An intelligent detection method for the operating state of electrical equipment

Abstract

The present application provides an intelligent detection method for the operating state of electrical equipment, and this method is applied to the field of data detection. The method includes: obtaining monitoring data of a hybrid gas circuit breaker in multiple dimensions; determining the degree of change coordination of the monitoring data in each dimension based on the monitoring data in multiple dimensions; determining the likelihood of abnormal performance of each data point in the monitoring data based on the degree of change coordination and the occurrence probability of the monitoring data; determining abnormal data points in the monitoring data based on the likelihood of abnormal performance; and judging the operating state of the hybrid gas circuit breaker based on the abnormal data points. In this method, it is possible to reduce the redundancy of the algorithm, improve the accuracy of abnormal data identification, and also improve the accuracy of detecting the operating state of the hybrid gas circuit breaker.

Description

Technical Field

[0001] This application relates to the field of data detection, and more specifically, to an intelligent detection method for the operating state of electrical equipment. Background Art

[0002] Common electrical equipment includes generators, transformers, power lines, circuit breakers, etc., which enable the normal operation of the power system and the normal transmission of electricity in the power system. According to different functions, structures, production materials, etc., the usage environment requirements of different electrical equipment are different. The usage environment includes low-pressure environment, low-temperature environment, dedicated power supply, etc. For example, low-temperature resistant cables, hybrid gas circuit breakers, etc. can be used in low-pressure environments and low-temperature environments due to materials or protective sleeves.

[0003] During the operation of electrical equipment, monitoring software is usually used to monitor and display environmental parameters. When the parameters are abnormal, audible and visual alarm information is issued to prompt the source of the abnormal information, and hierarchical protection measures can be taken when necessary.

[0004] Based on the Markov clustering algorithm in the process of identifying abnormal data, since the Markov algorithm itself needs to perform iterative operations to stretch the differences between different data points, but the transition probabilities between some abnormal data and normal data are small, some overlapping clusters will be lost after iteration, which will interfere with abnormal identification. At the same time, the abnormal changes of data are random, and when any dimension of data has abnormal changes, it is very likely to drive other dimensions of data, including the operating state of the electrical equipment itself, to change. Therefore, it is impossible to accurately judge the transfer relationship between each data and accurately judge the operating state of the electrical equipment only according to the frequency. Summary of the Invention

[0005] This application provides an intelligent detection method for the operating state of electrical equipment. By analyzing the abnormal possibility of each data point, this method can quickly determine the abnormal data point, and then judge the operating state of the hybrid gas circuit breaker through this abnormal data point. In this way, it can not only reduce the redundancy of the algorithm, improve the accuracy of abnormal data recognition, but also improve the accuracy of judging the operating state of the hybrid gas circuit breaker.

[0006] An embodiment of this application provides an intelligent detection method for the operating state of electrical equipment, which includes:

[0007] Obtain monitoring data of multiple dimensions of the circuit breaker through sensors;

[0008] For the first dimension among the multiple dimensions, construct a data window that matches the target data point at any moment in the first dimension; wherein, the first dimension is any one of the multiple dimensions; based on the first data point set of the monitoring data in the first dimension and the second data point sets respectively corresponding to the monitoring data in each second dimension, determine the correlation of co-variation generated by the first dimension; wherein, the second dimension is any one of the multiple dimensions and the state indication data, and the second dimension is different from the first dimension; based on the target data point and the first data point set, determine the possibility coefficient that the target data point is an outlier; adjust the correlation of co-variation generated by the first dimension using the possibility coefficient to obtain the degree of co-variation of the first dimension;

[0009] Based on the degree of co-variation and the occurrence probability of the monitoring data, determine the possibility of abnormal performance of each data point in the monitoring data;

[0010] Based on the possibility of abnormal performance, determine the abnormal data points in the monitoring data;

[0011] Based on the abnormal data points, detect the operating state of the circuit breaker.

