Online monitoring system for distribution box operation data based on sensors

Abstract

The present invention relates to the field of electrical digital data processing, and in particular to a sensor-based online monitoring system for distribution box operation data, comprising a data processing module, used for screening at least one suspected abnormal point in the acquired distribution box operation data; a data classification module, used for obtaining an abnormal classification index of the suspected abnormal point according to the distribution characteristics between the suspected abnormal point and its adjacent data points for any suspected abnormal point; an abnormality detection module, used for performing adaptive abnormality detection on the suspected abnormal point according to the abnormal classification index of the suspected abnormal point and its adjacent data points, obtaining the abnormal degree of the suspected abnormal point, obtaining an optimal abnormal threshold according to the abnormal degree of each suspected abnormal point, and performing abnormality detection on the real-time distribution box operation data using the optimal abnormal threshold, thereby weakening the influence of noise interference data on the abnormal detection of distribution meter operation data, and further improving the accuracy of abnormal value detection on the distribution box operation data.

Description

Technical Field

[0001] The invention relates to the field of electrical digital data processing, and in particular to a sensor-based online monitoring system for distribution box operation data. Background Art

[0002] The distribution box operation data monitoring system greatly improves the reliability, safety and efficiency of the power system through real-time monitoring, data analysis and automated management, and supports more intelligent operation and maintenance management.

[0003] In the prior art, a threshold is set based on experience of problems with the distribution box operating data. According to the set threshold, if the operating data of the distribution box exceeds the set threshold, the system will trigger an alarm to remind maintenance personnel to pay attention to possible abnormal conditions. The above method of detecting abnormalities in the distribution box operating data by setting empirical thresholds has the advantages of rapid deployment, intuitive understanding, and easy implementation.

[0004] However, since the effectiveness and accuracy of traditional distribution box operation data monitoring methods rely on experience accumulation and reasonable setting of thresholds, this will cause the distribution box operation data collected by sensors during the operation of the distribution box to contain noise data or interference data caused by sensor equipment. These interference data will affect the abnormal value detection of the distribution box operation data, causing false alarms and reducing the reliability of the system.

[0005] Therefore, how to improve the accuracy of outlier detection in distribution box operation data has become an urgent problem to be solved. Summary of the invention

[0006] In view of this, an embodiment of the present invention provides a sensor-based online monitoring system for distribution box operation data to solve the problem of improving the accuracy of abnormal value detection on distribution box operation data.

[0007] An embodiment of the present invention provides a sensor-based online monitoring system for distribution box operation data, which includes the following modules:

[0008] A data processing module is used to obtain a distribution box operation data sequence within a preset period, map each distribution box operation data in the distribution box operation data sequence in a two-dimensional plane, one distribution box operation data corresponds to one data point in the two-dimensional plane, and screen all data points in the two-dimensional plane to obtain at least one suspected abnormal point;

[0009] A data classification module is used to obtain, for any suspected abnormal point, a mutation degree of the suspected abnormal point according to a position difference between the suspected abnormal point and any adjacent data point of the suspected abnormal point, obtain a persistence characteristic value of the suspected abnormal point according to distribution characteristics between adjacent data points of the suspected abnormal point, and obtain an abnormal classification index of the suspected abnormal point according to the mutation degree and the persistence characteristic value;

[0010] The anomaly detection module is used to perform adaptive anomaly detection on the suspected anomaly points according to the anomaly classification index of the suspected anomaly points and the adjacent data points of the suspected anomaly points, obtain the degree of anomaly of the suspected anomaly points, obtain the degree of anomaly of each of the suspected anomaly points, obtain the optimal anomaly threshold according to the degree of anomaly of each of the suspected anomaly points, and use the optimal anomaly threshold to perform anomaly detection on the real-time distribution box operation data.

[0011] Preferably, the data classification module obtains the mutation degree of the suspected outlier point according to the position difference between the suspected outlier point and any adjacent data point of the suspected outlier point, including:

[0012] Respectively obtaining an absolute value of a vertical coordinate difference and an absolute value of a horizontal coordinate difference between the suspected abnormal point and a previous data point of the suspected abnormal point;

[0013] Obtaining a ratio of the absolute values ​​of the ordinate difference to the absolute values ​​of the abscissa difference, using the absolute value ratio of the difference as an independent variable of an inverse tangent function, obtaining a corresponding inverse tangent function value, obtaining a first normalized result of the inverse tangent function value, and obtaining a second normalized result of the absolute values ​​of the ordinate difference;

[0014] A weighted sum is performed on the first normalization result and the second normalization result, and the corresponding weighted sum result is used as the mutation degree of the suspected abnormal point.

