5G integrated intelligent power supply cabinet near-end wireless inspection system

Abstract

The invention relates to the technical field of power supply cabinet inspection, and discloses a 5G integrated intelligent power supply cabinet near-end wireless inspection system, which comprises a data acquisition module, a fault detection module, a maintenance feedback module and a fault verification module, wherein the data acquisition module is used for acquiring state parameters and environmental parameters of each power supply cabinet; the fault detection module calculates the fault confidence of each power supply cabinet based on the state parameters, and identifies an abnormal power supply cabinet based on the fault confidence; the fault verification module corrects the fault confidence of the abnormal power supply cabinet based on the environment parameters and the state parameters; and the maintenance feedback module is used for arranging maintenance for the abnormal power supply cabinet and carrying out feedback optimization on identification of the abnormal power supply cabinet based on a maintenance result. According to the invention, the accuracy and operation and maintenance efficiency of fault detection of the power cabinet are improved, and a technical breakthrough from passive response to intelligent prevention is realized.

Description

Technical Field

[0001] This application relates to the technical field of power cabinet patrol inspection, and specifically relates to a 5G integrated intelligent power cabinet proximal wireless patrol inspection system. Background Art

[0002] The progress of Internet of Things and edge computing technologies has promoted the development of power cabinet patrol inspection towards automation and intelligence, and a batch of remote monitoring systems based on wireless sensor networks have emerged. However, there are still deficiencies in the existing technologies for power cabinet patrol inspection.

[0003] Currently, the intelligent patrol inspection systems for power cabinets in the industry identify power cabinet anomalies by presetting fixed thresholds. Most of the existing systems adopt a unified detection threshold, without considering the parameter baseline fluctuations caused by production batches, operation duration, and load differences of power cabinets, resulting in misjudgments. Although some existing technologies use algorithms such as SVM and random forest to train historical data to achieve fault classification, the causal interpretability of multi-parameter linkage anomalies is poor, and the adaptability to sudden environmental interference is insufficient. There is a coupling relationship among parameters such as current harmonics, temperature, and fan power. The existing technologies lack the ability to model the causal conduction between parameters and are difficult to identify compound faults.

[0004] For example, the Chinese patent application with the publication number CN118100440A discloses a fault detection system for a distributed DC power cabinet, including multiple DC power cabinets, multiple battery packs, and a main controller. Multiple data transmission modules are installed inside the multiple DC power cabinets, and multiple sensor modules are installed on the inner walls of the multiple DC power cabinets and multiple battery packs. The main controller includes a data acquisition module, a data processing module, a fault diagnosis module, and a communication module. For this fault detection system of the distributed DC power cabinet, the number and layout positions of the sensor modules can be determined according to the number of DC power cabinets and battery packs, and the algorithms and data processing capabilities of the main controller can be determined according to the requirements of fault detection. Adopting a distributed structure, the fault detection function is dispersed to multiple DC power cabinets and battery packs, reducing the load and fault risk of a single device. However, this technical solution still has the problems raised in the background art of this application: lacking the ability to model the causal conduction between parameters and being difficult to identify compound faults.

[0005] The information disclosed in this background art section is only intended to increase the understanding of the overall background of this application, and should not be regarded as an admission or any form of implication that this information constitutes the prior art already known to those of ordinary skill in the art. Summary of the Invention

[0006] The technical problem to be solved by this application is to overcome the defects of the existing technologies, provide a 5G integrated intelligent power cabinet proximal wireless patrol inspection system, improve the accuracy of power cabinet fault detection and the operation and maintenance efficiency, and achieve a technological breakthrough from passive response to intelligent prevention.

