A method, device and system for diagnosing communication faults in a distribution network

Abstract

The present invention relates to the field of electrical fault detection technology, and specifically to a distribution network communication fault diagnosis method, device and system. The method includes: obtaining load and current data of power grid nodes at different times; dividing time sub-segments based on the load, and determining the current frequency stability coefficient according to the current fluctuation; then, combining the distance between power grid nodes and the difference in the current frequency stability coefficient, determining the comprehensive stability difference index, and screening abnormal sub-segments according to the comprehensive stability difference index; finally, combining the time interval and time proportion of the abnormal sub-segments, determining the distribution density, and performing outlier detection based on the distribution density, determining abnormal current data, and marking it as fault data. This solution can more accurately identify abnormal points and improve the accuracy of fault detection.

Description

Technical Field

[0001] The present invention relates to the technical field of electrical fault detection, and in particular to a method, device and system for diagnosing communication faults in a distribution network. Background Art

[0002] The distribution network refers to a power grid that receives electric energy from the transmission network or regional power plants and distributes it locally or step by step according to voltage to various types of users through distribution facilities. The distribution network is at the end of the power grid and has a weak ability to cope with fault disturbances. In addition, many communication devices in the distribution network are installed outdoors and are easily affected by high voltage and high electromagnetic interference, causing current data at the distribution network nodes to fluctuate and distort. Data distortion will affect the real-time monitoring function of the power system, making it impossible for operators to accurately understand the operating status of the system, resulting in potential safety hazards not being handled in a timely manner, affecting the optimized operation and dispatching decisions of the power system, and reducing the operating efficiency of the system.

[0003] In related technologies, the LOF algorithm is often used to perform fault diagnosis on the current data of each node in the distribution network. In the LOF algorithm, the size of the k value (neighborhood reachable distance) has a significant impact on the anomaly detection results. The choice of k value determines the number of data points considered by the algorithm when calculating the local density, which affects the identification of anomalies. Traditional LOF algorithms usually use a fixed K value that remains unchanged throughout the detection process, but due to the different current fluctuations and data characteristics of the power grid at different nodes, the accuracy of anomaly identification and fault detection is low. Summary of the invention

[0004] In order to solve the technical problem in the related art that the current fluctuations and data characteristics of different nodes in the power grid are different, which leads to low accuracy in identifying abnormal points and low accuracy in fault detection, the present invention provides a distribution network communication fault diagnosis method, device and system, and the technical solutions adopted are as follows:

[0005] The present invention proposes a method for diagnosing distribution network communication faults, the method comprising:

[0006] Obtain load and current data of power grid nodes at different times; take any power grid node as a target node, and divide the target node into different time sub-segments according to the load values ​​of the target node at different times within a preset first time period;

[0007] Determine the current frequency stability coefficient of the target node in each time subsegment according to the quantity and extreme value fluctuation of the current data corresponding to the target node in each time subsegment;

[0008] Determine the comprehensive stability difference index of the target node in each time subsegment according to the difference in current frequency stability coefficient between the target node and any other power grid node in the same time subsegment, and the distance between the target node and other power grid nodes; and screen out abnormal subsegments of all time subsegments in the target node according to the comprehensive stability difference index;

[0009] According to the time intervals of different abnormal sub-segments in the target node and the time proportion of the abnormal sub-segments, the distribution density of the abnormal sub-segments is determined, and the adaptive neighborhood reachable distance of the target node is determined according to the distribution density. According to the adaptive neighborhood reachable distance of each power grid node, outlier detection is performed on the current data in the power grid, abnormal current data is determined, and marked as fault data.

[0010] Furthermore, the segmentation is performed according to the load values ​​of the target node at different times, and is divided into different time sub-segments, including:

[0011] Based on the APCA segmentation method, all moments are segmented according to the load of the target node at different moments to obtain time sub-segments.

[0012] Furthermore, the current frequency stability coefficient of the target node in each time subsegment is determined according to the quantity and extreme value fluctuation of the current data corresponding to the target node in each time subsegment, including:

[0013] The number of current data in each time sub-segment is normalized to the maximum and minimum values ​​to obtain the sub-segment quantity index;

[0014] Perform extreme value detection on each time sub-segment to obtain extreme value points, and determine the mean of the time intervals between any extreme value point and the two nearest extreme value points as the extreme value time interval of the extreme value point;

[0015] Calculate the product of the variance of the current data corresponding to all extreme value points and the variance of the extreme value time interval, and perform maximum and minimum value normalization on the inverse of the product to obtain the extreme value fluctuation index of the time sub-segment;

[0016] The product of the sub-segment quantity index and the extreme value fluctuation index of the same time sub-segment is taken as the current frequency stability coefficient of the corresponding time sub-segment.

