An external damage detection method and system for a power line

Abstract

The present invention relates to the technical field of power line detection, and discloses a method and system for detecting external damage of power lines. The steps of the method include: setting detection nodes on the power lines; continuously collecting power parameters of each detection node; compensating and correcting the power parameters of each detection node; establishing a time series of power parameters for each detection node; obtaining a component sequence of the power parameters of each detection node based on the time series; calculating the non-linear characteristic index of each time series and the non-linear characteristic index of each component sequence; performing preliminary external damage detection on each detection node based on the non-linear characteristic index of the time series; and performing supplementary external damage detection on each detection node based on the non-linear characteristic index of the component sequence. This application can improve the accuracy and early warning ability of power line external damage detection and has good environmental adaptability.

Description

Technical Field

[0001] The present invention relates to the technical field of power line detection, and particularly relates to a method and system for detecting external damage of power lines. Background Art

[0002] Power lines are widely distributed and are exposed to complex natural and social environments for a long time, making them vulnerable to external damage threats. Traditional external damage detection of power lines mainly relies on manual inspections or visual inspections, which are inefficient and prone to missed detections; in the current power industry, there are also some detection methods that judge whether external damage has occurred through power parameters, improving the detection efficiency and accuracy, but there are still some technical deficiencies.

[0003] There are significant problems with the data accuracy of existing detection technologies. Power lines are widely distributed, and in some remote areas or harsh environments, sensors are easily interfered with, resulting in errors in the collected data, which poses a risk of misjudgment in external damage detection and reduces the reliability of detection. Currently, most detection methods have insufficient accuracy in the early identification of external damage to power lines. Existing detection means often focus on detecting after a fault occurs or when obvious abnormalities appear, and cannot detect potential external damage risks in a timely manner. For example, when a power line is slightly damaged initially, the power parameters of the line may only change slightly, and it is difficult for existing technologies to capture these subtle changes and issue early warnings, increasing the risk of power supply interruption. Existing technologies lack in-depth mining of power parameters. However, most current detection systems only perform simple statistics and analysis on data and cannot make full use of the non-linear characteristics in the data. In addition, different environmental factors and power system states also have a certain impact on external damage detection, and existing technologies also lack consideration in this regard.

[0004] As disclosed in the Chinese patent with the authorization announcement number CN114019318B, an indicator for fault section location of power lines is disclosed, including a walking device that automatically walks on the cable and automatically avoids suspension clamps. The walking device is provided with walking wheels, and the walking wheels drive the indicator to move on the cable, and the walking wheels can automatically lift to avoid suspension clamps. The walking device is provided with a scanning device for detecting the aging degree of the cable. The scanning device is provided with a scanning rod and a scanning motor, and the scanning motor drives the scanning rod to scan the cable to analyze the aging degree of the cable. The walking device is also provided with a touching device for detecting whether the cable is damaged and leaking electricity. By setting multiple connecting rods, the detection wheel is pressed tightly against the cable. The setting of this invention can work on various cables, automatically avoid obstacles on the cable, improve the efficiency of troubleshooting, and reduce the danger of troubleshooting.

[0005] As disclosed in the patent application with the authorization announcement number CN212845650U, a high-voltage transmission line detection device is disclosed. The technical solution includes a base and an upper cover. The front and rear sides of the base are respectively fixedly connected to locking pieces through bolts, and the other outer wall of the locking piece is fixedly connected to the upper cover through bolts. Insulating pads are bonded to the upper side wall of the base and the bottom side wall of the upper cover, and a signal antenna is installed inside the signal slot. A limiting slot is opened on the upper outer wall of the upper cover, and a photovoltaic panel is embedded inside the limiting slot. This technical solution solves the problems of high-voltage transmission line detection in the current stage. Most of them use manual observation or are paired with drones for aerial photography to observe whether there are line damages, and most of the high-voltage transmission line detection devices in the current stage can only detect the line areas that have been damaged and cannot immediately detect line faults during daily inspections.

[0006] All of the above technical solutions have the problems raised in this background technology: it is difficult to detect subtle wire damages and the environmental adaptability is insufficient.

[0007] The information disclosed in this background technology section is only intended to increase the overall understanding of the present invention and should not be regarded as an admission or any form of implication that this information constitutes the prior art already known to those of ordinary skill in the art. Summary of the Invention

[0008] The technical problem to be solved by the present invention is to overcome the defects of the prior art and provide a method and system for detecting external damages of power lines, so as to improve the accuracy and early warning ability of power line external damage detection.

