Concentrator intelligent fusion terminal based on multi-sensor monitoring and maintenance

Abstract

The invention relates to the technical field of fault monitoring, in particular to a concentrator intelligent fusion terminal based on multi-sensor monitoring and maintenance. Obtaining a suspected fault moment according to discrete features of data distribution in the state data sequence of any dimension; obtaining a target moment according to the data difference characteristics of the suspected fault moment and the adjacent moment; state characterization points in a multi-dimensional space are constructed according to the data of all dimensions at the target moment, clustering is carried out, and the fault characterization credibility of the state cluster is obtained according to the distribution characteristics of the state characterization points in the state cluster and the data discrete characteristics between the state characterization points. According to the method, the abnormal degree of the latest moment is obtained according to the distance feature between the state characterization point corresponding to the latest moment and the nearest state cluster, the range feature of the nearest state cluster and the fault characterization credibility; the operation state of the concentrator is monitored according to the abnormal degree, and the monitoring accuracy of the concentrator terminal is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of fault monitoring, in particular to a concentrator intelligent fusion terminal based on multi-sensor monitoring and maintenance. BACKGROUND

[0002] The concentrator intelligent fusion terminal is an intelligent terminal device integrating functions of collection, processing, transmission and control, mainly used for data centralized management and intelligent fusion control in the power industry. The concentrator needs to be frequently maintained during operation to ensure normal work, and the multi-sensor monitoring can further realize intelligent and automatic management and maintenance of the concentrator terminal.

[0003] The existing technology usually monitors the sensors of the concentrator intelligent terminal separately, and each sensor judges the abnormality based on the threshold set by itself, ignoring the fact that device failure usually reflects in multiple characteristic dimensions. For example, the temperature rise monitored by the temperature sensor or the fan speed change monitored by the speed sensor cannot completely explain whether there is an abnormality, but if the temperature rise and fan speed anomaly are monitored at the same time, it can better reflect the fault of the fan failure causing the abnormal change of the temperature of the concentrator terminal. Therefore, in the existing technology, the abnormality of the concentrator terminal is judged only by the threshold set by each sensor, which is easy to cause fault false alarm and reduce the accuracy of terminal abnormality monitoring. SUMMARY

[0004] In order to solve the above technical problems, the purpose of the present application is to provide a concentrator intelligent fusion terminal based on multi-sensor monitoring and maintenance, and the technical solution adopted is as follows: A data acquisition module is used to acquire state data sequences of different dimensions for monitoring the running state of the concentrator; A data analysis module is used to obtain a suspected fault time according to the discrete features of data distribution in the state data sequence of any dimension, and obtain a target time according to the data difference features of the suspected fault time and the adjacent time in the any dimension; A state cluster analysis module is used to construct state representation points in a multi-dimensional space according to the data of all dimensions of the target time, cluster all state representation points to obtain different state clusters, and obtain a fault representation credibility of the state cluster according to the distribution features of the state representation points in the state cluster and the data discrete features between the state representation points; A fault monitoring module is used to acquire a nearest state cluster according to the state representation points corresponding to the data of all dimensions of the latest time, obtain an abnormality degree of the latest time according to the distance features of the state representation points corresponding to the latest time and the nearest state cluster, the range features of the nearest state cluster and the fault representation credibility, and monitor the running state of the concentrator according to the abnormality degree.

[0005] Further, the step of obtaining the suspected fault time according to the discrete feature of data distribution in the state data sequence of any dimension comprises: calculating the absolute value of the difference between the data at any time in the state data sequence and the average value of the data at all times, and normalizing to obtain the data difference degree at the any time; and taking the time when the data difference degree exceeds a preset difference threshold as the suspected fault time corresponding to the any dimension.

[0006] Further, the step of obtaining the target time according to the data difference feature between the suspected fault time and the adjacent time in the any dimension comprises: calculating the sum of the absolute values of the differences between the suspected fault time and the adjacent times before and after in the any dimension to obtain the adjacent difference value; calculating the reciprocal of the adjacent difference value and normalizing to obtain the neighborhood similarity of the suspected fault time; and taking the suspected fault time when the neighborhood similarity exceeds a preset similarity threshold as the target time.

