Power cable insulation performance monitoring method and system

Abstract

The invention relates to the technical field of cable insulation monitoring, in particular to a power cable insulation performance monitoring method and system. The method comprises the following steps: obtaining a dielectric loss angle tangent value signal of a power cable, and carrying out blind source separation on the dielectric loss angle tangent value signal; for each independent component of blind source separation, two composite discrimination indexes are constructed: a component fluctuation spectrum index, which is used for quantizing the stability of the signal from the frequency characteristic; and the environment coupling coefficient is used for judging a signal source from the correlation with the temperature and the load and the change trend of the environment coupling coefficient. And an aging discrimination factor is calculated by fusing the two indexes, and a real aging trend signal is identified from a plurality of independent components, so that the insulation performance of the power cable is monitored. According to the invention, the automation level and the anti-interference capability of cable insulation monitoring are obviously improved, and misjudgment caused by environmental interference is effectively avoided.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of cable insulation monitoring, and in particular to a power cable insulation performance monitoring method and system. BACKGROUND

[0002] In the modern power grid operation system, as the core carrier of power transmission, the insulation performance of power cable directly determines the power supply reliability and operation safety of the power grid. Cable insulation aging is an important cause of power grid failure, so real-time monitoring of insulation performance has become a key link in power grid operation and maintenance. Among them, the dielectric loss tangent is a core quantitative index representing the degree of insulation aging.

[0003] However, in the actual operating environment, the dielectric loss tangent signal collected is often submerged in strong interference such as low-frequency drift caused by temperature and load changes and high-frequency noise caused by complex electromagnetic environment, and the signal-to-noise ratio is extremely low. The traditional frequency domain filtering method is difficult to cope with such complex non-stationary interference, resulting in the inability to accurately restore the true aging trend.

[0004] The prior art introduces an independent component analysis (ICA) algorithm. As an advanced blind source separation technology, ICA can separate multiple statistically independent source signals from a mixed signal, and in theory can separate the real aging signal from various interference signals. However, the ICA algorithm has its inherent blind defect when applied to this scenario. The algorithm only completes mathematical separation, and the physical meaning, amplitude and order of the separation result are all uncertain. This means that after obtaining multiple separated independent components, the system cannot automatically and objectively identify which component represents the slowly changing and one-way insulation aging signal, which is highly related to environmental factors, and which is random high-frequency noise. This lack of ability to analyze and judge the physical characteristics of each component only by mathematical separation makes it difficult to achieve automated and high-precision monitoring, and is prone to false diagnosis results due to human error or algorithm confusion. SUMMARY

[0005] To solve the technical problem that the ICA algorithm has a blind defect when separating signals, cannot automatically identify the physical meaning of each component, and further leads to incorrect monitoring results, the present application provides solutions in the following aspects.

[0006] In a first aspect, the present application provides a power cable insulation performance monitoring method, which comprises the steps of: The medium loss tangent signal, temperature data and load current data of the power cable are acquired; blind source separation is performed on the medium loss tangent signal to obtain a plurality of independent components; a high frequency interval and a low frequency interval of each independent component are distinguished according to a preset frequency division threshold value, and a component fluctuation spectral index of each independent component is calculated based on a ratio of energy of the high frequency interval to energy of the low frequency interval; for each independent component, an environmental coupling coefficient of the independent component is acquired based on time correlation of the independent component with the temperature data and the load current data and a change trend of the independent component; an aging discrimination factor of each independent component is calculated based on the component fluctuation spectral index and the environmental coupling coefficient; the aging discrimination factor is negatively correlated with a product of the component fluctuation spectral index and the environmental coupling coefficient; a target signal is selected from the plurality of independent components according to the aging discrimination factor, and monitoring of the insulation performance of the power cable is realized based on the target signal.