[0012] Preferably, the determining the correlation of co-variation generated by the first dimension based on the first data point set of the monitoring data in the first dimension and the second data point sets respectively corresponding to the monitoring data in each second dimension includes:

[0013] Calculate the covariance between the first data point set and the second data point sets;

[0014] Take the sum of the distribution variances of the first data point set and the second data point sets as the denominator;

[0015] Take the cumulative result of the ratio of the covariance to the denominator over all second dimensions as the correlation of co-variation generated by the first dimension.

[0016] Preferably, the determining the possibility coefficient that the target data point is an outlier based on the target data point and the first data point set includes:

[0017] Take the ratio of the mean value of the data window where the target data point is located to the maximum data value of the first data point set as the possibility coefficient.

[0018] Preferably, the determining the possibility of abnormal performance of each data point in the monitoring data based on the degree of co-variation and the occurrence probability of the monitoring data includes:

[0019] Determine the transition probability between any two data points of the same dimension in the monitoring data based on the degree of change coordination and the occurrence probability of the monitoring data;

[0020] Determine the likelihood of abnormal performance of each data point in the monitoring data based on the transition probability between any two data points of the same dimension;

[0021] Preferably, the determining the transition probability between any two data points of the same dimension in the monitoring data based on the degree of change coordination and the occurrence probability of the monitoring data includes:

[0022] In the data window to which the first moment of the first dimension belongs and the data window to which the second moment of the first dimension belongs, respectively determine the data points in the same order; wherein, the first moment and the second moment are any two moments;

[0023] Based on the data points in the same order and the corresponding degree of change coordination, determine the local similarity degree between the data points in the same order;

[0024] Based on the occurrence probability of the monitoring data and the local similarity degree, determine the transition probability between the first data point and the second data point; wherein, the first data point is the data value of the monitoring data at the first moment of the first dimension, and the second data point is the data value of the monitoring data at the second moment of the first dimension.

[0025] Preferably, the determining the local similarity degree between the data points in the same order includes:

[0026] Calculate the difference between the data points in the same order in the data windows of the two moments as the first difference;

[0027] Calculate the difference between the degrees of change coordination of the data points in the same order in the data windows of the two moments as the second difference;

[0028] Take the cumulative result of the reciprocal of the absolute value of the product of the first difference and the second difference on the data window as the local similarity degree.

[0029] Preferably, the determining the transition probability between any two data points of the same dimension in the monitoring data based on the occurrence probability of the monitoring data and the local similarity degree includes:

[0030] Obtain the probability of the second data point appearing in the monitoring data;

[0031] Determine the distance between the first moment and the second moment;

[0032] Calculate the ratio of the probability to the distance, and use the product of the ratio and the local similarity as the transition probability between the first data point and the second data point.

[0033] Preferably, determining the possibility of abnormal performance of each data point in the monitoring data based on the transition probability between any two data points in the same dimension includes:

[0034] Construct a transition matrix based on the transition probability between any two data points in the same dimension;

[0035] Determine the mean square error between the elements in the same order in the current column and any row in the transition matrix;

[0036] Determine the mean value of the mean square error based on the mean square error between the elements in the same order in the current column and each row;

[0037] Normalize the difference between the mean square error and the mean value to obtain the possibility of abnormal performance of the data point corresponding to the current column.

[0038] Preferably, determining the abnormal data points in the monitoring data based on the possibility of abnormal performance includes:

[0039] In the monitoring data, the data points corresponding to the columns with the possibility of abnormal performance greater than the set threshold are the abnormal data points; the data points corresponding to the columns with the possibility of abnormal performance less than or equal to the set threshold are not the abnormal data points;

[0040] Generate and output the operation status detection result and the alarm information based on the abnormal data points.