[0007] Further, the step of clustering all state representation points to obtain different state clusters comprises: clustering by K-means clustering algorithm according to the distance between state representation points to obtain different state clusters.

[0008] Further, the step of obtaining the fault representation credibility of a state cluster according to the distribution feature of state representation points in the state cluster and the data discrete feature between state representation points comprises: calculating the average of the Euclidean distances between all state representation points in the state cluster and the nearest other state representation points to obtain the average nearest distance; calculating the reciprocal of the average nearest distance and normalizing to obtain the intra-cluster density; calculating the reciprocal of the Euclidean distance between the two state representation points farthest in the state cluster and normalizing to obtain the range concentration; calculating the sum of the data standard deviations of all dimensions corresponding to all state representation points in the state cluster to obtain the comprehensive discrete degree; calculating the reciprocal of the comprehensive discrete degree and normalizing to obtain the distribution uniformity; calculating the sum of the intra-cluster density and the range concentration to obtain the cluster feature value; and calculating the product of the cluster feature value and the distribution uniformity and positively correlating to map to obtain the fault representation credibility of the state cluster.

[0009] Further, the step of obtaining the nearest state cluster according to the state representation point corresponding to the data of all dimensions at the latest time comprises: taking the state cluster closest to the state representation point corresponding to the latest time as the nearest state cluster.

[0010] Further, the step of obtaining the abnormality degree of the latest time according to the distance feature of the state representation point corresponding to the latest time from the nearest state cluster, the range feature of the nearest state cluster and the fault representation credibility comprises: calculating the Euclidean distance between the state representation point corresponding to the latest time and the cluster center of the nearest state cluster to obtain an interval distance; calculating the maximum value of the Euclidean distance between the state representation points in the nearest state cluster and the cluster center to obtain an intra-cluster maximum distance; calculating the ratio of the intra-cluster maximum distance to the interval distance to obtain a closeness; and calculating the product of the closeness and the fault representation credibility of the nearest state cluster to obtain the abnormality degree of the latest time.

[0011] Further, the step of monitoring the running state of the concentrator according to the abnormality degree comprises: When the abnormality degree exceeds a preset abnormality threshold, the running state of the concentrator is abnormal.

[0012] The present application has the following advantages: In the present application, the suspected fault time can be determined according to the discrete feature of the historical data distribution to determine the time when the concentrator is likely to fail; since not all suspected fault times are caused by real data anomalies, the target time can be obtained to remove the times that cannot reflect the historical real data anomalies, thereby improving the accuracy of subsequent fault monitoring. The state representation point can reflect the data distribution of each dimension at the target time, and the different state clusters can distinguish the state representation points representing different multi-dimensional data features; meanwhile, the state cluster can facilitate the final fault monitoring of the concentrator at the latest time. The distribution of the state representation points in the state cluster can reflect the similarity degree of different dimensions and the correlation degree of multi-dimensional data, thereby determining whether the state cluster can accurately reflect the failure of the concentrator, and the fault representation credibility can reflect the credibility of the state cluster representing the failure of the concentrator, thereby further improving the accuracy of the final fault monitoring. The nearest state cluster can determine the state cluster that the state representation point at the latest time approaches, thereby determining whether the state representation point represents the failure of the concentrator. The abnormality degree can accurately reflect whether the concentrator fails at the latest time; and finally, the running state of the concentrator is monitored according to the abnormality degree, which reduces the false alarm rate compared to the method of determining the abnormality of each sensor based on the threshold set by itself, thereby improving the accuracy of the abnormality monitoring of the concentrator terminal. BRIEF DESCRIPTION OF DRAWINGS

[0013] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art and the advantages thereof, the drawings needed in the embodiments or the prior art description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and those skilled in the art can obtain other drawings according to these drawings without any creative effort.

[0014] Figure 1 A centralized concentrator intelligent fusion terminal based on multi-sensor monitoring maintenance is provided in an embodiment of the present application. DETAILED DESCRIPTION

[0015] In order to further clarify the technical means and effects taken by the present application to achieve the predetermined object, the specific embodiments, structure, features and effects of a centralized concentrator intelligent fusion terminal based on multi-sensor monitoring maintenance according to the present application are described in detail as follows in combination with the drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0016] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs.