[0009] To solve the above technical problems, the present invention provides the following technical solutions:

[0010] On the one hand, the present invention provides a method for detecting external damages of power lines, including the following steps: setting detection nodes on the power lines; continuously collecting the power parameters of each detection node;

[0011] Based on the power parameters of the adjacent nodes of each detection node, compensating and correcting the power parameters of each detection node;

[0012] Establishing a time series of the power parameters for each detection node; obtaining the component series of the power parameters of each detection node based on the time series;

[0013] Calculating the non-linear characteristic indexes of each time series and the non-linear characteristic indexes of each component series;

[0014] Conducting preliminary external damage detection for each detection node based on the non-linear characteristic indexes of the time series; conducting supplementary external damage detection for each detection node based on the non-linear characteristic indexes of the component series.

[0015] As a preferred solution of the external damage detection method for the power line described in the present invention, wherein: the method for compensating and correcting the power parameters of each detection node is as follows:

[0016] S100: Simultaneously collect the power parameters of all detection nodes;

[0017] S200: Compensate and correct the power parameters of any detection node, specifically including:

[0018] Obtain the power parameters of each adjacent node of the target detection node; obtain the historical power parameters of the target detection node and its adjacent nodes;

[0019] Calculate the accuracy rate of the target detection node and its adjacent nodes based on the historical power parameters; calculate the trust degree of each adjacent node based on the accuracy rate;

[0020] Calculate the compensation amount of the target detection node based on the trust degree of each adjacent node;

[0021] Compensate and correct the power parameters of the target detection node based on the compensation amount.

[0022] As a preferred solution of the external damage detection method for the power line described in the present invention, wherein: the calculation method of the accuracy rate of any detection node is as follows: count the abnormal power parameters in the historical power parameters of the detection node, and record the time stamp corresponding to each abnormal power parameter;

[0023] Obtain the maintenance records of the power line, and judge whether each abnormal power parameter corresponds to a line fault based on the maintenance records; mark the abnormal power parameters that do not correspond to line faults as false alarm power parameters; count the total number of historical power parameters; divide the difference between the total number of historical power parameters and the number of false alarm power parameters by the total number of historical power parameters to obtain the accuracy rate of the corresponding detection node.

[0024] As a preferred solution of the external damage detection method for the power line described in the present invention, wherein: the judgment method of the adjacent nodes of the target detection node is as follows: any node adjacent to the target detection node on the power line or with a distance less than a preset distance threshold from the target detection node is an adjacent node of the target detection node; the calculation method of the trust degree of any adjacent node of the target detection node is as follows:

[0025] Calculate the correlation coefficient between the historical power parameters of any adjacent node and the historical power parameters of the target detection node, and use it as the association degree between the corresponding adjacent node and the target detection node;

[0026] Obtain the distance between any adjacent node and the target detection node; calculate the trust degree of the corresponding adjacent node based on the correlation degree, distance between any adjacent node and the target detection node, and the accuracy of the corresponding adjacent node.

[0027] As a preferred solution of the external damage detection method for the power line described in the present invention, wherein: calculating the compensation amount of the target detection node based on the trust degree of each adjacent node specifically includes: performing weighted summation on the power parameters of each adjacent node based on the trust degree of each adjacent node to obtain the compensation amount of the target detection node; wherein, the weight coefficient of the power parameter of any adjacent node is the trust degree of the corresponding adjacent node divided by the sum of the trust degrees of all adjacent nodes.

[0028] Compensating and correcting the power parameters of the target detection node based on the compensation amount specifically includes:

[0029] Compensating and correcting the power parameters of the target detection node based on the weighted summation result of the original value and the compensation amount of the power parameters of the target detection node; wherein, the weight coefficients of the compensation amount and the original value are both related to the accuracy of the target detection node.

[0030] As a preferred solution of the external damage detection method for the power line described in the present invention, wherein: the component sequence includes an approximate component sequence and a detail component sequence; obtaining the component sequence of the power parameters of each detection node based on the time series specifically includes: selecting a wavelet basis function and a decomposition scale; performing wavelet transform on the time series of the power parameters of any detection node based on the selected wavelet basis function and decomposition scale to obtain the component sequence of the power parameters of the corresponding detection node;

[0031] The non-linear characteristic indexes of any time series or component sequence both include Lyapunov exponent, correlation dimension, and information entropy.

[0032] As a preferred solution of the external damage detection method for the power line described in the present invention, wherein: the preliminary external damage detection specifically includes:

[0033] Collect historical data of power parameters and establish a historical database of power parameters; any piece of historical data in the historical database includes the time series of the power parameters of a detection node without external damage abnormality, the power grid environment parameters at the corresponding detection node, and the power grid system parameters; wherein, the power grid system parameters include the total power generation of the power grid and the total load of the power grid; the power grid environment parameters include temperature, humidity, wind speed, and wire length;

[0034] Train a threshold prediction model of non-linear characteristic indexes based on the historical database;

[0035] Collect power grid system parameters and power grid environment parameters of each detection node in real time;

[0036] Calculate the maximum threshold of each non-linear characteristic index of the time series of each detection node based on the threshold prediction model;

[0037] Conduct preliminary external break detection for each detection node based on the non-linear characteristic indexes of the time series; if any non-linear characteristic index of any time series is greater than the corresponding maximum threshold, there is an external break anomaly in the corresponding detection node.