[0015] Preferably, the data classification module obtains the persistent characteristic value of the suspected abnormal point according to the distribution characteristics between the adjacent data points of the suspected abnormal point, including:

[0016] The suspected abnormal point and a preset number of data points after the suspected abnormal point form a first data sequence, calculate the ordinate difference and the abscissa difference between any two adjacent data points in the first data sequence, obtain the absolute value of the difference ratio between the ordinate difference and the abscissa difference, and obtain the mean value of the ratio of the first data sequence according to the absolute value of the difference ratio between every two adjacent data points in the first data sequence;

[0017] Calculate the absolute value of the ratio of the difference between the ordinate and the abscissa between the first data point and the last data point in the two-dimensional plane;

[0018] The absolute value of the difference between the mean value of the ratio and the absolute value of the difference ratio is obtained, the absolute value of the difference is normalized, and the corresponding normalized result is used as the persistence characteristic value of the suspected abnormal point.

[0019] Preferably, the data classification module obtains the abnormal classification index of the suspected abnormal point according to the mutation degree and the persistence characteristic value, including:

[0020] The mutation degree and the persistence characteristic value are weighted and summed, and the corresponding weighted summation result is used as the abnormal classification index of the suspected abnormal point.

[0021] Preferably, the anomaly detection module is used to perform adaptive anomaly detection on the suspected anomaly point according to the anomaly classification index of the suspected anomaly point and the adjacent data points of the suspected anomaly point to obtain the degree of anomaly of the suspected anomaly point, including:

[0022] Obtain a comparison result between the abnormal classification index of the suspected abnormal point and a preset abnormal classification threshold, determine the data type of the suspected abnormal point according to the comparison result, and obtain the abnormal degree of the suspected abnormal point according to the data type and the adjacent data points of the suspected abnormal point.

[0023] Preferably, the abnormality detection module determines the data type of the suspected abnormal point according to the comparison result, including:

[0024] If the comparison result is that the abnormality classification index of the suspected abnormal point is greater than or equal to the abnormality classification threshold, the suspected abnormal point is marked as a high abnormality point;

[0025] If the comparison result is that the abnormality classification index of the suspected abnormal point is less than the abnormality classification threshold, the suspected abnormal point is marked as a low abnormality point.

[0026] Preferably, the abnormality detection module obtains the abnormality degree of the suspected abnormal point according to the data type and the adjacent data points of the suspected abnormal point, including:

[0027] If the data type of the suspected outlier point is a high degree of outlier point, the KNN algorithm is used to obtain K nearest neighboring points of the suspected outlier point in the two-dimensional plane, and the maximum distance is obtained as the degree of outlier of the suspected outlier point according to the distance between the suspected outlier point and each of the nearest neighboring points;

[0028] If the data type of the suspected outlier point is a point of low degree of abnormality, a preset number of data points adjacent to the left and right sides of the suspected outlier point are obtained, and a second data sequence is formed with the suspected outlier point. The distance between each two adjacent data points in the second data sequence is obtained respectively to obtain the mean distance, and the mean distance is used as the degree of abnormality of the suspected outlier point.

[0029] Preferably, the abnormality detection module obtains the optimal abnormality threshold according to the abnormality degree of each suspected abnormal point, including:

[0030] According to the abnormality degree of all suspected abnormal points, the average abnormality degree is obtained. For any suspected abnormal point, if the abnormality degree of the suspected abnormal point is greater than the average abnormality degree, the suspected abnormal point is marked as a real abnormal point, and the real abnormal point among all suspected abnormal points is obtained;

[0031] According to the abnormality degree of each real abnormal point, the minimum abnormality degree is selected as the optimal abnormality threshold.

[0032] Preferably, the data processing module selects at least one suspected abnormal point from all data points in the two-dimensional plane, including:

[0033] An abnormal operation data threshold is obtained. If the data value of any data point in the two-dimensional plane is greater than the abnormal operation data threshold, the data point is marked as a suspected abnormal point, and all suspected abnormal points in the two-dimensional plane are obtained.