[0017] The specific scheme of the centralized concentrator intelligent fusion terminal based on multi-sensor monitoring maintenance provided by the present application is described in detail below in combination with the drawings.

[0018] Please refer to Figure 1 which shows a centralized concentrator intelligent fusion terminal block diagram based on multi-sensor monitoring maintenance provided by an embodiment of the present application. The terminal includes the following modules: The data acquisition module S1 is used to acquire state data sequences of different dimensions of the monitoring concentrator running state.

[0019] A plurality of sensors are installed in the centralized concentrator intelligent fusion terminal for data monitoring. In the embodiment of the present application, the terminal is monitored by temperature sensors, humidity sensors, voltage and current sensors, fan speed sensors, and smoke sensors. In order to perform data fusion analysis, the data acquisition frequency of all sensors is the same, which is 1 time per second in the embodiment of the present application. The implementer can determine the type and frequency of the collected data according to the implementation scene. The data sequence collected by each sensor is normalized, and the normalization method is linear normalization in the embodiment of the present application, thereby obtaining state data sequences of different dimensions of the monitoring concentrator running state. The state data sequence under each dimension can reflect the running state characteristics of the concentrator. In the embodiment of the present application, the range of the state data sequence is the historical data of half a year before the latest time, and the state data sequence will be updated with the update of the latest time. The implementer can determine it according to the implementation scene.

[0020] The data analysis module S2 is configured to obtain a suspected fault time according to a discrete feature of data distribution in the state data sequence of any dimension, and obtain a target time according to a data difference feature between the suspected fault time and an adjacent time in any dimension.

[0021] Since the time range of the state data sequence is wide, the state data sequence contains data in various abnormal states of the concentrator during operation. Therefore, the time point that may represent the concentrator fault anomaly in the state data sequence is determined first. In the monitoring process, since the duration of the fault only accounts for a very small part of the concentrator operation process, the operation of the concentrator is normal in most time ranges, and the state data corresponding to the normal period is relatively stable. Therefore, the average level of the data in the state data sequence is close to the data level during normal operation to a large extent. Then, the suspected fault time can be obtained according to the discrete feature of data distribution in the state data sequence of any dimension. Preferably, in the embodiment of the present application, the step of obtaining the suspected fault time includes: calculating the absolute value of the difference between the data at any time in the state data sequence and the average value of the data at all times and normalizing to obtain the data difference degree of the any time. When the operation of the any time is more normal, the data is closer to the average value of the data at all times, and the data difference degree is smaller. When the operation of the any time is more abnormal, the abnormal data deviates more from the average value of the data at all times, and the data difference degree is larger. The time point at which the data difference degree exceeds the preset difference threshold is regarded as the suspected fault time corresponding to the any dimension. The data difference degree is obtained by linear normalization. For the data at the normal time, the data difference degree tends to 0 after linear normalization. For the data at the obviously abnormal time, the data difference degree tends to 1 after linear normalization, and the two tend to the ends of the value range. Therefore, in the embodiment of the present application, the preset difference threshold is 0.5, and exceeding the preset difference threshold means that the any time has obvious data abnormality feature, which is more likely to represent the concentrator fault anomaly in the any dimension.