[0038] As a preferred scheme of the external break detection method for the power line described in the present invention, wherein: the input of the threshold prediction model includes power grid system parameters and power grid environment parameters at any detection node; the output of the threshold prediction model includes the maximum threshold of each non-linear characteristic index of the time series of the corresponding detection node;

[0039] Training the threshold prediction model of the non-linear characteristic index based on the historical database specifically includes:

[0040] Set retrieval conditions; the retrieval conditions include a set of power grid system parameters and a set of power grid environment parameters;

[0041] Mark each piece of historical data that meets the retrieval conditions in the historical database as sample historical data;

[0042] Calculate the non-linear characteristic indexes of the time series of each piece of sample historical data;

[0043] Extract the maximum value of each non-linear characteristic index of the time series of the sample historical data respectively, as the maximum threshold of each non-linear characteristic index corresponding to the retrieval conditions;

[0044] Form a piece of training data with the retrieval conditions and the maximum threshold of each non-linear characteristic index corresponding thereto;

[0045] Repeat to obtain at least N pieces of training data; train the threshold prediction model based on the training data.

[0046] As a preferred scheme of the external break detection method for the power line described in the present invention, wherein: the supplementary external break detection specifically includes:

[0047] Collect power grid system parameters and power grid environment parameters of the corresponding detection node in real time;

[0048] Encode the power grid system parameters and the power grid environment parameters of the corresponding detection node, and form a power grid feature vector;

[0049] Encode the non-linear characteristic indexes of each component sequence of any detection node into the feature vector of each component sequence;

[0050] Input the power grid feature vector and the feature vectors of each component sequence into the trained external damage detection model to obtain the supplementary external damage detection results for the corresponding detection nodes.

[0051] In a second aspect, the present invention provides an external damage detection system for a power line, including a data acquisition module, a correction module, a processing module, a prediction module, and an external damage detection module; wherein:

[0052] The data acquisition module is used to continuously collect the power parameters of each detection node, and collect the power grid system parameters and the power grid environment parameters at each detection node.

[0053] The correction module compensates and corrects the power parameters of the detection node according to the power parameters of the adjacent nodes of each detection node.

[0054] The processing module is used to establish a time series and a component sequence of the power parameters for each detection node, and calculate the non-linear feature indexes of each time series and component sequence.

[0055] The prediction module predicts the maximum threshold of each non-linear feature index of the time series of each detection node based on the historical database.

[0056] The external damage detection module is used to perform preliminary external damage detection and supplementary external damage detection on each detection node.

[0057] Compared with the prior art, the beneficial effects achieved by the present invention are as follows:

[0058] This application introduces a trustworthiness evaluation mechanism for detection nodes, dynamically calculates the trustworthiness by comprehensively considering factors such as the accuracy of historical data, the correlation degree with the data of adjacent nodes, and the distance between nodes, weights the data according to the trustworthiness of adjacent nodes, compensates and corrects the power parameters, avoids misjudgment caused by incorrect data of individual nodes, realizes multi-point linkage detection, and effectively improves the accuracy of detection.

[0059] Perform supplementary external damage detection based on the non-linear feature indexes of the component sequence, fuse multi-scale non-linear features, use classifiers such as support vector machines to judge the external damage risk, comprehensively consider the power parameter fluctuation characteristics at different scales and the influence of the power grid and the environment, and analyze the abnormal fluctuation of power parameters more comprehensively and accurately to detect external damage situations that are difficult to directly observe. Description of the Drawings

[0060] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts. Among them:

[0061] Figure 1 Flow chart of an external damage detection method for a power line provided by the present invention;

[0062] Figure 2 Structural schematic diagram of an external damage detection system for a power line provided by the present invention;

[0063] Figure 3 Flow chart of a method for compensating and correcting power parameters of any detection node provided by the present invention. Detailed implementation manners

[0064] The technical solution of the present invention will be described in detail below through the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations on the technical solution of the present invention. Without conflict, the technical features in the embodiments of the present invention and the embodiments can be combined with each other.

[0065] Embodiment 1

[0066] This embodiment introduces an external damage detection method for a power line. Referring to Figure 1 , the method includes the following steps:

[0067] Set detection nodes on the power line; continuously collect the power parameters of each detection node;

[0068] Set detection nodes at key positions along the power line, such as poles and towers, cable branch points, etc.; install sensors at each detection node to continuously collect the power parameters at the detection node; the power parameters can be any one of current, voltage, and power.