[0017] Furthermore, the comprehensive stability difference index of the target node in each time subsegment is determined according to the difference in current frequency stability coefficient between the target node and any other grid node in the same time subsegment, and the distance between the target node and the other grid nodes, including:

[0018] Calculate the absolute value of the difference between the current frequency stability coefficient of the target node and any other grid node in the same time subsegment as the stability difference index;

[0019] The inverse of the distance between the target node and any other grid node is normalized to obtain a distance impact index;

[0020] The product of the stability difference index and the distance influence index between the target node and the same other power grid nodes is used as the stability difference influence coefficient between the target node and the corresponding other power grid nodes in the same time subsegment;

[0021] The mean of the stability difference influence coefficients between the target node and all other grid nodes is taken as the comprehensive stability difference index of the target node in the corresponding time subsegment.

[0022] Furthermore, the abnormal sub-segments of all time sub-segments in the target node are screened according to the comprehensive stability difference index, including:

[0023] The time sub-segment in which the comprehensive stability difference index is greater than the preset index threshold is regarded as an abnormal sub-segment.

[0024] Furthermore, according to the time intervals of different abnormal sub-segments in the target node and the time proportions of the abnormal sub-segments, the distribution density of the abnormal sub-segments is determined, including:

[0025] The middle moment of each abnormal sub-segment is taken as the representative moment of the corresponding abnormal sub-segment, and the time interval between the representative moments of another abnormal sub-segment that is closest to each abnormal sub-segment in time sequence is taken as the abnormal interval of the corresponding abnormal sub-segment;

[0026] Calculate the mean of the abnormal intervals of all abnormal sub-segments, and normalize the inverse of the mean as the abnormal distribution indicator;

[0027] The ratio of the number of moments belonging to the abnormal sub-segment to the number of all moments in the preset first time period is used as a moment proportion indicator;

[0028] The product of the abnormal distribution index and the time proportion index is calculated as the distribution density of the abnormal sub-segment.

[0029] Further, determining the adaptive neighborhood reachable distance of the target node according to the distribution density includes:

[0030] Perform negative correlation normalization on the distribution density to obtain the density impact weight;

[0031] The product of the density influence weight and the preset reachable distance is rounded up as the adaptive neighborhood reachable distance.

[0032] Furthermore, according to the adaptive neighborhood reachable distance of each grid node, outlier detection is performed on the current data in the grid to determine abnormal current data, including:

[0033] Based on the LOF algorithm, the adaptive neighborhood reachable distance is used as the K value for outlier detection to determine the abnormal current data.

[0034] On the other hand, a distribution network communication fault diagnosis device is also provided, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and when the processor executes the computer program, the steps of a distribution network communication fault diagnosis method as described in any one of the preceding items are implemented.

[0035] On the other hand, a distribution network communication fault diagnosis system is also provided, the system comprising:

[0036] An acquisition module is used to acquire load and current data of power grid nodes at different times; taking any power grid node as a target node, and segmenting the target node according to load values ​​at different times within a preset first time period, and dividing the target node into different time sub-segments;

[0037] A stability analysis module, used to determine the current frequency stability coefficient of the target node in each time sub-segment according to the quantity and extreme value fluctuation of the current data corresponding to the target node in each time sub-segment;

[0038] The anomaly detection module is used to determine the comprehensive stability difference index of the target node in each time subsegment according to the difference in current frequency stability coefficient between the target node and any other power grid node in the same time subsegment, and the distance between the target node and other power grid nodes; and screen out abnormal subsegments of all time subsegments in the target node according to the comprehensive stability difference index;

[0039] The fault identification module is used to determine the distribution density of abnormal sub-segments according to the time intervals of different abnormal sub-segments in the target node and the time proportion of the abnormal sub-segments, determine the adaptive neighborhood reachable distance of the target node according to the distribution density, perform outlier detection on the current data in the power grid according to the adaptive neighborhood reachable distance of each power grid node, determine the abnormal current data, and mark it as fault data.