[0022] Further, since the sensor may appear accidental noise data in the data acquisition process, the value thereof will also have a large difference with the data average level, and then be considered as a suspected fault time, so as to improve the accuracy of subsequent concentrator abnormal monitoring, the true abnormal time in the suspected fault time needs to be judged. Because the noise data presents random mutation and irregularity, and the concentrator fault anomaly usually lasts for a period of time, so the real data abnormal time presents the characteristics of continuous concentration. Therefore, the target time can be obtained according to the data difference characteristics of the suspected fault time and the adjacent time in the arbitrary dimension; preferably, in the embodiment of the present application, the step of obtaining the target time comprises: calculating the sum value of the absolute value of the difference between the suspected fault time and the adjacent time in the arbitrary dimension, obtaining the adjacent difference value; the greater the sum value of the absolute value of the difference between the suspected fault time and the adjacent time, the greater the adjacent difference value, which means that the discrete mutation characteristics of the suspected fault time are more obvious, and the suspected fault time is more likely to be caused by noise data. The reciprocal of the adjacent difference value is calculated and normalized to obtain the neighborhood similarity of the suspected fault time; the greater the neighborhood similarity means that the data between the adjacent times are more similar, and the suspected fault time is more likely to represent the real data abnormal time, and the suspected fault time whose neighborhood similarity exceeds the preset similarity threshold is taken as the target time, since the neighborhood similarity is obtained by linear normalization, therefore the neighborhood similarity of the suspected fault time corresponding to the noise tends to 0, and the neighborhood similarity of the suspected fault time corresponding to the real data anomaly tends to 1, both tend to the two ends of the value range; therefore, in the embodiment of the present application, the preset similarity threshold is 0.5; the target time represents the time when the concentrator appears real data anomaly in different dimensions in history.

[0023] The state cluster analysis module S3 is configured to construct state representation points in a multi-dimensional space according to the data of all dimensions of the target time; cluster all state representation points to obtain different state clusters; and obtain a fault representation credibility of the state cluster according to the distribution characteristics of the state representation points in the state cluster and the data dispersion characteristics between the state representation points.

[0024] After the target time in all dimensions is acquired, the multi-dimensional correlation feature of the fault can be analyzed; first, a state representation point in the multi-dimensional space is constructed according to the data of all dimensions of the target time, for example, in the corresponding state data sequence in the temperature dimension, a certain time is the target time, then the data of all dimensions corresponding to the target time is taken to construct a state representation point in the multi-dimensional space. The state representation points corresponding to all the target times in the history are distributed in the multi-dimensional space, and then the state representation points can be clustered to obtain different state clusters; preferably, in the embodiment of the present application, the state clusters are obtained by K-means clustering algorithm according to the distance between the state representation points. It should be noted that the K-means clustering algorithm belongs to the prior art, and the number of clustering clusters is obtained by the elbow method, and the specific clustering steps are not described again. By clustering, state representation points with similar multi-dimensional data correlation features can be clustered into a cluster, for example, when the concentrator appears abnormal circuit due to high humidity, the humidity, voltage and current monitoring dimensions will be abnormal, therefore, in the multi-dimensional space, the state clusters corresponding to the abnormal fault behavior will be formed between the state representation points with high abnormality in the three dimensions and low abnormality in other dimensions. After clustering, each state cluster in the clustering result can reflect a fault type of the concentrator.

[0025] Further, when the state representation points in the state cluster are more closely concentrated, it means that the similarity of the data of each dimension of the different state representation points in the state cluster is higher, the state cluster can more accurately reflect the fault type under the multi-dimensional data correlation feature, and the reliability of the fault type represented by the state cluster is higher. When the distribution of the state representation points in the state cluster presents a non-uniform feature of partial dispersion and partial concentration, it means that the state representation points in the state cluster have poor distribution concentration, the similarity of the data of each dimension is lower, and the similarity of the multi-dimensional data correlation feature is weaker, which means that the state cluster is difficult to accurately represent a certain fault type of the concentrator, and the fault representation reliability of the state cluster is lower. Therefore, the fault representation reliability of the state cluster can be obtained according to the distribution feature of the state representation points in the state cluster and the data dispersion feature between the state representation points.

