteristics is as follows: Obtain the Euclidean distance between the main reflected wave and all reflected wave samples in the preset database; The Euclidean distances are sorted in ascending order to obtain the Euclidean distance sequence. The reflected wave samples with the smallest preset sample size in the Euclidean distance sequence are selected. Obtain the power distribution network operation fault category corresponding to a preset number of reflected wave samples; The operational fault type that appears most frequently in the reflected wave samples of a preset number of samples is set as the operational fault type of the main reflected wave.

10. A power distribution network operation status monitoring system based on multi-source data analysis, characterized in that, include: Time-frequency analysis effectiveness evaluation module, blind source separation effectiveness evaluation module, and distribution network operation fault category identification module; The time-frequency analysis effectiveness evaluation module is used to perform wavelet transform processing on the initial reflected wave signal obtained by the reflected wave signal receiving device of the distribution network, obtain the reflected wave signal after wavelet transform processing, and perform time-frequency analysis effectiveness evaluation based on the wavelet transform evaluation parameters in the wavelet transform process to determine whether wavelet transform optimization should be performed. The blind source separation effective evaluation module is used to perform blind source separation processing on the reacquired reflected wave signal to obtain the main reflected wave signal in the distribution network if wavelet transform optimization is performed; otherwise, it directly performs blind source separation processing on the reflected wave signal to obtain the main reflected wave signal in the distribution network. At the same time, it performs blind source separation effective evaluation based on the blind source separation evaluation parameters in the blind source separation process to determine whether blind source separation optimization should be performed. The power distribution network operation fault category identification module is used to obtain time-frequency features from the reacquired main reflected wave signal if blind source separation optimization is performed; otherwise, it directly obtains time-frequency features from the acquired main reflected wave signal and identifies the power distribution network operation fault category based on the acquired time-frequency features.

Images