[0007] The application firstly decomposes a mixed signal into a plurality of independent components through blind source separation, then proposes two composite indexes of a component fluctuation spectral index and an environmental coupling coefficient, the former judges the stationarity from the internal frequency characteristics of the signal, and the latter judges the source from the correlation with the external environment and the change trend of the signal, and finally calculates an aging discrimination factor by fusing the information in the two dimensions, so that the slow and stable real aging trend can be accurately distinguished from the intense and random environmental interference such as temperature, load current change and noise. This greatly improves the accuracy and reliability of the insulation performance monitoring, effectively avoids false alarms caused by environmental factors, and solves the technical problems of poor anti-interference ability and easy misjudgment existing in the prior art.

[0008] Preferably, the component fluctuation spectral index of the independent component satisfies the relationship: ; Wherein, is the component fluctuation spectral index of the i-th independent component; is the component fluctuation spectral index of the i-th independent component; is the frequency division threshold value of the low frequency interval and the high frequency interval; is the highest frequency of the i-th independent component; is the frequency of the i-th independent component; is the frequency of the i-th independent component; is the power spectral density function of the i-th independent component, which is obtained by Fourier transform on the i-th independent component.

[0009] ​​​This invention defines the component fluctuation spectrum index by the ratio of high-frequency and low-frequency power spectra. This approach makes the quantification of signal fluctuations more objective, accurate, and reproducible, providing a solid mathematical foundation for the accurate calculation of subsequent discriminant factors.

[0010] Preferably, obtaining the environmental coupling coefficient of the independent component includes: constructing a time series corresponding to the independent component, denoted as the independent component series; and also constructing a temperature series corresponding to the temperature data and a load current series corresponding to the load current data; the environmental coupling coefficient of the independent component satisfies the following relationship: ; in, It is the first The environmental coupling coefficient of an independent component; , They are the first The correlation coefficients between the sequences corresponding to each independent component and the temperature and load current sequences during the same period; It is the first The number of groups in the sequence corresponding to each independent component whose difference between adjacent data points is positive; It is the first The number of groups in the sequence corresponding to each independent component whose difference between adjacent data points is not positive; It is the number of data points in the sequence corresponding to the independent component; It is a preset micro value; It is a standard normalized function; It is the absolute value symbol.

[0011] This invention integrates the correlation between independent components and the two main sources of interference, temperature and load current, as well as the two key characteristics of whether the signal itself has a monotonically increasing trend, into a single formula. This comprehensive consideration of multiple dimensions and factors makes the judgment on whether an independent component is due to environmental interference or actual aging more comprehensive and robust, significantly improving the ability to distinguish between them.

[0012] Preferably, the difference between adjacent data points is the difference between the previous data point and the next data point.

[0013] Preferably, the aging discrimination factor of each independent component is calculated by a negative exponential function of the product of the component fluctuation spectrum index and the environmental coupling coefficient of the independent component.

[0014] This invention clarifies that the aging discrimination factor is calculated as a negative exponential function of the product of the component fluctuation spectrum index and the environmental coupling coefficient. This nonlinear mapping effectively amplifies the differences in discrimination factors between the target signal and the interference signal, resulting in higher discriminative power and clearer decision boundaries in the screening results, thereby improving the accuracy of the final screening.

[0015] Preferably, the filtering of the target signal from the plurality of independent components according to the aging discriminant factor comprises: determining the independent component with the largest aging discriminant factor as the target signal.

[0016] Preferably, the monitoring of the insulation performance of the power cable based on the target signal comprises: obtaining a scaling coefficient corresponding to the target signal by using an inverse process of the blind source separation; restoring the amplitude of the target signal by using the scaling coefficient, and superimposing the mean value of the dielectric loss tangent signal to obtain a pure dielectric loss tangent change curve; and performing anomaly detection on the pure dielectric loss tangent change curve by using a local outlier factor algorithm.

[0017] After successfully identifying the target signal, the inverse process of the blind source separation is further used to reconstruct a pure aging curve that excludes all interferences, and a local outlier factor algorithm is introduced for anomaly detection. This not only provides clearer and more intuitive aging trends for implementers, but also realizes automatic early warning of potential sudden failures.

[0018] Preferably, the obtaining of the dielectric loss tangent signal of the power cable comprises: synchronously collecting a voltage reference signal and a total current signal of the power cable; calculating a phase difference between the voltage reference signal and the total current signal; and determining a loss angle based on the phase difference to calculate the dielectric loss tangent signal.