[0040] The present invention has the following beneficial effects:

[0041] This scheme obtains the load and current data of the grid nodes at different times; divides the time sub-segments based on the load, and determines the current frequency stability coefficient according to the current fluctuation; then, combined with the distance of the grid nodes and the difference in the current frequency stability coefficient, determines the comprehensive stability difference index, and screens the abnormal sub-segments according to the comprehensive stability difference index; because it combines the current fluctuations of multiple different grid nodes, compared with directly analyzing the abnormality based on the current fluctuations of a single grid node, this scheme can combine the current similarity characteristics of multiple nodes to conduct objective regional analysis of the fault, thereby avoiding the abnormal fluctuations of a single grid node affecting the fault analysis of the grid node; finally, combined with the time interval and time proportion of the abnormal sub-segment, determine the distribution density, and perform outlier detection based on the distribution density, determine the abnormal current data, and mark it as fault data. This scheme can combine the current fluctuation distribution and data characteristics of multiple grid nodes, determine the adaptive neighborhood reachable distance based on the distribution density, and can more accurately identify abnormal points and improve the accuracy of fault detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0043] Figure 1 A flow chart of a method for diagnosing communication faults in a distribution network provided by an embodiment of the present invention;

[0044] Figure 2 A schematic diagram of a distribution network system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0045] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of a distribution network communication fault diagnosis method, device and system proposed by the present invention, its specific implementation method, structure, features and effects in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0046] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0047] A specific scheme of a distribution network communication fault diagnosis method provided by the present invention is described in detail below with reference to the accompanying drawings.

[0048] See also Figure 1 , which shows a flow chart of a distribution network communication fault diagnosis method provided by an embodiment of the present invention, the method comprising:

[0049] S101: Obtain load and current data of power grid nodes at different times; take any power grid node as a target node, and within a preset first time period, segment the target node according to the load values ​​at different times, and divide it into different time sub-segments.

[0050] See also Figure 2 , Figure 2 A schematic diagram of a distribution network system is provided for an embodiment of the present invention. In the figure, the power grid is divided into several major ends, namely, power transmission, power distribution, and power consumption. The power grid nodes for power distribution are distribution facilities for local distribution or step-by-step distribution according to voltage. Different distribution facilities are located in different positions, and power distribution to different power consumption ends is achieved according to the principle of local distribution or the principle of step-by-step voltage distribution.

[0051] Many communication devices in the distribution network are installed outdoors and are easily affected by high voltage and high electromagnetic interference, causing fluctuations and distortions in the current data of the grid nodes in the distribution network. Data distortion will affect the real-time monitoring function of the power system, making it impossible for operators to accurately understand the operating status of the system, resulting in potential safety hazards not being dealt with in a timely manner, affecting the optimized operation and scheduling decisions of the power system, and reducing the operating efficiency of the system.

[0052] Based on this, this solution improves the accuracy of fault diagnosis by diagnosing the current conditions in the distribution network.

[0053] The preset first time period may be, for example, 24 hours, that is, the load and current data of different power grid nodes may be collected every 10 seconds within 24 hours, so as to perform statistics as a data basis.

[0054] The load refers to the electricity demand load at the electricity consumption end, which is objectively affected by time. For example, during the day, the electricity consumption is high and the load is large, while in the middle of the night, the electricity consumption is low and the load is small. Different areas are connected to different electricity consumption ends, such as the difference between industrial areas and residential areas, and their loads will also change.

[0055] Therefore, segmentation can be performed based on load to facilitate abnormality analysis. Furthermore, in some embodiments of the present invention, segmentation is performed based on the load values ​​of the target node at different times and divided into different time sub-segments, including: based on the APCA segmentation method, all times are segmented in time series according to the load of the target node at different times to obtain time sub-segments.

[0056] Among them, the APCA segmentation method is an adaptive piecewise constant approximation (APCA) algorithm, which can divide the values ​​that are similar and adjacent into a group. In this scheme, this method is used to realize the time segmentation of the load, so that the moments with similar loads are divided into a time sub-segment, and different time sub-segments are obtained. The loads in each time sub-segment are similar, eliminating the impact of load changes, so that the reliability and accuracy of the subsequent current fluctuation analysis are stronger.

[0057] S102: Determine a current frequency stability coefficient of the target node in each time sub-segment according to the quantity and extreme value fluctuation of the current data corresponding to the target node in each time sub-segment.

[0058] Among them, the more violent the extreme value fluctuation in each time sub-segment, the worse the stability, and the smaller the amount of current data, the greater the load change and the worse the stability. Therefore, the stability of the current frequency fluctuation itself can be analyzed based on the amount of current data and the extreme value fluctuation. In the embodiment of the present invention, the current frequency stability coefficient characterizes the stability of the current frequency fluctuation.