[0041] The beneficial effects of this application are as follows:

[0042] In this solution, after obtaining the monitoring data of multiple dimensions of the hybrid gas circuit breaker, the degree of change coordination of the monitoring data in multiple dimensions is analyzed by combining the state indication data; in this way, it is possible to avoid misjudging the normal data change as abnormal due to only analyzing the change of single-dimensional data to identify the abnormal degree. Then, by combining the degree of change coordination with the occurrence probability of the monitoring data, the possibility of abnormal performance of each data point is analyzed, so as to avoid the randomness of abnormal changes resulting in inaccurate judgment of the transfer relationship between each data, thus making the subsequent detection of abnormal data more accurate. Finally, based on the possibility of abnormal performance of each data point, the abnormal data points in the monitoring data can be quickly determined, and then the operation status of the hybrid gas circuit breaker can be detected according to the abnormal data; in this way, it can not only reduce the redundancy of the algorithm, improve the accuracy of abnormal data recognition, but also improve the accuracy of supporting the judgment of the operation status of the hybrid gas circuit breaker. Description of the Drawings

[0043] Figure 1 It is a schematic diagram of the implementation process of an intelligent detection method for the operating state of electrical equipment provided by an embodiment of the present application. Detailed implementation manners

[0044] Next, the technical solutions in the present application will be clearly and elaborately described in conjunction with the accompanying drawings. Among them, in the description of the embodiments of the present application, unless otherwise specified, " / " means "or". For example, A / B may mean A or B. The "and / or" in the text is only a description of the association relationship of the associated objects, indicating that there can be three relationships. For example, A and / or B may mean: A exists alone, A and B exist simultaneously, and B exists alone. In addition, in the description of the embodiments of the present application, "a plurality" means two or more than two.

[0045] Hereinafter, the terms "first" and "second" are only used for descriptive purposes and cannot be construed as implying or indicating relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of such features.

[0046] Next, the technical solutions provided by the embodiments of the present application will be introduced. The embodiments of the present application provide an intelligent detection method for the operating state of electrical equipment. Refer to Figure 1 , Figure 1 It is a schematic diagram of the implementation process of an intelligent detection method for the operating state of electrical equipment provided by an embodiment of the present application. The method includes:

[0047] 101. Obtain monitoring data of multiple dimensions of the hybrid gas circuit breaker.

[0048] Here, the electrical equipment may be a dedicated circuit breaker, a low-temperature-resistant cable, or a hybrid gas circuit breaker. In this embodiment, taking the hybrid gas circuit breaker as an example, monitoring data of multiple dimensions are obtained through a variety of sensors, and the sensors include but are not limited to temperature and humidity sensors, voltage sensors, current sensors, and density sensors; the monitoring data of multiple dimensions include but are not limited to temperature, humidity, voltage, current, and gas density.

[0049] The hybrid gas circuit breaker collects the above various types of relevant data through the corresponding sensors, records the relevant historical data within 24 hours, and stores the monitoring data of multiple dimensions collected in a computer or a host computer. In other embodiments, it can also be uploaded and stored in a cloud server.

[0050] 102. Determine the change coordination degree of the monitoring data in multiple dimensions based on the change response trends of the data in different dimensions.

[0051] Here, the change coordination degree is used to represent the change relationship between the monitoring data changes in one dimension and the data in other dimensions. Through the monitoring data of multiple dimensions, the change coordination degree of the data points in each dimension at any moment is determined.

[0052] Taking any one dimension of the hybrid gas circuit breaker as the reference dimension and the remaining dimensions as the control dimensions, analyze the data change response trends between the monitoring data of each dimension in the control dimensions and other control dimensions as well as the reference dimension, so as to reflect the change coordination degree of the monitoring data of each dimension through this data change response trend.

[0053] Since the hybrid gas circuit breaker has extremely high requirements for the operating environment, at this time, for any abnormal change in the dimension data, it may affect the working efficiency and safety of the hybrid gas circuit breaker to varying degrees. And the data fluctuations generated by the monitoring data of any one dimension are to a certain extent normal fluctuations. Therefore, only analyzing the changes in the data of a single dimension to identify the degree of abnormality will misjudge normal data changes as abnormalities and reduce the accuracy of anomaly detection.

[0054] In the embodiment of the present application, by monitoring the change coordination degree of the data in multiple dimensions, the problem of misjudging normal data changes as abnormalities caused by only analyzing the changes in the data of a single dimension to identify the degree of abnormality is avoided.

[0055] In some embodiments, by analyzing the co-variation relationship of the corresponding data between multiple dimensions and combining the local and global values of the data points in the dimension itself, the change coordination degree of any data point in the dimension is determined.