[0034] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0035] The present invention provides a sensor-based online monitoring system for distribution box operation data, comprising a data processing module, which is used to obtain a distribution box operation data sequence within a preset period, map each distribution box operation data in the distribution box operation data sequence in a two-dimensional plane, one distribution box operation data corresponds to a data point in the two-dimensional plane, and at least one suspected abnormal point is screened from all the data points in the two-dimensional plane; a data classification module, which is used to obtain the mutation degree of any suspected abnormal point according to the position difference between the suspected abnormal point and any adjacent data point of the suspected abnormal point, obtain the persistent characteristic value of the suspected abnormal point according to the distribution characteristics between the adjacent data points of the suspected abnormal point, and obtain the abnormal classification index of the suspected abnormal point according to the mutation degree and the persistent characteristic value; an abnormality detection module, which is used to perform adaptive abnormality detection on the suspected abnormal point according to the abnormal classification index of the suspected abnormal point and the adjacent data points of the suspected abnormal point, obtain the abnormality degree of the suspected abnormal point, obtain the abnormality degree of each of the suspected abnormal points, obtain the optimal abnormal threshold according to the abnormality degree of each of the suspected abnormal points, and use the optimal abnormal threshold to perform abnormality detection on real-time distribution box operation data. Among them, all suspected abnormal points are screened out from the distribution box operation data according to the empirical threshold. The suspected abnormal points include noise interference data and real abnormal data caused by equipment failure. Since the noise interference data will affect the detection of the real abnormal data, the classification abnormality index of the suspected abnormal point is obtained according to the mutation degree and continuous characteristic value of the suspected abnormal point, and the classification abnormality index is used to perform adaptive abnormality detection on the suspected abnormal point to obtain the abnormality degree of the suspected abnormal point. After that, the optimal abnormal threshold is obtained according to the abnormality degree, and the optimal abnormal threshold is used to perform abnormality detection on the real-time distribution box operation data, thereby weakening the influence of noise interference data on the abnormality detection of the distribution meter operation data, thereby improving the accuracy of abnormal value detection of the distribution box operation data. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments or the description of the prior art will be 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 labor.

[0037] Figure 1 It is a structural block diagram of a sensor-based online monitoring system for distribution box operation data provided in Embodiment 1 of the present invention;

[0038] Figure 2 It is a schematic diagram of obtaining the abnormality degree of a low abnormality degree point provided by the first embodiment of the present invention. DETAILED DESCRIPTION

[0039] Embodiments of the present disclosure are described in detail below, and examples of the embodiments are shown in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present disclosure, but should not be understood as limiting the present disclosure.

[0040] It should be noted that the terms "first", "second", etc. in the specification of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the present disclosure.

[0041] In order to illustrate the technical solution of the present invention, specific embodiments are provided below for illustration.

[0042] See also Figure 1 , is a structural block diagram of a sensor-based online monitoring system for distribution box operation data provided in Embodiment 1 of the present invention, such as Figure 1 As shown, the system may include:

[0043] The data processing module 11 is used to obtain a distribution box operation data sequence within a preset period, map each distribution box operation data in the distribution box operation data sequence in a two-dimensional plane, and one distribution box operation data corresponds to one data point in the two-dimensional plane, and screen out at least one suspected abnormal point from all data points in the two-dimensional plane.

[0044] First, obtain the distribution box operation data sequence within a preset period. The specific method is: the period set in the embodiment of the present invention is 1 day. There is no restriction on the preset period and it can be set according to the implementation scenario. The distribution box operation data sequence including the distribution box operation current data collected within one day is obtained. To facilitate the subsequent processing of the distribution box operation data in the distribution box operation data sequence, each distribution box operation data in the distribution box operation data sequence is mapped to a two-dimensional plane, wherein the two-dimensional plane is constructed with time t as the horizontal axis and the distribution box operation data value a as the vertical axis, and one data point in the two-dimensional plane corresponds to one distribution box operation data.

[0045] In the prior art, abnormal data in the distribution box operation data is obtained by using an abnormal operation data threshold. Since the accuracy of this technology depends on the accumulation of experience and the reasonable setting of the threshold, the abnormal data obtained by the traditional distribution box operation data monitoring method may contain noise interference data and real abnormal data caused by equipment failure, and the noise interference data will affect the detection of real abnormal data. In order to facilitate the subsequent distinction of whether the collected abnormal data is real abnormal data, in an embodiment of the present invention, after obtaining data points corresponding to all distribution box operation data in the distribution box operation data sequence in a two-dimensional plane, at least one suspected abnormal point is screened out from all data points. The specific method includes:

[0046] Obtain an abnormal operation data threshold. If the distribution box operation data is greater than the abnormal operation data threshold, mark the distribution box operation data as a suspected abnormal point, and obtain at least one suspected abnormal point in all the distribution box operation data in the two-dimensional plane.

[0047] In one embodiment, an abnormal operation data threshold is obtained based on a box plot and a 3sigma anomaly detection method. If the data point value corresponding to the operation data of any distribution box is greater than the abnormal operation data threshold, the data point is marked as a suspected abnormal point, thereby obtaining at least one suspected abnormal point in a two-dimensional plane.

[0048] It should be noted that the abnormal operation data threshold set in the embodiment of the present invention is an empirical value and can be set according to the specific implementation scenario. Among them, the box plot and 3sigma anomaly detection method belong to the existing technology and will not be repeated here.

[0049] At this point, at least one suspected abnormal point is screened out from all the data points in the two-dimensional plane.

[0050] The data classification module 12 is used to obtain, for any suspected outlier point, the degree of mutation of the suspected outlier point based on the position difference between the suspected outlier point and any adjacent data point of the suspected outlier point, obtain the persistence characteristic value of the suspected outlier point based on the distribution characteristics between the adjacent data points of the suspected outlier point, and obtain the abnormal classification index of the suspected outlier point based on the degree of mutation and the persistence characteristic value.