[0069] Based on the power parameters of the adjacent nodes of each detection node, compensate and correct the power parameters of each detection node; the method is as follows:

[0070] S100: Simultaneously collect the power parameters of all detection nodes;

[0071] S200: Compensate and correct the power parameters of any detection node. Referring to Figure 3 , specifically including:

[0072] S201: Obtain the power parameters of each adjacent node of the target detection node; obtain the historical power parameters of the target detection node and its adjacent nodes; the determination method of the adjacent nodes of the target detection node is as follows: any node adjacent to the target detection node on the power line or with a distance less than the preset distance threshold from the target detection node is an adjacent node of the target detection node.

[0073] S202: Calculate the accuracy rate of the target detection node and its neighboring nodes based on the historical power parameters; calculate the trust level of each neighboring node based on the accuracy rate.

[0074] The calculation method of the accuracy rate of any detection node is as follows: Count the abnormal power parameters in the historical power parameters of the detection node, and record the time stamp corresponding to each abnormal power parameter. The abnormal power parameters can be identified by setting the threshold range of the normal fluctuation of the power parameters, or based on the moving average method, calculate the average value of several consecutive historical power parameters, and the historical power parameters with a large deviation from the average value are abnormal power parameters. Obtain the maintenance records of the power line, and judge whether each abnormal power parameter corresponds to a line fault based on the maintenance records; mark the abnormal power parameters that do not correspond to the line fault as false alarm power parameters; count the total number of historical power parameters; divide the difference between the total number of historical power parameters and the number of false alarm power parameters by the total number of historical power parameters to obtain the accuracy rate of the corresponding detection node.

[0075] The calculation method of the trust level of any neighboring node of the target detection node is as follows:

[0076] Calculate the correlation coefficient between the historical power parameters of any neighboring node and the historical power parameters of the target detection node as the correlation degree between the corresponding neighboring node and the target detection node; preferably, in the embodiment of the present application, the historical power parameters of the neighboring node and the target detection node are aligned in time in advance, and then the Pearson correlation coefficient is calculated as the correlation degree. The greater the correlation degree, the higher the consistency and stronger the correlation of the power parameters between the neighboring node and the target detection node.

[0077] Obtain the distance between any neighboring node and the target detection node; calculate the trust level of the corresponding neighboring node based on the correlation degree, distance between any neighboring node and the target detection node, and the accuracy rate of the corresponding neighboring node. The formula is as follows:

[0078] ;

[0079] where C represents the trust level of any neighboring node of the target detection node; A represents the accuracy rate of the corresponding neighboring node; R represents the correlation degree between the corresponding neighboring node and the target detection node; d represents the distance between the corresponding neighboring node and the target detection node; represents the preset distance threshold.

[0080] S203: Calculate the compensation amount of the target detection node based on the trust level of each neighboring node; the method is as follows: Based on the trust level of each neighboring node, perform a weighted sum of the power parameters of each neighboring node to obtain the compensation amount of the target detection node; wherein, the weight coefficient of the power parameter of any neighboring node is the trust level of the corresponding neighboring node divided by the sum of the trust levels of all neighboring nodes.

[0081] S204: Compensate and correct the power parameters of the target detection node based on the compensation amount; let the power parameter of the target detection node after compensation and correction be E, and the original value of the power parameter of the target detection node be , and the compensation amount is , then E is obtained by and through weighted summation; wherein, the weight coefficient of is w, and

[0082] ;

[0083] wherein, represents the accuracy rate of the target detection node; represents the maximum value among the accuracy rates of all detection nodes; represents the adjustment coefficient, which is set by those skilled in the art based on actual requirements.

[0084] This application introduces a trust level evaluation mechanism for each detection node, and dynamically calculates the trust level of the node according to factors such as the accuracy of the node historical data, the correlation with the data of adjacent nodes, and the distance between nodes. For any node, weighted processing of the node data according to the trust level of the nearby nodes can, to a certain extent, avoid misjudgment caused by incorrect data of individual detection nodes, realize the detection of external damage to power lines with multi-point linkage, improve the detection accuracy, and thus solve the problem that the accuracy of the data of each detection node in long-distance power lines is affected by factors such as sensor accuracy and communication interference.

[0085] Establish a time series of power parameters for each detection node; obtain the component series of the power parameters of each detection node based on the time series;

[0086] This application continuously collects the power parameters of each detection node, and after compensation and correction, arranges them into a time series of power parameters. The power parameters after compensation and correction can provide a more accurate and comprehensive data basis for subsequent extraction of non-linear indicators, meet the very high requirements for data quality in non-linear indicator calculation, avoid the problem that noise and other interference factors in the power system affect the data quality and further affect the calculation accuracy of non-linear indicators.

[0087] For example, in a power system, by compensating and correcting the power parameters of a detection node with obvious sensor system errors using the relatively stable and accurate power parameters of adjacent nodes, the balancing effect of these adjacent nodes can be utilized to avoid misjudgment and make the final external damage detection more accurate.

[0088] In addition, the present application can also specifically set the time series length and sampling frequency of power parameters according to the characteristics of the power system where the power line is located, avoiding the problem that the time series cannot contain enough data to accurately calculate the non-linear characteristic index; for example, for a high-frequency circuit system with rapid changes, the length of the time series can be slightly shorter, but a higher sampling frequency is required.