[0026] Preferably, in the embodiments of the present application, the step of obtaining the fault representation credibility comprises: calculating the average value of the Euclidean distance between all state representation points in the state cluster and the nearest other state representation points to obtain the average nearest distance; the average nearest distance represents the proximity between the state representation points in the cluster, and the smaller the average nearest distance, the more concentrated the state representation points in the state cluster. The reciprocal of the average nearest distance is calculated and normalized to obtain the intra-cluster density; the greater the intra-cluster density, the higher the similarity of the dimensional data, the more obvious the correlation characteristics of the multi-dimensional data, and the higher the credibility of the state cluster reflecting the concentrator fault. The reciprocal of the Euclidean distance between the two state representation points farthest apart in the state cluster is calculated and normalized to obtain the range concentration; the greater the range concentration, the smaller the range of the state cluster, and the more concentrated the state representation points, the higher the credibility of the state cluster reflecting the concentrator fault. The sum of the data standard deviations of all dimensions corresponding to all state representation points in the state cluster is calculated to obtain the comprehensive dispersion; the comprehensive dispersion reflects the data dispersion degree of each dimension of the state representation points in the cluster, and the smaller the comprehensive dispersion, the more uniform the numerical distribution of each dimension, and the smaller the dispersion degree, which means that the distribution of the state representation points in the state cluster is more uniform. The reciprocal of the comprehensive dispersion is calculated and normalized to obtain the distribution uniformity; the greater the distribution uniformity, the higher the credibility of the state cluster representing the concentrator fault. The sum of the intra-cluster density and the range concentration is calculated to obtain the cluster feature value; the greater the cluster feature value, the smaller the range of the state cluster, and the closer the distance between adjacent state representation points. The product of the cluster feature value and the distribution uniformity is calculated and positively correlated to obtain the fault representation credibility of the state cluster; when there are multiple clusters of state representation points in the state cluster, and there is a certain distance between each state representation point cluster, the values of the intra-cluster density and the range concentration are still large at this time, but the state representation points in the state cluster are distributed in multiple clusters, the intra-cluster density feature is weak, which leads to errors in the fault representation credibility; therefore, by calculating the product of the cluster feature value and the distribution uniformity, the accuracy of the fault representation credibility of the state cluster is further improved. The greater the fault representation credibility, the more the state cluster can reflect the abnormal situation of the concentrator; obtaining the fault representation credibility can improve the accuracy of subsequent judgment of whether it is abnormal at the latest time. The formula for obtaining the fault representation credibility comprises:

[0027] wherein R represents the fault representation credibility, represents normalization, and L represents the average nearest distance, represents the intra-cluster density, and H represents the Euclidean distance between the two state representation points farthest apart in the state cluster, represents the range concentration, represents the cluster feature value, and S represents the comprehensive dispersion, Indicates the uniformity of distribution. The constant 2 in the denominator is to make a positive correlation mapping to the numerator, the value range of the numerator is 0 to 2, and after the positive correlation mapping, it is 0 to 1, which is convenient for subsequent analysis.

[0028] The fault monitoring module S4 is configured to obtain a nearest state cluster according to the state feature points corresponding to the data of all dimensions at the latest time; obtain an abnormality degree at the latest time according to the distance feature of the state feature point corresponding to the latest time and the nearest state cluster, the range feature of the nearest state cluster, and the fault feature credibility; and monitor the running state of the concentrator according to the abnormality degree.

[0029] After obtaining the fault feature credibility of different state clusters, whether the concentrator has a fault anomaly can be determined according to the state feature point corresponding to the latest time. First, the nearest state cluster is obtained according to the state feature points corresponding to the data of all dimensions at the latest time, which specifically includes: taking the state cluster closest to the state feature point corresponding to the latest time as the nearest state cluster; the closer the state feature point corresponding to the latest time is to the nearest state cluster, the closer the state feature represented by the state feature point corresponding to the latest time is to the state feature represented by the nearest state cluster. Therefore, the abnormality degree at the latest time is obtained according to the distance feature of the state feature point corresponding to the latest time and the nearest state cluster, the range feature of the nearest state cluster, and the fault feature credibility.