[0019] Preferably, the blind source separation is implemented by using an independent component analysis algorithm.

[0020] In a second aspect, the present application provides a power cable insulation performance monitoring system, which comprises a memory and a processor, and the memory stores computer program instructions which, when executed by the processor, implement the power cable insulation performance monitoring method of the first aspect of the present application.

[0021] By using the above technical solution, the power cable insulation performance monitoring method of the first aspect of the present application is generated into a computer program and stored in the memory to be loaded and executed by the processor, so that a terminal device is manufactured according to the memory and the processor, and is convenient to use.

[0022] The beneficial effects of the present application: in view of the problem that the prior art cannot accurately distinguish between real aging and environmental interference, the present application proposes a discrimination mechanism based on blind source separation and multi-dimensional feature fusion. It constructs two composite indexes of component fluctuation spectrum index and environmental coupling coefficient, and accurately images each independent signal component after independent component separation from the internal stationarity and external correlation of the signal, and finally calculates the aging discrimination factor, so as to intelligently identify the real aging signal. This method can effectively filter out strong interference such as temperature and load, and significantly improve the monitoring accuracy; the present application also introduces a local outlier factor algorithm for anomaly detection. This not only provides clearer and more intuitive aging trend for the implementer, but also realizes automatic early warning of potential sudden failure. BRIEF DESCRIPTION OF DRAWINGS

[0023] Figure 1 A flowchart of a power cable insulation performance monitoring method provided by an embodiment of the present application is shown in the figure. Figure 2 A curve graph of the original source signal provided by the embodiment of the present application is shown in the figure. Figure 3 A dielectric loss tangent value signal curve graph provided by an embodiment of the present application is shown in the figure. Figure 4 A curve graph of a plurality of independent components output by the ICA algorithm provided by an embodiment of the present application is shown in the figure. Figure 5 A pure dielectric loss tangent value change curve graph provided by an embodiment of the present application is shown in the figure. Figure 6 A structural block diagram of a power cable insulation performance monitoring system provided by an embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0024] The first aspect of the embodiment of the present application provides a power cable insulation performance monitoring method, as shown in the figure. Figure 1 The method comprises steps S100-S600: Step S100, obtaining the dielectric loss tangent value signal, temperature data and load current data of the power cable.

[0025] It should be noted that this step aims to obtain the necessary basic data for subsequent insulation state analysis and evaluation. Among them, the dielectric loss tangent value (tan ) is an internationally recognized core physical quantity that can sensitively reflect the internal energy loss of dielectric and the aging degree of insulating material. However, under online monitoring environment, the measured value is easily affected by the running temperature and load current fluctuation of the cable, resulting in drift and disturbance unrelated to real aging. Therefore, synchronous acquisition of these three groups of key data signals is the data basis for building subsequent interference elimination and accurate aging trend identification model.

[0026] Specifically, a non-intrusive high-frequency current transformer is installed on the cable grounding wire to measure the total current signal flowing through the cable insulation layer in real time. This signal includes capacitive current and a weak resistive current related to insulation loss. Simultaneously, the system uses a capacitive voltage divider to synchronously acquire the voltage reference signal from the power grid.

[0027] In one feasible implementation, the acquisition device samples the aforementioned current and voltage signals at a frequency of 50Hz. Subsequently, in the data processing unit, a Fast Fourier Transform (FFT) algorithm is used to calculate the phase difference between the voltage and the total current signal. This phase difference is compared with... The complementary angle is the loss angle. And then calculate The value is the dielectric loss tangent signal. It is calculated from the values ​​at each sampling time. The values ​​are arranged in chronological order, thus constructing a time series of the dielectric loss tangent signal, which serves as the core input for subsequent analysis. Regarding the sampling time, data from 1000 time points is considered; implementers can choose according to their needs.

[0028] At the same time, temperature and load current data of the cable are acquired using temperature and current sensors deployed on the surface of the cable at the same 50Hz acquisition frequency.

[0029] At this point, three sets of key data have been obtained for subsequent analysis.