[0056] Specifically, when abnormal changes occur in the monitoring data of any dimension, it is very likely to drive the monitoring data of other dimensions, including the operating state of the hybrid gas circuit breaker itself, to change. Therefore, by analyzing the change relationship between the monitoring data changes of any one dimension and the data of other dimensions, the possibility of abnormal changes in the monitoring data of any one dimension can be judged.

[0057] Among them, taking any data point at any moment in the reference dimension as the target data point, the data window matching the target data point can be constructed by taking the target data point as the center and the preset time length as the radius; where the preset time length can be a custom-set time period; in this way, by constructing a corresponding data window for the target data point, it is convenient to analyze the local prominence degree of the target data point in the data window.

[0058] By analyzing the data change response trends between the data in the currently located dimension and each of the other control dimensions and reference dimensions, the data change coordination degree at any moment in any dimension can be obtained as shown in formula (1):

[0059] (1);

[0060] where x and y respectively represent any two dimensions, respectively represent the data point sets corresponding to the dimensions (for example, the first data point set and the second data point set); \(x_t\) represents a moment t in dimension x; represents the mean value of the monitored data values within the data window where the moment t is located in dimension x, represents the maximum value in all the data of the entire dimension; N represents the total number of dimensions.

[0061] \(cov()\) represents the covariance operation, represents the covariance of the data point sets corresponding to any two dimensions x and y. represents the data point sets corresponding to the two dimensions x and y 、 the sum of variances.

[0062] where, the larger the value, the higher the change coordination degree at the current moment t, and the greater the possibility of abnormal data change, and the larger the possibility coefficient the value; while the covariance and variance distribution reflect the correlation of the data distributions between the two data point sets. The larger the covariance, the stronger the positive correlation between the data in the two data point sets, and the smaller the variance, the more concentrated the distribution of the elements in the two data point sets, and the greater the correlation of the data. In this way, by analyzing the co-variation relationship of the corresponding data between dimensions and combining the numerical sizes of the data points themselves and globally in the dimension, the accuracy of anomaly detection can be improved.

[0063] 103. Based on the change coordination degree and the occurrence probability of the monitored data, determine the abnormal appearance possibility of each data point in the monitored data.

[0064] Here, after obtaining the change coordination degree of the data points at any moment in any dimension, combined with the occurrence probability of the data, analyze the possibility that each data point appears as abnormal data, that is, the abnormal appearance possibility of each data point. Among them, the occurrence probability of the monitored data can be obtained through the number of times any data point appears in the overall monitored data and the corresponding time period of the monitored data.

[0065] In some possible implementation manners, by varying the degree of coordination and the occurrence frequency of the monitoring data, the transition probability between different data points in the same dimension is determined. After that, based on the transition probability between different data points in the same dimension, the possibility of abnormal performance of each data point is obtained, so as to reduce the inability to accurately judge the transition relationship between data points caused by the randomness of abnormal changes, thereby improving the accuracy of subsequent abnormal data monitoring.

[0066] Here, the occurrence probability of the monitoring data includes the occurrence frequencies of data points at different times in the same dimension. For example, for different times t and r in dimension x, the occurrence probability of the data point at time t in the monitoring data and the occurrence probability of the data point at time r in the monitoring data are obtained. After obtaining the degree of change coordination of the data point at any time in dimension x, the local similarity degree between the data points in the same order is analyzed through the degrees of change coordination of the data points within the window where the first time is located. By combining the occurrence probability of the monitoring data with this local similarity degree, the transition probability between any two data points in the same dimension can be accurately obtained.

[0067] In some possible implementation manners, after obtaining the degree of change coordination of any data point, in the process of using the Markov algorithm to perform anomaly detection on the data points in any dimension, it is necessary to construct a Markov chain and obtain the transition probability between each data point. The transition probability represents the similarity relationship between any two data points, and according to the algorithm principle, the difference between relatively similar data points is amplified to identify the abnormal components in the data; however, in the original algorithm, when constructing the Markov chain, it is obtained according to the frequency of occurrence of each pair of adjacent states (different data points), and the abnormal changes of the data are random, and the transition relationship between each data point cannot be accurately judged only based on the frequency.