[0059] Furthermore, in some embodiments of the present invention, the current frequency stability coefficient of the target node in each time sub-segment is determined based on the number of current data corresponding to the target node in each time sub-segment and the extreme fluctuation, including: performing maximum and minimum value normalization processing on the number of current data in each time sub-segment to obtain a sub-segment quantity index; performing extreme value detection on each time sub-segment to obtain an extreme point, and determining the mean of the time intervals between any extreme point and the two nearest extreme points as the extreme time interval of the extreme point; calculating the product of the variance of the current data corresponding to all extreme points and the variance of the extreme time interval, and performing maximum and minimum value normalization processing on the inverse of the product to obtain the extreme fluctuation index of the time sub-segment; and taking the product of the sub-segment quantity index and the extreme fluctuation index of the same time sub-segment as the current frequency stability coefficient of the corresponding time sub-segment.

[0060] Among them, the maximum and minimum value normalization is a linear normalization method well known to those skilled in the art, and will not be further limited or elaborated on.

[0061] Among them, the sub-segment quantity index is the number of moments in the time sub-segment. Since the moments are periodic collection moments, the sub-segment quantity index can also be analogized to the duration of the time sub-segment. That is, the larger the sub-segment quantity index is, the longer the duration of the corresponding time sub-segment is, the smaller the load change is, and the more stable the overall situation is.

[0062] Among them, extreme value detection is the detection of maximum value points and minimum value points. In the embodiment of the present invention, only the maximum value point can be analyzed as the extreme value point, or the maximum value point and the minimum value point can be analyzed together as the extreme value points. The overall impact of the analysis result is selected according to the actual detection needs.

[0063] In the embodiment of the present invention, the mean of the time intervals between any extreme point and the two nearest extreme points is determined as the extreme time interval of the extreme point, and the extreme time interval characterizes the frequency characteristics of the current fluctuation. The variance of the extreme time interval is subsequently calculated. The larger the variance, the more unstable the frequency. Similarly, the variance of the current data is calculated. The larger the value, the more unstable it is.

[0064] Afterwards, the product of the variance of the current data corresponding to all extreme points and the variance of the extreme time interval is calculated, and the inverse of the product is normalized to the maximum and minimum values ​​to obtain the extreme value fluctuation index of the time sub-segment. The extreme value fluctuation index characterizes the extreme value fluctuation characteristics. The larger the value of the extreme value fluctuation index, the smaller the fluctuation of the current data and the more stable its overall frequency fluctuation.

[0065] Since the larger the sub-segment quantity index is, the longer the corresponding time sub-segment lasts, the smaller the load change is, and the overall stability is, and the larger the value of the extreme value fluctuation index is, the smaller the fluctuation of the current data is, and the more stable the overall frequency fluctuation is; therefore, the product of the sub-segment quantity index and the extreme value fluctuation index can be calculated to obtain the current frequency stability coefficient of the corresponding time sub-segment.

[0066] S103: Determine the comprehensive stability difference index of the target node in each time sub-segment based on the difference in current frequency stability coefficient between the target node and any other power grid node in the same time sub-segment, and the distance between the target node and other power grid nodes; and screen out abnormal sub-segments of all time sub-segments in the target node based on the comprehensive stability difference index.

[0067] Among them, since the current frequency stability coefficients of different power grid nodes have certain differences, the greater the difference between the stability of any power grid node as a whole and the stability of other power grid nodes, the more abnormal the power grid node is. Based on this, abnormal analysis can be performed and the time can be limited to the same time sub-segment.

[0068] It should be noted that the time sub-segments obtained by dividing the target node should have similar stability for other nodes according to the objective load demand in order to meet the normal power distribution of the power grid. Therefore, the time sub-segments divided by the target node can be used as a benchmark for abnormal analysis.

[0069] For example, all moments between 10:00-11:00 in the target node are taken as a time sub-segment. Then, when performing data analysis on other power grid nodes, the data of all moments between 10:00-11:00 in other power grid nodes are also used to calculate the current frequency stability coefficient, thereby facilitating comparative analysis.