[0089] Further, in order to improve the data quality, after establishing the time series of power parameters, a stationarity test is performed on it. If the time series is non-stationary, operations such as differencing are performed to make it stationary, preventing errors in the calculation results caused by the non-stationarity of the time series.

[0090] The component sequence includes an approximation component sequence and a detail component sequence; obtaining the component sequence of the power parameters of each detection node based on the time series specifically includes: selecting a wavelet basis function and a decomposition scale; performing wavelet transform on the time series of the power parameters of any detection node based on the selected wavelet basis function and decomposition scale to obtain the component sequence of the power parameters of the corresponding detection node.

[0091] In the embodiment of the present application, using the selected wavelet basis function and decomposition scale, wavelet transform is performed on the preprocessed time series of power parameters to decompose the time series into different scales. For example, performing 3-layer wavelet decomposition to obtain an approximation component sequence with a scale of 3 and 3 detail component sequences. The component sequences of the present application describe the characteristics of power parameters from different scale and frequency perspectives. The approximation component sequence retains the main trend and low-frequency information of the power parameters, while the detail component sequences contain the high-frequency details and fluctuation information of the power parameters at different scales.

[0092] Calculate the non-linear characteristic index of each time series and the non-linear characteristic index of each component sequence;

[0093] The non-linear characteristic index of any time series or component sequence includes Lyapunov exponent, correlation dimension, and information entropy.

[0094] The magnitude of the Lyapunov exponent reflects the speed of system state separation. When the power system operates normally, the Lyapunov exponents of power parameters such as current are usually within a relatively stable and small range. When an external break occurs in a power line, the changes in power parameters become complex and unstable, and the Lyapunov exponent will increase abnormally. Therefore, in the preferred preliminary external break detection method of the embodiments of this application, the Lyapunov exponent exceeding a certain threshold is used as one of the features for judging the external break of the wire. The correlation dimension describes the complexity of the system and can reflect the complex law of the change of power parameters in the power system. When the power system operates normally, the correlation dimensions of power parameters such as current are relatively stable. The external break of the wire causes irregular changes in power parameters, and the correlation dimension will also mutate. Therefore, in the preferred preliminary external break detection method of the embodiments of this application, the correlation dimension exceeding the corresponding threshold is used as one of the features for judging the external break of the wire. The information entropy is used to measure the uncertainty or degree of chaos of the system, and the information entropy of power parameters can reflect the disorder of their changes. The external break of the line will increase the disorder of power parameters, and the information entropy will increase accordingly. Therefore, in the preferred preliminary external break detection method of the embodiments of this application, the information entropy exceeding the corresponding threshold is used as one of the features for judging the external break of the wire.

[0095] This application captures the abnormal fluctuations of power parameters such as current and voltage caused by the external break of the power line by calculating the non-linear indicators of the time series of power parameters, thereby discovering tiny signs of external break and realizing the early warning of the risk of large-area external break, which is more forward-looking than the external break screening that conducts power line inspections after the external break occurs.

[0096] Based on the non-linear characteristic indicators of the time series, conduct preliminary external break detection for each detection node; specifically including:

[0097] S10: Collect historical data of power parameters and establish a historical database of power parameters; any piece of historical data in the historical database includes the time series of power parameters of a detection node without external break anomalies, the grid environment parameters at the corresponding detection node, and the grid system parameters; wherein, the grid system parameters include the total power generation of the grid and the total load of the grid; the grid environment parameters include temperature, humidity, wind speed, and wire length;

[0098] This application reduces the misjudgment of wire external damage by considering power grid system parameters and power grid environment parameters. More specifically, according to the power grid system parameters and power grid environment parameters collected in real time, this application dynamically adjusts the maximum thresholds of non-linear characteristic indicators such as Lyapunov exponents and correlation dimensions, so as to adapt to the changes of power parameters under different environments and working conditions, and avoid false alarms or missed alarms of external damage abnormalities. For example, when the total power generation and total load of the power grid change (such as the change in the power generation of renewable energy and the access of large-scale electrical loads), it will cause disturbances to the power grid stability, resulting in fluctuations in power parameters such as power grid frequency, power, current, and voltage even without wire external damage. Temperature and humidity will affect the physical characteristics of electrical equipment, further leading to fluctuations in power parameters. Strong winds will cause phenomena such as galloping and phase-to-phase discharge of transmission lines, and then cause violent fluctuations in power parameters. The wire length is the distance between the detection node and the starting end of the wire, and the starting end of the wire includes, but is not limited to, power generation stations, distribution stations, substations, etc. Longer transmission lines will increase the loss and signal delay during power transmission, affecting the stability of power parameters.