[0068] According to the analysis, for the data points of any dimension, when abnormal changes occur, different degrees of responses will appear in the monitoring data of other dimensions. Therefore, the relationship between the data points at any two times in the same dimension can be obtained by combining the data changes at the same time in other dimensions, that is, the degree of change coordination. At the same time, since the data points that generate anomalies will have a large difference from the data points at other times in the local time period, therefore, by analyzing the time difference between the data at two time points and the data difference in the local area, the transition probability between any two data points in the same dimension in the Markov chain can be obtained. As shown in formula (2):

[0069] (2);

[0070] Among them, t and r respectively represent any two times (i.e., the first time and the second time) in dimension x. Denote the monitoring data value at time r, Denote the data value corresponding to the x dimension at time r The probability of occurrence in the monitoring data, Denote the time interval between time t and time r, Denote the difference between the values of the a-th data points in the same order in the windows where time t and time r are located, Denote the difference between the change coordination degrees of the a-th data points in the same order in the data windows where time t and time r are located.

[0071] Among them, Denote the numerical difference after weighting the data points in the same order in the corresponding data windows of two times by the coordination degree; the smaller the difference, the more similar the local data distribution between any two data points in the same order, and the local similarity The larger the value, the higher the probability of data t transferring to data r; at the same time, the smaller the time interval between the two times, the smaller the probability that the data states at the two times change, the greater the probability that they are in the same state, and the greater the transfer probability; and the data value corresponding to time r The higher the probability of occurrence in the monitoring data, it indicates that after the state transfer, the monitoring data changes to The higher the probability; that is The larger the value, the greater the probability that the monitoring data value at time t changes to the monitoring data value at time r after the state transfer.

[0072] In this way, by analyzing the numerical and coordination degree differences between data points at different times in the same dimension, combined with the theory of distance and probability of the algorithm itself, the corrected transfer probability between any two data is obtained. This operation avoids the randomness of abnormal changes, which makes it impossible to accurately judge the transfer relationship between each data according to the frequency, and makes the subsequent detection of abnormal data more accurate.

[0073] Furthermore, based on the transfer probability between any two data points in the same dimension, determine the possibility of abnormal performance of each data point in the monitoring data.

[0074] Here, the likelihood of abnormal performance of each data point is used to represent the probability that each data point is an abnormal data point. In this way, the higher the likelihood of abnormal performance, the higher the probability that the data point is an abnormal data point, and the lower the likelihood of abnormal performance, the lower the probability that the data point is an abnormal data point. After obtaining the transition probability between any two data points, a transition matrix is constructed; the mean square error between the elements in the same order in different columns and rows of the transition matrix is calculated, as well as the mean value of the mean square errors, so as to further determine the likelihood of abnormal performance of each data point according to the mean square error and the mean value. In this way, by combining the degree of change coordination with the occurrence probability of monitoring data, the transition probability between any two data can be accurately obtained, thus avoiding the inability to accurately judge the transition relationship between data due to the randomness of abnormal data changes, and further facilitating the subsequent determination of abnormal data points.

[0075] In some possible implementation manners, a transition matrix is constructed through the transition probability, and the likelihood of abnormal performance of each data point can be quickly calculated from the transition probabilities in the transition matrix, which can be implemented through the following process:

[0076] First, based on the transition probability between any two data points of the same dimension, a transition matrix is constructed.

[0077] Here, after obtaining the transition probability between any two data, a transition matrix is constructed therefrom; at this time, both the horizontal and vertical axes of the matrix are in accordance with the chronological positions corresponding to the data.

[0078] Second, determine the mean square error between the elements in the same order in the current column and any row of the transition matrix.

[0079] Here, taking the current column as the nth column as an example, first determine the elements in the same order in the nth column and any row in the transition matrix. For example, the elements in the same order as the first, second, third, etc. elements in the nth column and that row. Calculate the mean square error between each two elements in the same order respectively. For example, the mean square error between the element in the nth column and the element in the same order as the ith element in any row.