[0070] Furthermore, in some embodiments of the present invention, a comprehensive stability difference index of the target node in each time sub-segment is determined based on the difference in current frequency stability coefficients between the target node and any other power grid node in the same time sub-segment, and the distance between the target node and other power grid nodes, including: calculating the absolute value of the difference in current frequency stability coefficients between the target node and any other power grid node in the same time sub-segment as a stability difference index; normalizing the inverse of the distance between the target node and any other power grid node to obtain a distance influence index; taking the product of the stability difference index and the distance influence index between the target node and the same other power grid node as the stability difference influence coefficient between the target node and the corresponding other power grid node in the same time sub-segment; and taking the average of the stability difference influence coefficients between the target node and all other power grid nodes as the comprehensive stability difference index of the target node in the corresponding time sub-segment.

[0071] Among them, the stability difference index represents the stability difference between two power grid nodes in the same time sub-segment.

[0072] It should be noted that the various grid nodes in the distribution network are interconnected through power lines to form an electrical network. This physical connection allows electrical energy to flow between grid nodes, resulting in a certain correlation between currents in different grid nodes, and the distribution of various grid nodes in the distribution network usually has a certain geographical pattern. Grid nodes that are geographically close are usually connected by shorter power lines, and the impedance characteristics of these lines are relatively similar. Under the same power supply conditions, the impact of line impedance on current is relatively consistent. When current flows between these grid nodes, due to the similar impedance, the change trend of the current will also be closer, making the frequency stability of the current relatively similar. Therefore, distance is also an important indicator for stability difference analysis.

[0073] In the same area, whether it is the start and stop of equipment in industrial production or the electricity consumption habits of residents, there is usually a certain regularity. When the current frequency stability of a certain period of time is greatly different from that of other days in the preset time period, the possibility of being affected by noise is greater. Because noise interference is usually random and uncertain. For example, natural phenomena such as electromagnetic interference and lightning can cause fluctuations and distortions in the current signal, thereby destroying the frequency stability of the current.

[0074] Among them, the distance impact index represents the distance characteristics of two power grid nodes in the same time sub-segment. The larger the value of the distance impact index is, the closer the distance between the two power grid nodes is, that is, the more similar they should be to each other under normal circumstances, and the greater the impact.

[0075] In the embodiment of the present invention, the product of the stability difference index and the distance influence index between the target node and the same other power grid node is used as the stability difference influence coefficient between the target node and the corresponding other power grid node in the same time subsegment.

[0076] The stability difference index is weighted by the distance influence index. The larger the value of the stability difference index and the larger the value of the distance influence index, the closer the two are to each other and the greater the stability difference. At this time, the current change between the target node and the corresponding power grid node is more abnormal. When the value of the stability difference index is large but the value of the distance influence index is small, it means that the distance is far and the distance influence is small. When the value of the stability difference index is small but the value of the distance influence index is small, it means that the stability between the two is similar and the mutual influence is small. When the value of the stability difference index is small but the value of the distance influence index is large, it means that the distance is close and the similarity is higher, and the current change is more normal.

[0077] Based on the above analysis, the stability difference influence coefficient can characterize the objective stability difference between the target node and the corresponding other power grid nodes. Therefore, the mean value of the stability difference influence coefficient between the target node and all other power grid nodes is calculated as the comprehensive stability difference index of the target node in the corresponding time sub-segment.

[0078] It can be understood that the larger the comprehensive stability difference index is, the greater the difference between the target node and all other power grid nodes in the time sub-segment, that is, the more abnormal the target node itself is in the corresponding time sub-segment. Therefore, in an embodiment of the present invention, an abnormality analysis can be performed on each time sub-segment of the target node based on the comprehensive stability difference index.

[0079] Furthermore, in some embodiments of the present invention, abnormal sub-segments of all time sub-segments in the target node are screened according to the comprehensive stability difference index, including: taking the time sub-segment whose comprehensive stability difference index is greater than a preset index threshold as an abnormal sub-segment.

[0080] The preset indicator threshold is a threshold value of the comprehensive stability difference indicator. Optionally, the preset indicator threshold may be specifically 0.7, for example, that is, the time sub-segment when the comprehensive stability difference indicator is greater than 0.7 is regarded as an abnormal sub-segment.

[0081] S104: Determine the distribution density of the abnormal sub-segments according to the time intervals of different abnormal sub-segments in the target node and the time proportion of the abnormal sub-segments, determine the adaptive neighborhood reachable distance of the target node according to the distribution density, perform outlier detection on the current data in the power grid according to the adaptive neighborhood reachable distance of each power grid node, determine the abnormal current data, and mark it as fault data.

[0082] Among them, the distribution density of the abnormal sub-segments can represent the overall abnormal distribution of the current target node within the preset first time period. Anomaly detection based on this can improve detection accuracy.