[0099] S20: Train a threshold prediction model for non-linear characteristic indicators based on the historical database;

[0100] The input of the threshold prediction model includes power grid system parameters and power grid environment parameters at any detection node; the output of the threshold prediction model includes the maximum threshold of each non-linear characteristic indicator of the time series corresponding to the detection node.

[0101] Training a threshold prediction model for non-linear characteristic indicators based on the historical database specifically includes:

[0102] Set retrieval conditions; the retrieval conditions include a set of power grid system parameters and a set of power grid environment parameters; more specifically, a set of power grid system parameters includes a value of the total power generation of the power grid and a value of the total load of the power grid; a set of power grid environment parameters includes one value each of temperature, humidity, wind speed, and wire length;

[0103] Mark each piece of historical data in the historical database that meets the retrieval conditions as sample historical data;

[0104] Meeting the retrieval conditions does not necessarily mean that each power grid system parameter and power grid environment parameter is exactly equal to the corresponding item in the retrieval conditions; preferably, in the embodiments of this application, a certain error range is set for each parameter to prevent the failure to retrieve historical data that meets the conditions. For example, the retrieval conditions include a wire length of 10 km, and the error range set for the wire length is 5%. Historical data with a wire length ranging from 9.5 km to 10.5 km all meet this item of the retrieval conditions regarding the wire length.

[0105] Calculate the non - linear characteristic indexes of the time series of each sample's historical data;

[0106] Extract the maximum value of each non - linear characteristic index of the time series of the sample's historical data respectively as the maximum threshold of each non - linear characteristic index corresponding to the retrieval condition;

[0107] Form a piece of training data with the retrieval condition and the maximum threshold of each non - linear characteristic index corresponding thereto;

[0108] Repeat to obtain at least N pieces of training data; train a threshold prediction model based on the training data. N is a positive integer. The threshold training model is any one of a linear regression model, a polynomial regression model, and a support vector regression model; when enough sample data is collected, the trained threshold prediction model can calculate the maximum threshold of each non - linear characteristic index under different power grid environment parameters and power grid system parameters, that is, the maximum normal value of each non - linear characteristic index in the case of no external break anomaly.

[0109] S30: Collect power grid system parameters and power grid environment parameters of each detection node in real - time;

[0110] S40: Calculate the maximum threshold of each non - linear characteristic index of the time series of each detection node based on the threshold prediction model;

[0111] S50: Conduct a preliminary external break detection for each detection node based on the non - linear characteristic indexes of the time series; if any non - linear characteristic index of any time series is greater than the corresponding maximum threshold, there is an external break anomaly in the corresponding detection node. When there is an external break anomaly in a certain detection node, it means that there is an external break in the power line at this detection node, thus affecting the stability of the change law of the power parameters of this detection node.

[0112] Preferably, set an abnormal redundancy for each non - linear characteristic index; if any non - linear characteristic index of any time series is greater than the sum of the corresponding maximum threshold and the corresponding abnormal redundancy, there is an external break anomaly in the corresponding detection node. This application further preferably uses 5% as the abnormal redundancy of the corresponding non - linear characteristic index, which can reduce a certain number of misjudgments without significantly affecting the sensitivity of the preliminary external break detection.

[0113] Conduct a supplementary external break detection for each detection node based on the non - linear characteristic indexes of the component series.

[0114] The supplementary external break detection specifically includes:

[0115] Collect power grid system parameters and power grid environment parameters of the corresponding detection node in real - time;

[0116] Encode the power grid system parameters and the power grid environment parameters of the corresponding detection nodes, and form a power grid feature vector;

[0117] Encode the non-linear feature indexes of each component sequence of any detection node into the feature vector of each component sequence;

[0118] Input the power grid feature vector and the feature vector of each component sequence into the trained external break detection model to obtain the supplementary external break detection result of the corresponding detection node.

[0119] The external break detection model is a classifier constructed based on any one of the support vector machine, Gaussian mixture model, and variational autoencoder; the external break detection model is iteratively trained through an optimization algorithm so that it can comprehensively consider the characteristics of power parameters at each scale (i.e., each component sequence) and the influence of the power grid and environment, and output an external break detection result that is difficult to directly observe. Among them, the non-linear feature indexes at the high-frequency scale can highlight the rapid changes in power parameters caused by sudden external force impacts, and the non-linear feature indexes at the low-frequency scale can reflect the changes in power parameters caused by long-term environmental factors, such as the influence of slow temperature changes on line resistance. Existing external break detections based on power parameters are basically carried out based on time series, ignoring the characteristics of power parameter fluctuations at different scales. In contrast, this application can analyze the abnormal fluctuations of power parameters more comprehensively and accurately, improving the accuracy of external break detection.