[0030] Step S200: Perform blind source separation on the dielectric loss tangent signal to obtain multiple independent components.

[0031] It should be noted that the original acquired dielectric loss tangent signal is a complex hybrid signal, typically containing at least three types of signal sources linearly superimposed: a signal representing the true aging trend of irreversible insulation degradation, a reversible drift signal caused by temperature and load changes, and a high-frequency noise signal generated by electromagnetic interference or measurement errors. For example... Figure 2 The graph shown is a curve of the original source signal. In this graph, the horizontal axis represents the sampling points, and the vertical axis represents the amplitude of the original signal. The graph contains three curves: the blue curve is the aging trend signal curve, the green curve is the high-frequency noise signal curve, and the orange curve is the environmental drift signal curve. The original source signal is existing technology and is the fundamental reason for the existence of ICA, so it will not be elaborated on here.

[0032] like Figure 3The figure shows the dielectric loss tangent signal curve. In this figure, the horizontal axis represents the sampling point, and the vertical axis represents the amplitude of the mixed dielectric loss tangent signal. The figure also contains three curves: the blue curve is the curve after mixing the aging trend signal, the green curve is the curve after mixing the high-frequency noise signal, and the orange curve is the curve after mixing the environmental drift signal.

[0033] Blind source separation (BSS) technology can recover the original, statistically independent source signals to a maximum extent when both the source signals and the mixing matrix are unknown. The purpose of this step is to utilize this technology to recover the mixed signals... The signal is decomposed into a series of purer signal components with independent physical meaning.

[0034] Specifically, Independent Component Analysis (ICA) is the preferred technique for achieving blind source separation. ICA is a blind source separation tool whose core idea is to find an unmixing matrix that maximizes the non-Gaussianity of the output signal components, thereby achieving the highest statistical independence among the components.

[0035] The dielectric loss tangent signal obtained in step S100 is used as input. The ICA algorithm is executed to output multiple independent signal components, i.e., independent components. The physical meaning of these independent components is unknown at the beginning of the separation process; they may correspond to real insulation aging signals, high-frequency noise signals, or temperature- and load-related environmental drift signals. For example... Figure 4 The graph shown represents the output of multiple independent signal components after executing the ICA algorithm. The horizontal axis represents the sampling points, and the vertical axis represents the amplitude of each independent component. The graph also contains three curves: the blue curve corresponding to IC-1 is the aging trend signal curve separated by ICA; the green curve corresponding to IC-3 is the high-frequency noise signal curve separated by ICA; and the orange curve corresponding to IC-2 is the environmental drift signal curve separated by ICA. The ICA algorithm is existing technology and will not be elaborated upon further here.

[0036] Thus, multiple independent components obtained from the decomposition of the dielectric loss tangent signal were acquired.

[0037] Step S300: Distinguish the high-frequency and low-frequency ranges of each independent component according to a preset frequency division threshold, and calculate the component fluctuation spectrum index of each independent component based on the ratio of the energy of the high-frequency range to the energy of the low-frequency range.

[0038] It should be noted that each independent component usually exhibits significant different characteristics in the frequency domain, while insulation aging is an extremely slow physical and chemical process, and the corresponding signal component must exhibit slowly varying signal characteristics with highly concentrated energy in the extremely low frequency band in the frequency domain; relatively, various noises and periodic interferences are mainly distributed in the high frequency band; therefore, the present application can preliminarily distinguish the potential aging signal from the high frequency interference by quantifying the energy distribution characteristics of each independent component in the frequency domain.

[0039] Specifically, the Fourier transform is performed on the first independent component separated by ICA to obtain the power spectral density function .

[0040] According to the above logic, the component fluctuation frequency spectrum index of the first independent component satisfies the following relationship: ; wherein, f is the frequency division threshold of the low frequency interval and the high frequency interval; is the highest frequency of the first independent component; is the frequency of the first independent component; is the power spectral density function of the first independent component, which is obtained by performing Fourier transform on the first independent component.