[0083] Furthermore, in some embodiments of the present invention, the distribution density of abnormal sub-segments is determined based on the time intervals of different abnormal sub-segments in the target node and the moment proportion of the abnormal sub-segments, including: taking the middle moment of each abnormal sub-segment as the representative moment of the corresponding abnormal sub-segment, and taking the time interval between the representative moments of another abnormal sub-segment that is closest to each abnormal sub-segment in time sequence as the abnormal interval of the corresponding abnormal sub-segment; calculating the mean of the abnormal intervals of all abnormal sub-segments, normalizing the opposite of the mean as an abnormal distribution index; taking the ratio of the number of moments belonging to the abnormal sub-segment to the number of all moments in a preset first time period as a moment proportion index; and calculating the product of the abnormal distribution index and the moment proportion index as the distribution density of the abnormal sub-segment.

[0084] Among them, the representative moment is the midpoint of the duration corresponding to the abnormal sub-segment, for example, between 10:00-11:00, the middle moment is 10:30. The representative moment can characterize the position of the abnormal sub-segment in the time series. The time interval between the representative moments of the two nearest abnormal sub-segments is used as the abnormal interval of the corresponding abnormal sub-segment. Thus, the abnormal interval of each abnormal sub-segment is calculated and averaged. The larger the mean, the farther the two corresponding abnormal sub-segments are apart, that is, the lower the abnormal distribution density. Therefore, the inverse of the mean is normalized as the abnormal distribution index. The larger the value of the abnormal distribution index, the higher the abnormal distribution density.

[0085] Among them, the ratio of the number of moments belonging to the abnormal sub-segment to the number of all moments in the preset first time period is used as the moment proportion index, that is, the larger the moment proportion index is, the more the abnormal sub-segment accounts for in the time series and the higher the abnormal distribution density is.

[0086] In summary, the product of the abnormal distribution index and the moment proportion index is calculated as the distribution density of the abnormal sub-segment.

[0087] It is understandable that different power grid nodes can calculate the distribution density of their adaptive abnormal sub-segments, which is convenient for subsequent abnormal detection of power grid nodes according to the distribution density of the abnormal sub-segments.

[0088] Furthermore, in some embodiments of the present invention, the adaptive neighborhood reachable distance of the target node is determined based on the distribution density, including: performing negative correlation normalization processing on the distribution density to obtain a density influence weight; and rounding up the product of the density influence weight and the preset reachable distance as the adaptive neighborhood reachable distance.

[0089] Among them, the adaptive neighborhood reachable distance, that is, the reachable distance when performing neighborhood detection, can specifically use the LOF algorithm for anomaly detection in the embodiment of the present invention, and the adaptive neighborhood reachable distance can be used as the K value of the LOF algorithm (the kth other power grid node closest to the central power grid node).

[0090] In the embodiment of the present invention, the preset reachable distance is a preset value. Specifically, the preset reachable distance may be 20, for example.

[0091] Among them, the density impact weight represents the impact value of the distribution density on the adaptive neighborhood reachable distance. It can be understood that the larger the distribution density, the denser the abnormal sub-segments. At this time, the k value needs to be reduced to achieve more refined abnormal analysis. The smaller the distribution density, the sparser the abnormal sub-segments. At this time, the k value can be increased to ensure the overall abnormal analysis efficiency.

[0092] In an embodiment of the present invention, negative correlation normalization can be used to process the distribution density, that is, the larger the distribution density, the smaller the value of the density influence weight, and the density influence weight value range is (0,1). In an embodiment of the present invention, negative correlation normalization can be achieved using exp(-x), where exp represents an exponential function with a natural constant as the base, x represents the distribution density, and exp(-x) represents the density influence weight.

[0093] Thus, the abnormality detection of the LOF algorithm is performed to obtain abnormal current data, and the abnormal current data is marked as fault data.

[0094] This scheme obtains the load and current data of the grid nodes at different times; divides the time sub-segments based on the load, and determines the current frequency stability coefficient according to the current fluctuation; then, combined with the distance of the grid nodes and the difference in the current frequency stability coefficient, determines the comprehensive stability difference index, and screens the abnormal sub-segments according to the comprehensive stability difference index; because it combines the current fluctuations of multiple different grid nodes, compared with directly analyzing the abnormality based on the current fluctuations of a single grid node, this scheme can combine the current similarity characteristics of multiple nodes to conduct objective regional analysis of the fault, thereby avoiding the abnormal fluctuations of a single grid node affecting the fault analysis of the grid node; finally, combined with the time interval and time proportion of the abnormal sub-segment, determine the distribution density, and perform outlier detection based on the distribution density, determine the abnormal current data, and mark it as fault data. This scheme can combine the current fluctuation distribution and data characteristics of multiple grid nodes, determine the adaptive neighborhood reachable distance based on the distribution density, and can more accurately identify abnormal points and improve the accuracy of fault detection.