[0120] Embodiment 2

[0121] This embodiment is the second embodiment of the present invention; based on the same inventive concept as Embodiment 1, referring to Figure 2 , this embodiment introduces an external break detection system for a power line, including a data acquisition module, a correction module, a processing module, a prediction module, an external break detection module, and a display module; where:

[0122] The data acquisition module is used to continuously collect the power parameters of each detection node, and collect the power grid system parameters and the power grid environment parameters at each detection node;

[0123] The correction module compensates and corrects the power parameters of the detection node according to the power parameters of the adjacent nodes of each detection node; this module first obtains the power parameters and historical power parameters of the adjacent nodes of the target detection node, calculates the accuracy rate and confidence level, and then calculates the compensation amount; based on the compensation amount, the power parameters of the target detection node are corrected, so as to improve the data accuracy and provide a reliable data basis for subsequent analysis.

[0124] The processing module is used to establish the time series and component series of power parameters for each detection node, and calculate the non-linear characteristic indexes of each time series and component series; this module selects appropriate wavelet basis functions and decomposition scales, performs wavelet transform on the time series to obtain the approximate component series and detail component series, and then calculates the non-linear characteristic indexes (including Lyapunov exponent, correlation dimension, information entropy) of the time series and component series, so as to mine the non-linear characteristics of power parameters from different scales and frequency perspectives.

[0125] The prediction module is used to predict the maximum threshold of each non-linear characteristic index of the time series of each detection node based on the historical database; this module first establishes the historical database, and then trains the threshold prediction model of the non-linear characteristic index according to the historical data; the input of the trained model is the power grid system parameters and power grid environment parameters, and the output is the maximum threshold of each non-linear characteristic index of the time series of the corresponding detection node.

[0126] The external damage detection module is used to perform preliminary external damage detection and supplementary external damage detection on each detection node. The preliminary external damage detection includes judging whether each non-linear characteristic index of the time series exceeds the corresponding maximum threshold to determine whether there is an external damage abnormality in the detection node; the supplementary external damage detection includes inputting the power grid feature vector and the feature vector encoded by the non-linear characteristic index of the component series into the external damage detection model to obtain the supplementary external damage detection result.

[0127] The display module is used to summarize the results of the preliminary external damage detection and supplementary external damage detection, and output the detection results in an intuitive way (such as interface display, alarm prompt, etc.). When an external damage abnormality of the power line is detected, relevant personnel are notified in time for handling.

[0128] The specific function implementation of each of the above modules refers to the relevant content in the external damage detection method of the power line described in Embodiment 1, and will not be elaborated here.

[0129] Those skilled in the art should understand that the 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 complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0130] The embodiments of the present invention have been described above in conjunction with the accompanying drawings. However, the present invention is not limited to the above specific embodiments. The above specific embodiments are merely illustrative and not restrictive. Under the inspiration of the present invention, those of ordinary skill in the art can also make many forms without departing from the purpose and scope of the present invention. All of these fall within the protection scope of the present invention.

Claims

1. A method for detecting external damage of a power line, characterized in that: The following steps are involved: Setting up detection nodes on power lines; Continuously collect power parameters of each detection node; Based on the power parameters of the neighboring nodes of each detection node, the power parameters of each detection node are compensated and corrected; the method is as follows: S100: Collecting power parameters of all detection nodes simultaneously; S200: compensating and correcting the power parameters of any detection node, specifically including: Obtain power parameters of each neighboring node of the target detection node; obtain historical power parameters of the target detection node and its neighboring nodes; Calculating the accuracy of the target detection node and its adjacent nodes based on the historical power parameters; calculating the trust of each adjacent node based on the accuracy; Calculate the compensation amount of the target detection node based on the trust of each neighboring node; Performing compensation correction on the power parameters of the target detection node based on the compensation amount; Establishing a time series of power parameters for each detection node; acquiring a component series of power parameters of each detection node based on the time series; Calculate the nonlinear characteristic index of each time series and the nonlinear characteristic index of each component series; A preliminary external failure detection is performed on each detection node based on the nonlinear characteristic index of the time series; a supplementary external failure detection is performed on each detection node based on the nonlinear characteristic index of the component series.

2. A method for detecting external damage to a power line according to claim 1, characterized in that: The accuracy of any detection node is calculated as follows: count the abnormal power parameters in the historical power parameters of the detection node, and record the timestamp corresponding to each abnormal power parameter; Obtain maintenance records of power lines, and determine whether each abnormal power parameter corresponds to a line fault based on the maintenance records; Marking abnormal power parameters that do not correspond to line faults as false alarm power parameters; Count the total number of historical power parameters; The difference between the total number of historical power parameters and the number of false alarm power parameters is divided by the total number of historical power parameters to obtain the accuracy of the corresponding detection node.