[0051] Since the acquired suspected abnormal points include noise interference data and real abnormal data caused by equipment failure, the data characteristics of real abnormal data are usually manifested as data mutation or drastic fluctuation. These real abnormal data will exceed the normal range and will last for a period of time, thereby affecting the trend of the overall distribution box operation data, while the data characteristics of noise interference data are usually manifested as random, irregular small-amplitude offsets, which will not cause drastic changes in the data and will not have an obvious trend of change. Therefore, in an embodiment of the present invention, for any suspected abnormal point in a two-dimensional plane, first, according to the position difference between the suspected abnormal point and any adjacent data point of the suspected abnormal point, the degree of mutation of the suspected abnormal point is obtained, and then, according to the distribution characteristics between the adjacent data points of the suspected abnormal point, the persistence characteristic value of the suspected abnormal point is obtained, and finally, according to the degree of mutation and the persistence characteristic value, the abnormal classification index of the suspected abnormal point is obtained.

[0052] Wherein, the mutation degree of the suspected outlier point is obtained according to the position difference between the suspected outlier point and any adjacent data point of the suspected outlier point, including:

[0053] Respectively obtaining an absolute value of a vertical coordinate difference and an absolute value of a horizontal coordinate difference between the suspected abnormal point and a previous data point of the suspected abnormal point;

[0054] Obtaining a ratio of the absolute values ​​of the ordinate difference to the absolute values ​​of the abscissa difference, using the absolute value ratio of the difference as an independent variable of an inverse tangent function, obtaining a corresponding inverse tangent function value, obtaining a first normalized result of the inverse tangent function value, and obtaining a second normalized result of the absolute values ​​of the ordinate difference;

[0055] A weighted sum is performed on the first normalization result and the second normalization result, and the corresponding weighted sum result is used as the mutation degree of the suspected abnormal point.

[0056] In one implementation, taking the ith suspected abnormal point as an example, the coordinates of the ith suspected abnormal point are recorded as ( , ), the coordinates of the previous data point of the i-th suspected outlier point are ( , ), respectively obtain the absolute value of the ordinate difference and the absolute value of the abscissa difference between the ith suspected outlier point and the data point before the ith suspected outlier point, use the absolute value ratio of the ordinate difference and the abscissa difference as the independent variable of the inverse tangent function, and obtain the corresponding inverse tangent function value. According to the absolute value of the ordinate difference and the inverse tangent function value, the mutation degree of the ith suspected outlier point is obtained. The expression of the mutation degree of the ith suspected outlier point is:

[0057]

[0058] in, Indicates the mutation degree of the i-th suspected abnormal point, represents the coordinates of the i-th suspected outlier point, Represents the coordinates of the previous data point of the i-th suspected outlier point, () represents the inverse tangent function, Norm() represents the normalization function, || represents the absolute value symbol, represents the first weight, Represents the second weight.

[0059] It should be noted that It reflects the difference between the ordinate value of the i-th suspected outlier point and its previous data point. The larger the value is, the greater the numerical difference between the ith suspected abnormal point and its previous data point is, indicating that the fluctuation degree of the ith suspected abnormal point is greater, and the more likely it is to be real abnormal data caused by equipment failure, and the greater the mutation degree of the corresponding ith suspected abnormal point is.

[0060] Preferably, the first weight and the second weight set in the embodiment of the present invention are empirical values, that is, There is no restriction on this, and it can be set according to the implementation scenario.

[0061] Wherein, according to the distribution characteristics between the adjacent data points of the suspected abnormal point, the persistent characteristic value of the suspected abnormal point is obtained, including:

[0062] The suspected abnormal point and a preset number of data points after the suspected abnormal point form a first data sequence, calculate the ordinate difference and the abscissa difference between any two adjacent data points in the first data sequence, obtain the absolute value of the difference ratio between the ordinate difference and the abscissa difference, and obtain the mean value of the ratio of the first data sequence according to the absolute value of the difference ratio between every two adjacent data points in the first data sequence;

[0063] Calculate the absolute value of the ratio of the difference between the ordinate and the abscissa between the first data point and the last data point in the two-dimensional plane;

[0064] The absolute value of the difference between the mean value of the ratio and the absolute value of the difference ratio is obtained, the absolute value of the difference is normalized, and the corresponding normalized result is used as the persistence characteristic value of the suspected abnormal point.