[0041] In the relationship, the numerator represents the total energy of the signal of the first independent component in the high frequency interval, and the denominator represents the total energy of the signal of the first independent component in the low frequency interval. Therefore, when the value of the component fluctuation frequency spectrum index is larger, it indicates that the energy of the independent component is more concentrated in the high frequency part, and the possibility of high frequency noise or periodic interference is larger. Conversely, when the value of the component fluctuation frequency spectrum index is smaller, it indicates that the energy of the component is mainly concentrated in the low frequency part, and the possibility of slowly varying signal is larger. It should be noted that f is the frequency division threshold for defining the low frequency interval and the high frequency interval, and the setting thereof is the key to realize effective separation. The threshold is selected according to the characteristic frequency for distinguishing the real aging trend and the periodic environmental interference. The real insulation aging of the power cable is an extremely low frequency process with months or years as a unit, while the main low frequency interference is caused by the daily variation cycle of temperature, load and other factors, and the characteristic frequency thereof is about 0.01 Hz.

[0042] ​​​​Hz. In one possible implementation, the frequency division threshold value is set to Hz. This setting can ensure that the energy of the main interference such as the daily variation cycle is counted in the high frequency part of the relationship formula, while the energy of the real aging trend is attributed to the low frequency part in the denominator. The calculated component fluctuation spectrum index can extremely effectively represent the physical properties of the signal components, laying the foundation for subsequent accurate identification. Those skilled in the art can understand that the value of the frequency division threshold value can be adaptively adjusted under other specific working conditions, for example, when the weekly variation is the main interference.

[0043] It should also be noted that the highest frequency is usually half of the sampling frequency, and the highest frequency of the analysis is set to half of the sampling frequency, which is not an arbitrary choice, but strictly follows the basic guidelines in the field of digital signal processing, that is, the Nyquist-Shannon sampling theorem, to ensure the effectiveness and accuracy of the spectrum analysis. The Nyquist-Shannon sampling theorem is prior art and will not be described here.

[0044] So far, the component fluctuation spectrum index that can represent the frequency distribution characteristics of each independent component has been obtained.

[0045] Step S400, for each independent component, based on the time correlation of the independent component with the temperature data and the load current data, and the change trend of the independent component, an environmental coupling coefficient of the independent component is obtained.

[0046] It should be noted that according to the component fluctuation spectrum index of each independent component, we can distinguish low frequency signal components and high frequency signal components; for low frequency signal components, they may contain real insulation aging signals, or they may contain drift interference signals caused by changes in temperature, load and other environmental factors, which also exhibit low frequency characteristics. The two types of signals have similar shapes in the time domain, which is the main source of misjudgment of traditional methods.

[0047] Since the real insulation aging is essentially a permanent degradation of the inherent properties of the cable insulation material, this degradation trend is unidirectional accumulation with time and cannot be reversed, and has weak direct correlation with temperature, load and other external environmental factors; while the environmental drift interference is highly related to changes in temperature and load, and is usually reversible fluctuation. This step aims to build a comprehensive index that can quantify both environmental correlation and trend reversibility.

[0048] Specifically, in order to perform subsequent quantitative analysis, the original data signal and the separated independent components need to be constructed into a structured time series.

[0049] For the construction of independent component sequence: the sequence is the direct output of the ICA algorithm, when the input medium loss tangent signal time sequence contains data points, the algorithm outputs multiple independent signal components, which also contain data points. Each of the output signal components, whose data points correspond one-to-one with the timestamps of the original signal, itself constitutes an independent component sequence.

[0050] For the construction of temperature sequence and load current sequence: the discrete data points synchronously collected in step S100 are sorted according to their respective timestamps to form the temperature sequence and the load current sequence. The temperature sequence can be expressed as , wherein is the temperature data at the sampling time. The load current sequence can be expressed as , wherein is the load current data at the sampling time. Since the temperature data, load current data and medium loss tangent signal are collected at the same frequency, the corresponding sampling times are the same, that is, the number of data points of the time sequences of the two environmental factors and the number of data points of the independent components are the same, and both contain data points of the sampling time.