[0030] Preferably, in the embodiment of the present application, the step of obtaining the abnormality degree includes: calculating the Euclidean distance between the state feature point corresponding to the latest time and the cluster center of the nearest state cluster to obtain an interval distance; the smaller the interval distance is, the more likely it is that the state feature point belongs to the nearest state cluster. Calculating the maximum value of the Euclidean distance between the state feature point and the cluster center in the nearest state cluster to obtain an intra-cluster maximum distance; the intra-cluster maximum distance can reflect the range size of the state cluster. Calculating the ratio of the intra-cluster maximum distance to the interval distance to obtain a closeness; the greater the interval distance is than the intra-cluster maximum distance, the smaller the closeness is, which means that the state feature point is farther away from the nearest state cluster, and the less likely the concentrator is to have a fault at the latest time; the smaller the interval distance is than the intra-cluster maximum distance, the greater the closeness is, which means that the state feature point is in the nearest state cluster and close to the cluster center, and the more likely the concentrator is to have an anomaly at the latest time. Calculating the product of the closeness and the fault feature credibility of the nearest state cluster to obtain the abnormality degree at the latest time; the greater the fault feature credibility of the nearest state cluster is, the higher the credibility of the nearest state cluster in representing the fault anomaly of the concentrator is, and at the same time, the closer the state feature point is to the nearest state cluster, the greater the possibility of the concentrator having an anomaly at the latest time is, and the higher the abnormality degree is. The higher the abnormality degree is, the closer the state feature point at the latest time is to the nearest state cluster, and the higher the credibility of the nearest state cluster in representing the fault of the concentrator is. The formula for obtaining the abnormality degree includes:

[0031] In the formula, F represents the abnormality degree at the latest time, D represents the maximum distance in the cluster, and d represents the interval distance, represents the closeness, represents the fault representation credibility of the recent state cluster.

[0032] Finally, the running state of the concentrator can be monitored according to the abnormality degree, specifically including: when the abnormality degree exceeds a preset abnormality threshold, the running state of the concentrator is abnormal; in the embodiment of the present application, the preset abnormality threshold is 0.6, and the implementer can determine it according to the implementation scene. When the concentrator is abnormal, timely warning and notification of maintenance personnel for inspection and maintenance are performed, so as to realize the function of the concentrator intelligent terminal for self-monitoring of fault abnormalities through multiple sensors, and reduce the frequency of daily maintenance of personnel. At this point, whether the concentrator intelligent terminal is abnormal is analyzed through the multi-dimensional data correlation of multiple sensors, compared with the method of only judging the abnormality based on the threshold set by each sensor, the fault false alarm rate is reduced, and the accuracy of the concentrator terminal abnormality monitoring is improved.

[0033] In summary, the embodiment of the present application provides a concentrator intelligent fusion terminal based on multi-sensor monitoring and maintenance; a suspected fault time is obtained according to the discrete features of data distribution in the state data sequence of any dimension; a target time is obtained according to the data difference features of the suspected fault time and the adjacent time; a state representation point in a multi-dimensional space is constructed according to the data of all dimensions of the target time and is clustered, and a fault representation credibility of a state cluster is obtained according to the distribution features of the state representation points in the state cluster and the data discrete features between the state representation points; a recent state cluster is obtained according to the state representation point corresponding to the latest time. The distance features of the state representation point corresponding to the latest time and the recent state cluster, the range features of the recent state cluster, and the fault representation credibility are used to obtain the abnormality degree at the latest time; the running state of the concentrator is monitored according to the abnormality degree, and the monitoring accuracy of the concentrator terminal is improved.

[0034] It should be noted that the above-mentioned embodiments of the present application are only for description, and do not represent the advantages and disadvantages of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are also possible or may be advantageous.

[0035] Each embodiment in the specification is described in a progressive manner, and the same or similar parts between each embodiment can be referred to each other, and each embodiment mainly describes the difference from other embodiments.

Claims

1. A concentrator intelligent fusion terminal based on multi-sensor monitoring and maintenance, characterized by: The terminal includes the following modules: A data acquisition module is used to obtain status data sequences of different dimensions for monitoring the operation status of the concentrator; A data analysis module is configured to obtain a suspected fault moment based on the discrete characteristics of data distribution in the state data sequence of any dimension; and obtain a target moment based on the data difference characteristics between the suspected fault moment and adjacent moments in the any dimension; A state cluster analysis module is used to construct state representation points in a multidimensional space based on data of all dimensions at the target moment; cluster all state representation points to obtain different state clusters; and obtain the fault representation credibility of the state cluster based on the distribution characteristics of the state representation points in the state cluster and the data discrete characteristics between the state representation points; The fault monitoring module is used to obtain the latest state cluster based on the state representation points corresponding to the data of all dimensions at the latest moment; Obtaining the abnormality degree at the latest moment based on the distance feature between the state representation point corresponding to the latest moment and the nearest state cluster, the range feature of the nearest state cluster, and the fault representation credibility; The operating status of the concentrator is monitored according to the abnormality degree.