[0065] In one implementation, taking the ith suspected abnormal point as an example, the preset number set in the embodiment of the present invention is 2, and there is no limit on the preset number, which can be set according to the implementation scenario, and the two adjacent data points after the ith suspected abnormal point are obtained, and the first data sequence is formed with the ith suspected abnormal point, and the ordinate difference and the abscissa difference between any two adjacent data points in the first data sequence are calculated, and the absolute value of the difference ratio between the ordinate difference and the abscissa difference is obtained. According to the absolute value of the difference ratio of every two adjacent data points in the first data sequence, the mean of the ratios of the first data sequence is obtained, and the absolute value of the difference ratio of the ordinate difference and the abscissa difference between the first data point and the last data point in the two-dimensional plane is calculated. According to the absolute value of the difference between the mean of the ratios and the absolute value of the difference ratio, the persistence characteristic value of the ith suspected abnormal point is obtained, and the expression of the persistence characteristic value of the ith suspected abnormal point is:

[0066]

[0067] in, represents the persistent eigenvalue of the i-th suspected outlier point, represents the coordinates of the i-th suspected outlier point, represents the coordinates of the next data point after the ith suspected outlier point, and n represents the number of data points after the ith suspected outlier point. represents the coordinates of the first data point in the two-dimensional plane, represents the coordinates of the last data point in the two-dimensional plane, Norm() represents the normalization function, and || represents the absolute value sign.

[0068] It should be noted that reflects the overall slope of all data points in the two-dimensional plane, It reflects the average slope between the i-th suspected outlier point and its n adjacent data points. The larger the value of is, the larger the average slope of the i-th suspected abnormal point is, corresponding to The larger the difference is, the greater the difference between the average slope of the ith suspected abnormal point and the overall slope is, that is, the greater the possibility that the ith suspected abnormal point is the real abnormal data caused by equipment failure, and the greater the abnormal persistence of the ith suspected abnormal point.

[0069] Among them, according to the mutation degree and persistence characteristic value, the abnormal classification indicators of suspected abnormal points are obtained, including:

[0070] The mutation degree and the persistence characteristic value are weighted and summed, and the corresponding weighted summation result is used as the abnormal classification index of the suspected abnormal point.

[0071] In one implementation, taking the ith suspected abnormal point as an example, a weighted sum is performed on the mutation degree of the ith suspected abnormal point and the persistence characteristic value of the ith suspected abnormal point to obtain the abnormal classification index of the ith suspected abnormal point. Then, the expression of the abnormal classification index of the ith suspected abnormal point is:

[0072]

[0073] in, represents the abnormal classification index of the i-th suspected abnormal point, Indicates the mutation degree of the i-th suspected abnormal point, represents the persistent eigenvalue of the i-th suspected outlier point, represents the weight of the mutation degree, Represents the weight of the persistent eigenvalue.

[0074] It should be noted that The larger the value of , the greater the fluctuation degree of the i-th suspected abnormal point, that is, the greater the mutation degree of the i-th suspected abnormal point, the greater the abnormal degree of the corresponding i-th suspected abnormal point, that is, the greater the possibility that the i-th suspected abnormal point is a high abnormal degree point; The larger the value of , the greater the abnormal persistence of the i-th suspected abnormal point, and the greater the abnormal degree of the corresponding i-th suspected abnormal point, that is, the greater the possibility that the i-th suspected abnormal point is a high abnormal degree point.

[0075] Preferably, the weight of the mutation degree and the weight of the persistence characteristic value set in the embodiment of the present invention are empirical values, that is, There is no restriction on this, and it can be set according to the implementation scenario.

[0076] At this point, the abnormal classification index of the i-th suspected abnormal point is obtained according to the distribution characteristics between the i-th suspected abnormal point and the data points adjacent to the i-th suspected abnormal point.

[0077] Similarly, according to the above step of obtaining the abnormal classification index of the i-th suspected abnormal point, the abnormal classification index of each suspected abnormal point in the two-dimensional plane is obtained.

[0078] The anomaly detection module 13 is used to perform adaptive anomaly detection on the suspected anomaly points according to the anomaly classification index of the suspected anomaly points and the adjacent data points of the suspected anomaly points, obtain the degree of anomaly of the suspected anomaly points, obtain the degree of anomaly of each suspected anomaly point, obtain the optimal anomaly threshold according to the degree of anomaly of each suspected anomaly point, and use the optimal anomaly threshold to perform anomaly detection on the real-time distribution box operation data.

[0079] In order to highlight the abnormal characteristics of data points with a high degree of abnormality and reduce the abnormal characteristics of points with a low degree of abnormality, adaptive abnormality detection is performed on points with different degrees of abnormality, thereby weakening the influence of noise interference on the abnormal detection of distribution box operation data. Therefore, in an embodiment of the present invention, for any suspected abnormal point in a two-dimensional plane, first, according to the abnormal classification index of the suspected abnormal point and the adjacent data points of the suspected abnormal point, an adaptive abnormality detection is performed on the suspected abnormal point to obtain the abnormal degree of the suspected abnormal point. In order to more accurately distinguish whether each suspected abnormal point in the two-dimensional plane is a real abnormal point, according to the abnormal degree of each suspected abnormal point, an optimal abnormal threshold is obtained. When the optimal abnormal threshold is used to perform abnormal detection on the real-time distribution box operation data, the real abnormal points can be effectively distinguished, thereby improving the detection capability of the abnormal detection system for the abnormal state of the distribution box, and ultimately ensuring the safety and reliability of the power system.