[0051] After the above time sequences are constructed, the environmental coupling coefficient can be calculated. The time correlation of the independent component and the environmental factors, i.e. temperature and load current, is taken as the numerator, and the one-way increasing trend of the independent component is taken as the denominator. An independent component that is highly related to the environment and has reversible trend fluctuations will have a larger environmental coupling coefficient.

[0052] According to the above logic, the environmental coupling coefficient of the th independent component satisfies the relationship: ; wherein , are the correlation coefficients of the th independent component corresponding sequence and the temperature sequence and the load current sequence in the same period, respectively; is the number of groups of adjacent data points in the th independent component corresponding sequence that are positive when subtracted; is the number of groups of adjacent data points in the th independent component corresponding sequence that are not positive when subtracted; is the number of data points of the independent component sequence, which is also the total number of data points in the medium loss tangent signal time sequence, for example, 1000 consecutive data points can be taken for analysis; This is a preset microvalue used to prevent the denominator from being 0. It can be set to 0.001 or as needed. It is the standard normalization function, which normalizes the denominator. Mapped to interval; It is the absolute value symbol.

[0053] In this relationship, the numerator part Quantified the first The stronger the linear correlation between an independent component and external environmental factors such as temperature and load current, the stronger the correlation. The more likely an independent component is to be caused by environmental changes, the more likely the core term in the denominator is to be affected. Used to measure the The strength of the unidirectional increasing property of the sequence corresponding to each independent component varies. For a purely monotonically increasing sequence, this value approaches 1; for a fluctuating sequence, the value approaches 0. Therefore, when an independent component is an environmental drift, its numerator is large and its denominator is small, ultimately leading to... The value increases significantly. Conversely, for true aging signals, the numerator is smaller and the denominator is larger. The value will be very small.

[0054] It should be added that in this relation... , The preferred correlation coefficient is Pearson correlation coefficient, used to quantify the first... The strength of the linear correlation between each independent component and temperature and load current. In this relationship... and The difference between adjacent data points is calculated by subtracting the data of the previous data point from the data of the next data point.

[0055] To visually illustrate the role of the core term in the denominator, in order to include Taking a sequence of data points as an example, the number of sets of differences between corresponding adjacent data points is: , The number of groups with positive differences. If the number of groups has non-positive differences, then: ; Therefore, the first The core term in the denominator of the relationship for the environmental coupling coefficients of each independent component satisfies the following relationship: ; When the sequence is purely monotonically increasing At this point, the core term in the denominator is 1; when the sequence fluctuates up and down, The core term in the denominator approaches 0, thus quantifying the unidirectional increasing property.

[0056] At this point, the environmental coupling coefficient capable of quantifying the coupling degree of each independent component and environmental factors is obtained.

[0057] Step S500, based on the component fluctuation spectrum index and the environmental coupling coefficient, the aging discrimination factor of each independent component is calculated.

[0058] It should be noted that in combination with the physical nature of power cable insulation aging and the characteristics of the actual monitoring scene, a signal component that can accurately represent the real insulation aging needs to meet two core attributes: first, from the frequency domain characteristics, the degradation period of insulation aging lasts for several months to several years, which is an extremely low frequency physical process, so its signal energy is highly focused on the extremely low frequency band, and there is no high frequency fluctuation or periodic oscillation characteristics. Second, from the time domain correlation, the aging trend is unidirectional accumulation and irreversible, and the correlation with external environmental factors such as temperature and load is very weak. This step aims to combine the component fluctuation spectrum index and the environmental coupling coefficient to build an aging discrimination factor, which can determine the signal component that meets the characteristics of real insulation aging, and provide a basis for judging the selection of target aging signals from multiple independent components.

[0059] Specifically, according to the above logic, the aging discrimination factor of the first independent component is calculated as follows: The relationship is satisfied: ; Wherein, is the component fluctuation spectrum index of the first independent component; is the environmental coupling coefficient of the first independent component ; is the natural exponential function. In this relationship, only when and

[0060] are both small, their product is the smallest, and the value of the exponential term is closest to its maximum value 1. Therefore, the larger the value of the aging discrimination factor , the more the corresponding independent component meets the characteristics of the real insulation aging signal that is low frequency and independent of the environment. Step S600, according to the aging discrimination factor, the target signal is selected from the multiple independent components, and the monitoring of the insulation performance of the power cable is realized based on the target signal.