3. A method for detecting external damage to a power line according to claim 2, characterized in that: The neighboring nodes of the target detection node are determined as follows: any node that is adjacent to the target detection node on the power line or whose distance to the target detection node is less than a preset distance threshold is a neighboring node of the target detection node; the trustworthiness of any neighboring node of the target detection node is calculated as follows: Calculate the correlation coefficient between the historical power parameters of any adjacent node and the historical power parameters of the target detection node as the correlation degree between the corresponding adjacent node and the target detection node; Get the distance between any adjacent node and the target detection node; The trust of any adjacent node is calculated based on the correlation and distance between the adjacent node and the target detection node, as well as the accuracy of the adjacent node.

4. A method for detecting external damage to a power line as claimed in claim 3, characterized in that: The method of calculating the compensation amount of the target detection node based on the trust of each neighboring node specifically includes: based on the trust of each neighboring node, performing weighted summation on the power parameters of each neighboring node to obtain the compensation amount of the target detection node; wherein the weight coefficient of the power parameter of any neighboring node is the trust of the corresponding neighboring node divided by the sum of the trusts of all neighboring nodes; The compensating and correcting the power parameter of the target detection node based on the compensation amount specifically includes: Based on the weighted sum of the original value of the power parameter of the target detection node and the compensation amount, the power parameter of the target detection node is compensated and corrected; wherein the weight coefficients of the compensation amount and the original value are both related to the accuracy of the target detection node.

5. A method for detecting external damage to a power line as claimed in claim 4, characterized in that: The component sequence includes an approximate component sequence and a detail component sequence; obtaining a component sequence of the power parameters of each detection node based on the time sequence specifically includes: selecting a wavelet basis function and a decomposition scale; performing a wavelet transform on the time series of the power parameters of any detection node based on the selected wavelet basis function and the decomposition scale to obtain a component sequence of the power parameters of the corresponding detection node; The nonlinear characteristic indicators of any time series or component series include Lyapunov exponent, correlation dimension, and information entropy.

6. A method for detecting external damage to a power line according to claim 5, characterized in that: The preliminary external damage detection specifically includes: Collect historical data of power parameters and establish a historical database of power parameters; any historical data in the historical database includes a time series of power parameters of a detection node without external abnormalities, power grid environmental parameters at the corresponding detection node, and power grid system parameters; wherein the power grid system parameters include the total power generation power of the power grid and the total load of the power grid; the power grid environmental parameters include temperature, humidity, wind speed, and power line length; Training a threshold prediction model for nonlinear characteristic indicators based on the historical database; Collect grid system parameters and grid environment parameters of each detection node in real time; Calculate the maximum threshold of each nonlinear characteristic index of the time series of each detection node based on the threshold prediction model; A preliminary external failure detection is performed on each detection node based on the nonlinear characteristic indicators of the time series; if any nonlinear characteristic indicator of any time series is greater than the corresponding maximum threshold, the corresponding detection node has an external failure anomaly.

7. A method for detecting external damage to a power line according to claim 6, characterized in that: The input of the threshold prediction model includes the power grid system parameters and the power grid environment parameters at any detection node; the output of the threshold prediction model includes the maximum threshold of each nonlinear characteristic index of the time series of the corresponding detection node; The threshold prediction model for training nonlinear characteristic indicators based on the historical database specifically includes: Setting search conditions; the search conditions include a set of power grid system parameters and a set of power grid environment parameters; Mark each piece of historical data in the historical database that meets the search criteria as sample historical data; Calculate the nonlinear characteristic index of the time series of each sample historical data; The maximum value of each nonlinear characteristic index of the time series of the sample historical data is extracted respectively as the maximum threshold value of each nonlinear characteristic index corresponding to the retrieval condition; The retrieval condition and the maximum threshold of each corresponding nonlinear feature index form a piece of training data; Repeatedly obtain at least N training data; and train a threshold prediction model based on the training data.

8. A method for detecting external damage to a power line according to claim 7, characterized in that: The supplementary external damage detection specifically includes: Collect power grid system parameters and power grid environment parameters of corresponding detection nodes in real time; Encode the grid system parameters and the grid environment parameters of the corresponding detection nodes and form a grid feature vector; Encode the nonlinear characteristic index of each component sequence of any detection node into a feature vector of each component sequence; The power grid feature vector and the feature vector of each component sequence are input into a trained external failure detection model to obtain a supplementary external failure detection result of a corresponding detection node.

9. A power line external failure detection system, which is used to implement a power line external failure detection method according to any one of claims 1 to 8, characterized in that: It includes data acquisition module, correction module, processing module, prediction module and external damage detection module; among which: The data acquisition module is used to continuously collect the power parameters of each detection node, and collect the power grid system parameters and the power grid environment parameters at each detection node; The correction module compensates and corrects the power parameters of each detection node according to the power parameters of the neighboring nodes of each detection node; The processing module is used to establish a time series and a component series of power parameters for each detection node, and calculate the nonlinear characteristic index of each time series and component series; The prediction module predicts the maximum threshold of each nonlinear characteristic index of the time series of each detection node based on the historical database; The external breach detection module is used to perform preliminary external breach detection and supplementary external breach detection on each detection node.

Images