[0080] Among them, according to the abnormal classification index of the suspected abnormal point and the adjacent data points of the suspected abnormal point, the suspected abnormal point is adaptively detected to obtain the abnormal degree of the suspected abnormal point, including:

[0081] (1) Obtaining a comparison result between the abnormality classification index of the suspected abnormal point and a preset abnormality classification threshold, and determining the data type of the suspected abnormal point according to the comparison result.

[0082] Specifically, if the comparison result is that the abnormality classification index of the suspected abnormal point is greater than or equal to the abnormality classification threshold, the suspected abnormal point is marked as a high abnormality point;

[0083] If the comparison result is that the abnormality classification index of the suspected abnormal point is less than the abnormality classification threshold, the suspected abnormal point is marked as a low abnormality point.

[0084] In one implementation, taking the i-th suspected abnormal point as an example, the abnormal classification threshold set in the embodiment of the present invention is 0.7. There is no restriction on the preset abnormal classification threshold and it can be set according to the implementation scenario. If the abnormal classification index of the i-th suspected abnormal point is greater than or equal to 0.7, the i-th suspected abnormal point is marked as a high abnormality point.

[0085] If the abnormal classification index of the i-th suspected abnormal point is less than 0.7, the i-th suspected abnormal point is marked as a low abnormal degree point.

[0086] (2) Obtaining the degree of abnormality of the suspected abnormal point based on the data type and the adjacent data points of the suspected abnormal point.

[0087] If the data type of the suspected outlier point is a high degree of outlier point, the KNN algorithm is used to obtain K nearest neighboring points of the suspected outlier point in the two-dimensional plane, and the maximum distance is obtained as the degree of outlier of the suspected outlier point according to the distance between the suspected outlier point and each of the nearest neighboring points;

[0088] If the data type of the suspected outlier point is a point of low degree of abnormality, a preset number of data points adjacent to the left and right sides of the suspected outlier point are obtained, and a second data sequence is formed with the suspected outlier point. The distance between each two adjacent data points in the second data sequence is obtained respectively to obtain the mean distance, and the mean distance is used as the degree of abnormality of the suspected outlier point.

[0089] In one implementation, taking the ith suspected outlier point as an example, if the data type of the ith suspected outlier point is a high degree of outlier point, the K value is set to 4, and the KNN algorithm is used to obtain the distance between the ith suspected outlier point and all other data points, and the maximum distance value among the K minimum distances is selected as the degree of outlier of the ith suspected outlier point, thereby highlighting the abnormal characteristics of the ith suspected outlier point, wherein the KNN algorithm is a prior art and will not be described in detail here.

[0090] It should be noted that the K value set in the embodiment of the present invention is an empirical value and can be set according to the implementation scenario.

[0091] If the data type of the i-th suspected abnormal point is a low abnormal degree point, refer to Figure 2 , which is a schematic diagram of obtaining the abnormality degree of a low abnormality point, where Figure 2 The two-dimensional plane in the figure has time t as the horizontal axis and the distribution box operation data value a as the vertical axis. The i-th suspected abnormal point corresponds to Figure 2 The peak point H in the embodiment of the present invention is set to a preset number of 4. There is no limit on the preset number and it can be set according to the implementation scenario. A total of 4 adjacent data points on the left and right sides of the i-th suspected abnormal point are obtained, that is, 2 adjacent data points on the left and right sides of the i-th suspected abnormal point are obtained, and a second data sequence is formed with the i-th suspected abnormal point. The distance between each two adjacent data points in the second data sequence (i.e. Figure 2 The distances of L1, L2, L3 and L4 corresponding to the i-th suspected outlier are obtained, and the obtained distance mean is used as the abnormality degree of the i-th suspected outlier point, thereby reducing the abnormality degree of the i-th suspected outlier point and weakening the influence of the i-th suspected outlier point on the judgment of the real outlier point.

[0092] Similarly, according to the above step of obtaining the abnormality degree of the i-th suspected abnormal point, the abnormality degree of each suspected abnormal point in the two-dimensional plane is obtained.

[0093] Among them, according to the abnormality degree of each suspected abnormal point, the optimal abnormal threshold is obtained, including:

[0094] According to the abnormality degree of all suspected abnormal points, the average abnormality degree is obtained. For any suspected abnormal point, if the abnormality degree of the suspected abnormal point is greater than the average abnormality degree, the suspected abnormal point is marked as a real abnormal point, and the real abnormal point among all suspected abnormal points is obtained;

[0095] According to the abnormality degree of each real abnormal point, the minimum abnormality degree is selected as the optimal abnormality threshold.