[0061] Step S600, according to the aging discrimination factor, the target signal is selected from the multiple independent components, and the monitoring of the insulation performance of the power cable is realized based on the target signal.

[0062] ​​It should be noted that this step is the final execution link of the whole monitoring method. Through the layer-by-layer screening and quantitative evaluation of the preceding steps, each independent component has been assigned a clear aging discriminant factor. The purpose of this step is to accurately identify and extract the target signal from the numerous separated independent components based on the discriminant factor, and on this basis, to perform amplitude recovery and anomaly detection, and finally to output the monitoring conclusion of the cable insulation performance.

[0063] Specifically, the execution process of this step is as follows: First, compare the aging discriminant factors of all the independent components calculated in step S500, and determine the independent component with the largest aging discriminant factor as the target signal.

[0064] Then, in order to make the target signal have a true physical meaning and amplitude, it needs to be recovered. The scaling coefficient corresponding to the target signal in the original mixing matrix is obtained by using the inverse process of ICA. The scaling coefficient is used to recover the amplitude of the target signal, and the overall mean value of the original dielectric loss tangent signal is superimposed, so as to obtain a pure dielectric loss tangent change curve that can reflect the true aging accumulation process, which is free of noise and environmental interference. The overall mean value of the original signal is superimposed, because the independent component baseline is 0 after ICA separation due to the zero-mean preprocessing of the signal before ICA separation; superimposing the mean value can restore the true physical baseline of the target signal, and ensure that the final pure curve is consistent in order of magnitude with the actual measurement value.

[0065] Finally, the insulation performance is monitored based on the pure change curve.

[0066] As a preferred embodiment, the local outlier factor algorithm can be used for real-time anomaly detection of the curve. LOF is a classic density-based unsupervised anomaly detection algorithm that can effectively identify abnormal mutation points in data sequences. When the LOF algorithm detects an outlier point on the curve, it indicates that the accumulation rate of the dielectric loss tangent has abnormally accelerated, and the system can determine that the cable insulation performance is abnormal and trigger a warning. As shown in FIG. 6, which is a pure dielectric loss tangent change curve diagram, the horizontal axis represents time, and the vertical axis represents the recovered dielectric loss tangent. The blue curve is the pure dielectric loss tangent change curve, and the red dot is the outlier point detected by the LOF algorithm, which corresponds to the abnormal point of the cable insulation performance. Figure 5

[0067] In the LOF algorithm, the number of neighborhood points is a key parameter, which can be empirically optimized to 20. The preferred number of neighborhood points is 20, which is based on the fact that the power cable ​​​The daily / weekly monitoring period of the signal, usually 1000 data points cover 20 neighborhood points can balance the stability of local density calculation and the sensitivity of outlier identification, if the data sampling frequency is higher, such as N>5000, the number of neighborhood points can be adjusted accordingly. The LOF algorithm is prior art, and will not be described here.

[0068] The second aspect of the embodiment provides a power cable insulation performance monitoring system, as shown in the figure, the power cable insulation performance monitoring system comprises a memory and a processor, the memory stores computer program instructions, when the computer program instructions are executed by the processor, the first aspect of the application is realized. A power cable insulation performance monitoring method. Figure 6

[0069] The power cable insulation performance monitoring system also includes a communication bus and a communication interface and other components familiar to those skilled in the art, the settings and functions of which are known in the art, and therefore will not be described here.

[0070] In the present application, the aforementioned memory can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, device or apparatus. For example, the computer readable storage medium can be any suitable magnetic storage medium or magneto-optical storage medium, such as resistive random access memory, dynamic random access memory, static random access memory, enhanced dynamic random access memory, high bandwidth memory, hybrid memory cube, etc., or any other medium that can be used to store the required information and can be accessed by an application, module or both. Any such computer storage medium can be part of the device or accessible or connectable to the device.