2. The concentrator intelligent fusion terminal based on multi-sensor monitoring and maintenance according to claim 1 is characterized in that: The step of obtaining a suspected fault moment according to the discrete characteristics of data distribution in the state data sequence of any dimension comprises: The absolute value of the difference between the data at any moment in the state data sequence and the average value of the data at all moments is calculated and normalized to obtain the data difference at the any moment; the moment when the data difference exceeds the preset difference threshold is taken as the suspected fault moment corresponding to the any dimension.

3. The concentrator intelligent fusion terminal based on multi-sensor monitoring and maintenance according to claim 1 is characterized in that: The step of obtaining the target time according to the data difference characteristics between the suspected fault time and the adjacent time in the arbitrary dimension includes: The sum of the absolute values ​​of the differences between the suspected fault moment and the preceding and following adjacent moments is calculated in the arbitrary dimension to obtain an adjacent difference value; the inverse of the adjacent difference value is calculated and normalized to obtain a neighborhood similarity of the suspected fault moment; and the suspected fault moment whose neighborhood similarity exceeds a preset similarity threshold is used as the target moment.

4. The concentrator intelligent fusion terminal based on multi-sensor monitoring and maintenance according to claim 1 is characterized in that: The step of clustering all state representation points to obtain different state clusters includes: According to the distance between the state representation points, the K-means clustering algorithm is used to cluster the points and obtain different state clusters.

5. The concentrator intelligent fusion terminal based on multi-sensor monitoring and maintenance according to claim 1 is characterized in that: The step of obtaining the fault characterization credibility of the state cluster according to the distribution characteristics of the state characterization points in the state cluster and the data discrete characteristics between the state characterization points includes: Calculate the average value of the Euclidean distances between all state characterization points in the state cluster and the nearest other state characterization points to obtain the average closest distance; calculate the inverse of the average closest distance and normalize it to obtain the intra-cluster density; calculate the inverse of the Euclidean distance between the two state characterization points farthest apart in the state cluster and normalize it to obtain the range concentration; calculate the sum of the data standard deviations of all dimensions corresponding to all state characterization points in the state cluster to obtain the comprehensive discreteness; calculate the inverse of the comprehensive discreteness and normalize it to obtain the distribution uniformity; calculate the sum of the intra-cluster density and the range concentration to obtain the cluster characteristic value; calculate the product of the cluster characteristic value and the distribution uniformity and perform positive correlation mapping to obtain the fault characterization credibility of the state cluster.

6. The concentrator intelligent fusion terminal based on multi-sensor monitoring and maintenance according to claim 1 is characterized in that: The step of obtaining the most recent state cluster according to the state representation points corresponding to the data of all dimensions at the latest moment includes: The state cluster closest to the state representation point corresponding to the latest moment is taken as the latest state cluster.

7. The concentrator intelligent fusion terminal based on multi-sensor monitoring and maintenance according to claim 1 is characterized in that: The step of obtaining the abnormality degree at the latest moment according to the distance feature between the state representation point corresponding to the latest moment and the nearest state cluster, the range feature of the nearest state cluster, and the fault representation credibility includes: Calculate the Euclidean distance between the state representation point corresponding to the latest moment and the cluster center of the nearest state cluster to obtain the interval distance; calculate the maximum value of the Euclidean distance between the state representation point and the cluster center in the nearest state cluster to obtain the maximum distance within the cluster; calculate the ratio of the maximum distance within the cluster to the interval distance to obtain the proximity; calculate the product of the proximity and the fault representation credibility of the nearest state cluster to obtain the abnormality degree at the latest moment.

8. The concentrator intelligent fusion terminal based on multi-sensor monitoring and maintenance according to claim 1 is characterized in that: The step of monitoring the operating status of the concentrator according to the abnormality degree includes: When the abnormality level exceeds a preset abnormality threshold, the concentrator operation state is abnormal.

Images