[0096] In one implementation, based on the abnormality degree of each suspected abnormal point, the average abnormality degree is obtained. Taking the i-th suspected abnormal point as an example, if the abnormality degree of the i-th suspected abnormal point is greater than the average abnormality degree, the i-th suspected abnormal point is marked as a true abnormal point.

[0097] Similarly, according to the above steps of obtaining the real outliers, all the real outliers are screened out from all the suspected outliers in the two-dimensional plane.

[0098] All the obtained real abnormal points are sorted in descending order according to the degree of abnormality, and the smallest abnormality degree is selected as the optimal abnormality threshold. The optimal abnormality threshold is used to perform anomaly detection on the real-time distribution box operation data.

[0099] At this point, anomaly detection is performed on the real-time distribution box operation data based on the obtained optimal anomaly threshold.

[0100] In summary, the present invention provides a sensor-based online monitoring system for distribution box operation data, including a data processing module, which is used to obtain a distribution box operation data sequence within a preset period, map each distribution box operation data in the distribution box operation data sequence in a two-dimensional plane, one distribution box operation data corresponds to a data point in the two-dimensional plane, and at least one suspected abnormal point is screened from all data points in the two-dimensional plane; a data classification module, which is used to obtain the mutation degree of any suspected abnormal point according to the position difference between the suspected abnormal point and any adjacent data point of the suspected abnormal point, obtain the persistent characteristic value of the suspected abnormal point according to the distribution characteristics between the adjacent data points of the suspected abnormal point, and obtain the abnormal classification index of the suspected abnormal point according to the mutation degree and the persistent characteristic value; an abnormality detection module, which is used to perform adaptive abnormality detection on the suspected abnormal point according to the abnormal classification index of the suspected abnormal point and the adjacent data points of the suspected abnormal point, obtain the abnormality degree of the suspected abnormal point, obtain the abnormality degree of each of the suspected abnormal points, obtain the optimal abnormal threshold according to the abnormality degree of each of the suspected abnormal points, and use the optimal abnormal threshold to perform abnormality detection on real-time distribution box operation data. Among them, all suspected abnormal points are screened out from the distribution box operation data according to the empirical threshold. The suspected abnormal points include noise interference data and real abnormal data caused by equipment failure. Since the noise interference data will affect the detection of the real abnormal data, the classification abnormality index of the suspected abnormal point is obtained according to the mutation degree and continuous characteristic value of the suspected abnormal point, and the classification abnormality index is used to perform adaptive abnormality detection on the suspected abnormal point to obtain the abnormality degree of the suspected abnormal point. After that, the optimal abnormal threshold is obtained according to the abnormality degree, and the optimal abnormal threshold is used to perform abnormality detection on the real-time distribution box operation data, thereby weakening the influence of noise interference data on the abnormality detection of the distribution meter operation data, thereby improving the accuracy of abnormal value detection of the distribution box operation data.

[0101] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention.

Claims

1. The sensor-based online monitoring system for distribution box operation data is characterized by: The system comprises: A data processing module is used to obtain a distribution box operation data sequence within a preset period, and map each distribution box operation data in the distribution box operation data sequence in a two-dimensional plane, wherein the two-dimensional plane is constructed with time t as the horizontal axis and the distribution box operation data value a as the vertical axis, and one distribution box operation data corresponds to one data point in the two-dimensional plane, and at least one suspected abnormal point is screened out from all the data points in the two-dimensional plane; A data classification module is used for obtaining, for any suspected outlier point, a mutation degree of the suspected outlier point according to a position difference between the suspected outlier point and any adjacent data point of the suspected outlier point, the mutation degree refers to a numerical difference between the suspected outlier point and a previous data point, obtaining a persistence characteristic value of the suspected outlier point according to distribution characteristics between adjacent data points of the suspected outlier point, the persistence characteristic value refers to a difference between an average slope of the suspected outlier point and an overall slope of all data points in the two-dimensional plane, the average slope refers to an average of the slopes between the suspected outlier point and its n adjacent data points, and obtaining an abnormal classification index of the suspected outlier point according to the mutation degree and the persistence characteristic value; An anomaly detection module, used to perform adaptive anomaly detection on the suspected anomaly point according to the anomaly classification index of the suspected anomaly point and the adjacent data points of the suspected anomaly point, obtain the degree of anomaly of the suspected anomaly point, obtain the degree of anomaly of each of the suspected anomaly points, obtain the optimal anomaly threshold according to the degree of anomaly of each of the suspected anomaly points, and perform anomaly detection by comparing the optimal anomaly threshold with the real-time distribution box operation data; According to the abnormal classification index of the suspected abnormal point and the adjacent data points of the suspected abnormal point, adaptive abnormality detection is performed on the suspected abnormal point to obtain the abnormality degree of the suspected abnormal point, including: Obtaining a comparison result between the abnormal classification index of the suspected abnormal point and a preset abnormal classification threshold, determining the data type of the suspected abnormal point according to the comparison result, and obtaining the abnormal degree of the suspected abnormal point according to the data type and the adjacent data points of the suspected abnormal point; Determining the data type of the suspected abnormal point according to the comparison result includes: If the comparison result is that the abnormality classification index of the suspected abnormal point is greater than or equal to the abnormality classification threshold, the suspected abnormal point is marked as a high abnormality point; If the comparison result is that the abnormality classification index of the suspected abnormal point is less than the abnormality classification threshold, the suspected abnormal point is marked as a low abnormality point; Acquiring the abnormality degree of the suspected abnormal point according to the data type and the adjacent data points of the suspected abnormal point, including: If the data type of the suspected outlier point is a high degree of outlier point, the KNN algorithm is used to obtain K nearest neighboring points of the suspected outlier point in the two-dimensional plane, and the maximum distance is obtained as the degree of outlier of the suspected outlier point according to the distance between the suspected outlier point and each of the nearest neighboring points; If the data type of the suspected outlier point is a point of low degree of abnormality, a preset number of data points adjacent to the left and right sides of the suspected outlier point are obtained, and a second data sequence is formed with the suspected outlier point. The distance between each two adjacent data points in the second data sequence is obtained respectively to obtain the mean distance, and the mean distance is used as the degree of abnormality of the suspected outlier point.