[0080] Third, based on the mean square error between the elements in the same order in the current column and each row, determine the mean value of the mean square errors.

[0081] Here, after obtaining the mean square error between the elements in the same order in the current column and each row through the above process, further determine the mean value of these mean square errors.

[0082] Finally, perform a normalization process on the difference between the mean square error and the mean value to obtain the likelihood of abnormal performance of the data point corresponding to the current column.

[0083] Here, subtract the obtained multiple mean squared errors from the mean value to obtain multiple differences; calculate the reciprocal of each difference, and then calculate the mean value of the obtained multiple reciprocals. After that, perform a normalization process on this mean value to obtain the likelihood of abnormal performance of the data point corresponding to the current column. In this way, by determining the likelihood of abnormal performance generated by any data point, it is possible to avoid the problem that the transfer probability between abnormal data itself and normal data is relatively small, and some overlapping clusters will be lost after iteration, which will interfere with abnormal recognition. At the same time, it avoids the problem of too low abnormal detection efficiency caused by directly performing multiple iterations based on the Markov algorithm process, reduces the redundancy of the algorithm, and improves the accuracy of abnormal recognition.

[0084] In some possible implementation manners, during the process of clustering data using the Markov algorithm, iterative operations of matrix exponentiation and normalization are performed according to the transition probability, and a new transition matrix after normalization is obtained. According to the number of times the data points corresponding to any horizontal row appear simultaneously with other horizontal rows in several identical vertical columns in the transition matrix, several clusters are obtained; theoretically, for normal data points, the overlapping clusters generated after the above steps are fewer, while the overlapping clusters generated by abnormal data are more; however, the transfer probability between some abnormal data itself and normal data is relatively small, and some overlapping clusters will be lost after iteration, which will interfere with abnormal recognition.

[0085] After analysis, it can be seen that the transition matrix is symmetric along the diagonal before normalizing the transition probability for each column, and the premise for generating overlapping clusters of any data point is that the mutual differences of the matrix products between the corresponding column and several other horizontal rows are relatively small; and because normalization is performed column by column, the closer the transition probabilities corresponding to the same order between the current column and any row are, the higher the symmetry between the values of the current column and any row, which is manifested as more and relatively larger stretched transition probabilities are generated in the corresponding same column after matrix exponentiation and normalization to form a new matrix, and thus the greater the probability of generating more overlapping clusters in the end, and at the same time, it also indicates that the greater the likelihood of abnormal performance of the data point corresponding to the current column. Therefore, the likelihood of abnormal performance of the data point corresponding to the current column is as shown in formula (3):

[0086] (3);

[0087] In formula (3), is the likelihood of abnormal performance corresponding to the nth column, represents the transition probability corresponding to the i-th element in the nth column, represents the transition probability of the i-th element in the m-th row, represents the total number of rows in the matrix. MSE() represents the operation of mean squared error, Denotes the mean square error between the \(i\)-th element in the same order in the data of the \(n\)-th column and the \(m\)-th row (i.e., the \(i\)-th transition probability in the transition matrix). The smaller this value is, the smaller the difference between the two sequences, and the greater the degree of symmetry it indicates. Denotes the mean value of the mean square error obtained for the current column \(n\) and all rows.

[0088] Among them, Denotes the difference between the mean square error of the \(n\)-th column and the \(m\)-th row and the mean value of the mean square error obtained for the current column \(n\) and all rows. The smaller this difference is, the smaller the mutual difference of the matrix products between the corresponding column and several other horizontal rows, and the greater the possibility of generating an overlapping cluster. That is Denotes the possibility that the data points corresponding to the \(n\)-th column exhibit abnormal data. The larger this value is, the greater the possibility.

[0089] 104. Based on the possibility of exhibiting abnormality, determine abnormal data points in the monitored data.