[0095] On the other hand, the present solution also provides a distribution network communication fault diagnosis device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of any of the aforementioned methods are implemented.

[0096] On the other hand, the present solution also provides a distribution network communication fault diagnosis system, the system comprising:

[0097] An acquisition module is used to acquire load and current data of power grid nodes at different times; taking any power grid node as a target node, and segmenting the target node according to load values ​​at different times within a preset first time period, and dividing the target node into different time sub-segments;

[0098] A stability analysis module, used to determine the current frequency stability coefficient of the target node in each time sub-segment according to the quantity and extreme value fluctuation of the current data corresponding to the target node in each time sub-segment;

[0099] The anomaly detection module is used to determine the comprehensive stability difference index of the target node in each time subsegment according to the difference in the current frequency stability coefficient between the target node and any other power grid node in the same time subsegment, and the distance between the target node and other power grid nodes; and screen out the abnormal subsegments of all time subsegments in the target node according to the comprehensive stability difference index;

[0100] The fault identification module is used to determine the distribution density of abnormal sub-segments according to the time intervals of different abnormal sub-segments in the target node and the time proportion of the abnormal sub-segments, determine the adaptive neighborhood reachable distance of the target node according to the distribution density, perform outlier detection on the current data in the power grid according to the adaptive neighborhood reachable distance of each power grid node, determine the abnormal current data, and mark it as fault data.

[0101] It should be noted that the distribution network communication fault diagnosis device and the distribution network communication fault diagnosis system provided by this solution have the same beneficial effects as the corresponding distribution network communication fault diagnosis method, which will not be described in detail.

[0102] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0103] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. A method for diagnosing communication faults in a distribution network, characterized in that: The method comprises: Obtain load and current data of power grid nodes at different times; take any power grid node as a target node, and divide the target node into different time sub-segments according to the load values ​​of the target node at different times within a preset first time period; Determine the current frequency stability coefficient of the target node in each time subsegment according to the quantity and extreme value fluctuation of the current data corresponding to the target node in each time subsegment; Determine the comprehensive stability difference index of the target node in each time subsegment according to the difference in current frequency stability coefficient between the target node and any other power grid node in the same time subsegment, and the distance between the target node and other power grid nodes; and screen out abnormal subsegments of all time subsegments in the target node according to the comprehensive stability difference index; According to the time intervals of different abnormal sub-segments in the target node and the time proportion of the abnormal sub-segments, the distribution density of the abnormal sub-segments is determined, and the adaptive neighborhood reachable distance of the target node is determined according to the distribution density. According to the adaptive neighborhood reachable distance of each power grid node, outlier detection is performed on the current data in the power grid to determine the abnormal current data and mark it as fault data; The step of determining the current frequency stability coefficient of the target node in each time subsegment according to the quantity and extreme value fluctuation of the current data corresponding to the target node in each time subsegment includes: The number of current data in each time sub-segment is normalized to the maximum and minimum values ​​to obtain the sub-segment quantity index; Perform extreme value detection on each time sub-segment to obtain extreme value points, and determine the mean of the time intervals between any extreme value point and the two nearest extreme value points as the extreme value time interval of the extreme value point; Calculate the product of the variance of the current data corresponding to all extreme value points and the variance of the extreme value time interval, and perform maximum and minimum value normalization on the inverse of the product to obtain the extreme value fluctuation index of the time sub-segment; The product of the sub-segment quantity index and the extreme value fluctuation index of the same time sub-segment is taken as the current frequency stability coefficient of the corresponding time sub-segment.

2. A method for diagnosing communication faults in a distribution network as claimed in claim 1, characterized in that: The segmentation is performed according to the load values ​​of the target node at different times, and is divided into different time sub-segments, including: Based on the APCA segmentation method, all moments are segmented according to the load of the target node at different moments to obtain time sub-segments.