2. The sensor-based online monitoring system for distribution box operation data according to claim 1 is characterized in that: The data classification module obtains the mutation degree of the suspected abnormal point according to the position difference between the suspected abnormal point and any adjacent data point of the suspected abnormal point, including: Respectively obtaining an absolute value of a vertical coordinate difference and an absolute value of a horizontal coordinate difference between the suspected abnormal point and a previous data point of the suspected abnormal point; Obtaining a ratio between the absolute value of the ordinate difference and the absolute value of the abscissa difference as a difference absolute value ratio, using the difference absolute value ratio as an independent variable of an inverse tangent function, obtaining a corresponding inverse tangent function value, obtaining a first normalized result of the inverse tangent function value, and obtaining a second normalized result of the ordinate difference absolute value; A weighted sum is performed on the first normalization result and the second normalization result, and the corresponding weighted sum result is used as the mutation degree of the suspected abnormal point.

3. The sensor-based online monitoring system for distribution box operation data according to claim 1 is characterized in that: The data classification module obtains the persistent feature value of the suspected abnormal point according to the distribution characteristics between the adjacent data points of the suspected abnormal point, including: The suspected abnormal point and a preset number of data points after the suspected abnormal point form a first data sequence, the ordinate difference and the abscissa difference between any two adjacent data points in the first data sequence are calculated, and the absolute value of the ratio between the ordinate difference and the abscissa difference is obtained as a first absolute value, and the mean of the ratio of the first data sequence is obtained according to the first absolute value of every two adjacent data points in the first data sequence; Calculate the absolute value of the ratio between the longitudinal coordinate difference and the abscissa difference between the first data point and the last data point in the two-dimensional plane as the second absolute value; The absolute value of the difference between the ratio mean and the second absolute value is obtained, the absolute value of the difference is normalized, and the corresponding normalized result is used as the persistence characteristic value of the suspected abnormal point.

4. The sensor-based online monitoring system for distribution box operation data according to claim 1 is characterized in that: The data classification module obtains the abnormal classification index of the suspected abnormal point according to the mutation degree and the persistence characteristic value, including: The mutation degree and the persistence characteristic value are weighted and summed, and the corresponding weighted summation result is used as the abnormal classification index of the suspected abnormal point.

5. The sensor-based online monitoring system for distribution box operation data according to claim 1 is characterized in that: The abnormality detection module obtains the optimal abnormality threshold according to the abnormality degree of each suspected abnormal point, including: According to the abnormality degree of all suspected abnormal points, the average abnormality degree is obtained. For any suspected abnormal point, if the abnormality degree of the suspected abnormal point is greater than the average abnormality degree, the suspected abnormal point is marked as a real abnormal point, and the real abnormal point among all suspected abnormal points is obtained; According to the abnormality degree of each real abnormal point, the minimum abnormality degree is selected as the optimal abnormality threshold.

6. The sensor-based online monitoring system for distribution box operation data according to claim 1 is characterized in that: The data processing module selects at least one suspected abnormal point from all data points in the two-dimensional plane, including: An abnormal operation data threshold is obtained. If the data value of any data point in the two-dimensional plane is greater than the abnormal operation data threshold, the data point is marked as a suspected abnormal point, and all suspected abnormal points in the two-dimensional plane are obtained.

Images