[0090] Here, after obtaining the possibility of exhibiting abnormality for any data point, by judging the magnitude relationship between the possibility of exhibiting abnormality and a preset threshold, determine whether the data point is an abnormal data point. Obtain the data points that generate abnormalities according to the steps, and judge the operating state of the hybrid gas circuit breaker based on the abnormal data points.

[0091] In some possible implementation manners, by comparing the preset threshold and the possibility of exhibiting abnormality, judge whether the data points in the monitored data are abnormal data points, and output an alarm message for the abnormal data points. That is, the above step 104 can be implemented by the following steps 141 and 142:

[0092] 141. In the monitored data, determine the data points with a possibility of exhibiting abnormality greater than the preset threshold as the abnormal data points.

[0093] In the embodiments of the present application, a threshold of 0.7 can be set. Let the data points corresponding to the columns with a possibility of exhibiting abnormality greater than the set threshold be the abnormal data points; the data points corresponding to the columns with a possibility of exhibiting abnormality less than or equal to the set threshold are not abnormal data points. In this way, the abnormal data points in the monitored data can be determined quickly and simply.

[0094] 142. Based on the abnormal data points, generate an operating state detection result.

[0095] Here, first analyze the dimension to which the abnormal data point belongs and the source of the abnormal data point; among them, the source of the data point is used to represent the generation location of the abnormal data point, facilitating the positioning and adjustment of the abnormal data point. Then, based on the dimension to which the abnormal data point belongs, generate a detection result that matches the dimension to which the abnormal data point belongs, and output an alarm message carrying the source; in this way, the alarm message is targeted. In this way, by judging the numerical relationship between the abnormal possibility of the data point and the preset threshold, the abnormal data point is determined, and an alarm message is timely output for the abnormal data point to prompt that the operating state is abnormal, so that it is possible to not only have a lower computational complexity, but also accurately detect the abnormal data point and output the detection information and alarm message in a targeted manner. As an example, when the abnormal data point belongs to the dimension of temperature, the detection result may be that the temperature of the operating environment of the hybrid gas circuit breaker is abnormal, and the acquisition moment corresponding to the abnormal data point and the abnormal temperature are jointly used as the alarm message.

[0096] In the embodiment of the present application, by monitoring the change coordination degree of the data in multiple dimensions; in this way, it is possible to avoid the problem of misjudging the normal data change as abnormal due to only analyzing the change of the data in a single dimension to identify the abnormal degree. Then, by combining the change coordination degree with the occurrence probability of the monitored data, the abnormal possibility of each data point is analyzed, in this way, it is avoided that the randomness of the abnormal change leads to the inability to accurately judge the transfer relationship between each data, so that the subsequent detection of abnormal data is more accurate. Finally, through the abnormal possibility of each data point, the abnormal data points in the monitored data can be quickly determined; in this way, it is possible to not only reduce the redundancy of the algorithm, improve the accuracy of abnormal data recognition, but also improve the accuracy of the operating state detection of the hybrid gas circuit breaker.

[0097] It should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, rather than limiting them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

[0098] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between each embodiment can be referred to each other. The key point of each embodiment is to illustrate the differences from other embodiments.

Claims

1. An intelligent detection method for the operating state of an electrical device, characterized in that, The method includes: Obtaining monitoring data of a circuit breaker in multiple dimensions through sensors; For a first dimension among the multiple dimensions, constructing a data window that matches a target data point at any moment in the first dimension; wherein, the first dimension is any one of the multiple dimensions; based on a first data point set of the monitoring data in the first dimension and second data point sets respectively corresponding to the monitoring data in each second dimension, determining the correlation of co-variation generated by the first dimension; wherein, the second dimension is any one of the multiple dimensions and the status indication data, and the second dimension is different from the first dimension; based on the target data point and the first data point set, determining the likelihood coefficient that the target data point is an outlier; adjusting the correlation of co-variation generated by the first dimension with the likelihood coefficient to obtain the co-variation degree of change of the first dimension; Based on the co-variation degree of change and the occurrence probability of the monitoring data, determining the likelihood of each data point in the monitoring data being abnormally presented; Based on the likelihood of abnormal presentation, determining abnormal data points in the monitoring data; Based on the abnormal data points, detecting the operating status of the circuit breaker.