3. A method for diagnosing distribution network communication faults as claimed in claim 1, characterized in that: Determining the comprehensive stability difference index of the target node in each time subsegment according to the difference in current frequency stability coefficient between the target node and any other grid node in the same time subsegment, and the distance between the target node and the other grid nodes, includes: Calculate the absolute value of the difference between the current frequency stability coefficient of the target node and any other grid node in the same time subsegment as the stability difference index; The inverse of the distance between the target node and any other grid node is normalized to obtain a distance impact index; The product of the stability difference index and the distance influence index between the target node and the same other power grid nodes is used as the stability difference influence coefficient between the target node and the corresponding other power grid nodes in the same time subsegment; The mean of the stability difference influence coefficients between the target node and all other grid nodes is taken as the comprehensive stability difference index of the target node in the corresponding time subsegment.

4. A method for diagnosing distribution network communication faults as claimed in claim 1, characterized in that: The abnormal sub-segments of all time sub-segments in the target node are screened according to the comprehensive stability difference index, including: The time sub-segment in which the comprehensive stability difference index is greater than the preset index threshold is regarded as an abnormal sub-segment.

5. A method for diagnosing communication faults in a distribution network as claimed in claim 1, characterized in that: According to the time intervals of different abnormal sub-segments in the target node and the time proportions of the abnormal sub-segments, the distribution density of the abnormal sub-segments is determined, including: The middle moment of each abnormal sub-segment is taken as the representative moment of the corresponding abnormal sub-segment, and the time interval between the representative moments of another abnormal sub-segment that is closest to each abnormal sub-segment in time sequence is taken as the abnormal interval of the corresponding abnormal sub-segment; Calculate the mean of the abnormal intervals of all abnormal sub-segments, and normalize the inverse of the mean as the abnormal distribution indicator; The ratio of the number of moments belonging to the abnormal sub-segment to the number of all moments in the preset first time period is used as a moment proportion indicator; The product of the abnormal distribution index and the time proportion index is calculated as the distribution density of the abnormal sub-segment.

6. A method for diagnosing communication faults in a distribution network as claimed in claim 1, characterized in that: Determining an adaptive neighborhood reachable distance of a target node according to the distribution density includes: Perform negative correlation normalization on the distribution density to obtain the density impact weight; The product of the density influence weight and the preset reachable distance is rounded up as the adaptive neighborhood reachable distance.

7. A method for diagnosing communication faults in a distribution network as claimed in claim 1, characterized in that: According to the adaptive neighborhood reachable distance of each grid node, outlier detection is performed on the current data in the grid to determine abnormal current data, including: Based on the LOF algorithm, the adaptive neighborhood reachable distance is used as the K value for outlier detection to determine the abnormal current data.

8. A distribution network communication fault diagnosis device, the device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

9. A distribution network communication fault diagnosis system, characterized in that: The system comprises: An acquisition module is used to acquire load and current data of power grid nodes at different times; taking any power grid node as a target node, and segmenting the target node according to load values ​​at different times within a preset first time period, and dividing the target node into different time sub-segments; A stability analysis module, used to determine the current frequency stability coefficient of the target node in each time sub-segment according to the quantity and extreme value fluctuation of the current data corresponding to the target node in each time sub-segment; The anomaly detection module is used to determine the comprehensive stability difference index of the target node in each time subsegment according to the difference in current frequency stability coefficient between the target node and any other power grid node in the same time subsegment, and the distance between the target node and other power grid nodes; and screen out abnormal subsegments of all time subsegments in the target node according to the comprehensive stability difference index; A fault identification module is used to determine the distribution density of abnormal sub-segments according to the time intervals of different abnormal sub-segments in the target node and the time proportion of the abnormal sub-segments, determine the adaptive neighborhood reachable distance of the target node according to the distribution density, perform outlier detection on the current data in the power grid according to the adaptive neighborhood reachable distance of each power grid node, determine the abnormal current data, and mark it as fault data; The step of determining the current frequency stability coefficient of the target node in each time subsegment according to the quantity and extreme value fluctuation of the current data corresponding to the target node in each time subsegment includes: The number of current data in each time sub-segment is normalized to the maximum and minimum values ​​to obtain the sub-segment quantity index; Perform extreme value detection on each time sub-segment to obtain extreme value points, and determine the mean of the time intervals between any extreme value point and the two nearest extreme value points as the extreme value time interval of the extreme value point; Calculate the product of the variance of the current data corresponding to all extreme value points and the variance of the extreme value time interval, and perform maximum and minimum value normalization on the inverse of the product to obtain the extreme value fluctuation index of the time sub-segment; The product of the sub-segment quantity index and the extreme value fluctuation index of the same time sub-segment is taken as the current frequency stability coefficient of the corresponding time sub-segment.

Images