[0071] The above are the preferred embodiments of the present application, which do not limit the protection scope of the present application, therefore: any equivalent changes made in accordance with the structure, shape, principle of the present application should be covered within the protection scope of the present application.​

Claims

1. A method of monitoring the condition of the insulation of a power cable, characterized by The method comprises the steps of: obtaining a medium loss tangent signal, temperature data and load current data of a power cable; performing blind source separation on the medium loss tangent signal to obtain a plurality of independent components; distinguishing a high-frequency interval and a low-frequency interval of each independent component according to a preset frequency division threshold, and calculating a component fluctuation spectral index of the each independent component based on a ratio of energy of the high-frequency interval to energy of the low-frequency interval; for the each independent component, obtaining an environmental coupling coefficient of the independent component based on time correlation of the independent component with the temperature data and the load current data and a variation trend of the independent component; calculating an aging discriminant factor of the each independent component based on the component fluctuation spectral index and the environmental coupling coefficient; the aging discriminant factor is negatively correlated with a product of the component fluctuation spectral index and the environmental coupling coefficient; screening a target signal from the plurality of independent components according to the aging discriminant factor, and realizing monitoring of insulation performance of the power cable based on the target signal.

2. The method of claim 1, wherein the component fluctuation spectral index of the independent component satisfies a relationship: ； wherein, is the component fluctuation spectrum index of the th independent component; is the frequency division threshold between the low frequency interval and the high frequency interval; is the highest frequency of the th independent component; is the frequency of the th independent component; is the frequency of the th independent component; th independent component.

3. The power cable insulation performance monitoring method according to claim 1, characterized by, the obtaining of the environmental coupling coefficient of the independent component comprises: constructing a time sequence corresponding to the independent component, denoted as an independent component sequence; also constructing a temperature sequence corresponding to the temperature data and a load current sequence corresponding to the load current data; The environmental coupling coefficient of the independent component satisfies a relationship: ; wherein, is the ambient coupling coefficient of the th independent component; , are the correlation coefficients of the th independent component corresponding sequence and the temperature sequence, the load current sequence of the same period, respectively; is the number of groups of adjacent data points in the th independent component corresponding sequence that are positive after subtraction; is the number of groups of adjacent data points in the th independent component corresponding sequence that are not positive after subtraction; is the number of data points of the independent component corresponding sequence; is a preset infinitesimal value; is a standard normalization function; is an absolute value symbol.

4. The method of claim 3, wherein the difference between the adjacent data points is a difference between a latter data point and a former data point in the adjacent data points.

5. The power cable insulation performance monitoring method according to claim 1, characterized by, the aging discriminant factor of the each independent component is obtained by a negative exponential function of a product of the component fluctuation spectral index and the environmental coupling coefficient of the independent component.

6. The power cable insulation performance monitoring method according to claim 1, characterized by, the screening of the target signal from the plurality of independent components according to the aging discriminant factor comprises: determining the independent component with the largest aging discriminant factor as the target signal.

7. The power cable insulation performance monitoring method according to claim 1, characterized by, the monitoring of the insulation performance of the power cable based on the target signal comprises: obtaining a scaling coefficient corresponding to the target signal by using an inverse process of the blind source separation; restoring an amplitude of the target signal by using the scaling coefficient, and superimposing a mean value of the medium loss tangent signal to obtain a pure medium loss tangent change curve; performing anomaly detection on the pure medium loss tangent change curve by using a local outlier factor algorithm.

8. The power cable insulation performance monitoring method according to claim 1, characterized by, the obtaining of the medium loss tangent signal of the power cable specifically comprises: synchronously collecting a voltage reference signal and a total current signal of the power cable; calculating a phase difference between the voltage reference signal and the total current signal; determining a loss angle based on the phase difference, and calculating the medium loss tangent signal.

9. The power cable insulation performance monitoring method according to claim 1, characterized by, the blind source separation is implemented by using an independent component analysis algorithm.

10. A power cable insulation performance monitoring system, characterized by the power cable insulation performance monitoring system comprises a processor and a memory, and the memory stores computer program instructions which, when executed by the processor, realize a power cable insulation performance monitoring method according to any one of claims 1-9.

Images