2. The intelligent detection method for the operating state of an electrical device according to claim 1, wherein The determining the correlation of co-variation generated by the first dimension based on the first data point set of the monitoring data in the first dimension and the second data point sets respectively corresponding to the monitoring data in each second dimension includes: Calculating the covariance between the first data point set and the second data point sets; Taking the sum of the distribution variances of the first data point set and the second data point sets as the denominator; Taking the cumulative result of the ratio of the covariance to the denominator over all second dimensions as the correlation of co-variation generated by the first dimension.

3. An intelligent detection method for the operating state of an electrical device according to claim 1, characterized in that, The determining the likelihood coefficient that the target data point is an outlier based on the target data point and the first data point set includes: Taking the ratio of the mean of the data window where the target data point is located to the maximum data value of the first data point set as the likelihood coefficient.

4. An intelligent detection method for the operating state of an electrical device according to claim 1, characterized in that, The determining the likelihood of each data point in the monitoring data being abnormally presented based on the co-variation degree of change and the occurrence probability of the monitoring data includes: Based on the co-variation degree of change and the occurrence probability of the monitoring data, determining the transition probability between any two data points in the same dimension of the monitoring data; Based on the transition probability between any two data points in the same dimension, determining the likelihood of each data point in the monitoring data being abnormally presented.

5. The intelligent detection method for the operating state of an electrical device according to claim 4, characterized in that, The determining the transition probability between any two data points in the same dimension of the monitoring data based on the co-variation degree of change and the occurrence probability of the monitoring data includes: In the data window to which the first moment in the first dimension belongs and the data window to which the second moment in the first dimension belongs, respectively determining data points in the same order; wherein, the first moment and the second moment are any two moments; Based on the data points in the same order and the corresponding co-variation degree of change, determining the local similarity degree between the data points in the same order; Determine the transition probability between a first data point and a second data point based on the occurrence probability of the monitoring data and the local similarity degree; wherein, the first data point is the data value of the monitoring data at the first moment in the first dimension, and the second data point is the data value of the monitoring data at the second moment in the first dimension.

6. The intelligent detection method for the operating state of an electrical device according to claim 5, characterized in that The determining of the local similarity degree between the data points in the same order includes: Calculate the difference between the data points in the same order in the data windows where the two moments are located as the first difference; Calculate the difference between the change coordination degrees of the data points in the same order in the data windows where the two moments are located as the second difference; Take the accumulated result of the reciprocal of the absolute value of the product of the first difference and the second difference on the data window as the local similarity degree.

7. An intelligent detection method for the operating state of an electrical device according to claim 6, characterized in that, The determining of the transition probability between any two data points in the same dimension of the monitoring data based on the occurrence probability of the monitoring data and the local similarity degree includes: Obtain the probability of the occurrence of the second data point in the monitoring data; Determine the distance between the first moment and the second moment; Calculate the ratio of the probability to the distance, and take the product of the ratio and the local similarity as the transition probability between the first data point and the second data point.

8. An intelligent detection method for the operating state of an electrical device according to any one of claims 5 to 7, characterized in that, The determining of the possibility of abnormal performance of each data point in the monitoring data based on the transition probability between any two data points in the same dimension includes: Construct a transition matrix based on the transition probability between any two data points in the same dimension; Determine the mean square error between the elements in the same order in the current column and any row in the transition matrix; Determine the mean value of the mean square error based on the mean square error between the elements in the same order in the current column and each row; Perform normalization processing on the difference between the mean square error and the mean value to obtain the possibility of abnormal performance of the data point corresponding to the current column.

9. An intelligent detection method for the operating state of an electrical device according to claim 1, characterized in that, The determining of the abnormal data points in the monitoring data based on the possibility of abnormal performance includes: In the monitoring data, the data points corresponding to the columns with the possibility of abnormal performance greater than the set threshold are the abnormal data points; the data points corresponding to the columns with the possibility of abnormal performance less than or equal to the set threshold are not abnormal data points; Generate and output the running state detection result and the warning information based on the abnormal data points.

Images