[0007] To solve the above technical problems, the present application provides the following technical solutions:

[0008] A 5G integrated intelligent power cabinet proximal wireless inspection system, including a data acquisition module, a fault detection module, a maintenance feedback module, and a fault verification module; wherein:

[0009] The data acquisition module is used to collect the status parameters and environmental parameters of each power cabinet;

[0010] The fault detection module calculates the fault confidence level of each power cabinet based on the status parameters and identifies abnormal power cabinets based on the fault confidence level;

[0011] The fault verification module corrects the fault confidence level of the abnormal power cabinet based on the environmental parameters and status parameters;

[0012] The maintenance feedback module is used to arrange maintenance for the abnormal power cabinet and optimize the identification of the abnormal power cabinet based on the maintenance results.

[0013] As a preferred solution of the 5G integrated intelligent power cabinet proximal wireless inspection system of the present application, wherein: the fault detection module includes a data processing unit and a fault identification unit;

[0014] The data processing unit constructs a status feature vector of each power cabinet based on the status parameters; specifically includes: performing standardization processing on each status parameter respectively and arranging them in a specified order to form the status feature vector of the corresponding power cabinet;

[0015] The fault identification unit is configured with a reference feature vector for each power cabinet; the establishment method of the reference feature vector is as follows: continuously collect each status parameter during the stable operation of the power cabinet; for each status parameter, calculate the mean value after removing outliers to obtain the reference status parameter corresponding to the power cabinet item; perform standardization processing on each reference status parameter of each power cabinet and arrange them into the reference feature vector of each power cabinet.

[0016] As a preferred solution of the 5G integrated intelligent power cabinet proximal wireless inspection system of the present application, wherein: the fault identification unit is further configured with a fault identification strategy; the fault identification strategy is used to calculate the fault confidence level according to the status feature vector and the reference feature vector of each power cabinet, specifically including: for any power cabinet, calculate the Mahalanobis distance between the status feature vector and the reference feature vector; map the Mahalanobis distance to the fault confidence level through a non-linear activation function; the value range of the fault confidence level is ;

[0017] The fault identification unit is also configured with a confidence threshold; the fault identification strategy further includes: if the fault confidence of any power cabinet is higher than the confidence threshold, the corresponding power cabinet is an abnormal power cabinet.

[0018] As an optimal solution of the 5G integrated intelligent power cabinet proximal wireless inspection system described in the present application, wherein: the fault verification module includes a first correction unit; the first correction unit is configured with a first correction strategy for correcting the fault confidence of the abnormal power cabinet according to the state parameters;

[0019] The first correction strategy specifically includes:

[0020] Establish a causal association matrix for describing the causal association between state parameters;

[0021] The causal association matrix is a matrix of N rows and N columns, where N is the number of items of state parameters; the element in the i-th row and j-th column of the causal association matrix is denoted as , and the value ranges of both i and j are 1, 2,..., N; represents the causal strength coefficient of the j-th state parameter on the i-th state parameter;

[0022] Identify the abnormal state parameters in the abnormal power cabinet;

[0023] Based on the causal association matrix, determine the association parameters of each abnormal state parameter;

[0024] Based on the association parameters of each abnormal state parameter, identify whether there is fault conduction in the abnormal power cabinet; if there is fault conduction in the abnormal power cabinet, set the fault confidence to 1.

[0025] As an optimal solution of the 5G integrated intelligent power cabinet proximal wireless inspection system described in the present application, wherein: the method for identifying the abnormal state parameters in the abnormal power cabinet is as follows: calculate the difference between each state parameter of the abnormal power cabinet and the corresponding reference state parameter item by item as the deviation amount of each state parameter; the first correction unit is also configured with a deviation amount threshold for each state parameter; the state parameter with a deviation amount greater than the deviation amount threshold is the abnormal state parameter in the abnormal power cabinet;

[0026] The first correction unit is also configured with a causal strength threshold; based on the causal association matrix, determining the association parameters of each abnormal state parameter specifically includes: in the causal association matrix, the causal strength coefficient in the corresponding row of any abnormal state parameter is the associated causal strength coefficient of the abnormal state parameter; the association parameter of any abnormal state parameter includes all state parameters corresponding to the associated causal strength coefficients greater than the causal strength threshold;

[0027] Based on the associated parameters of each abnormal state parameter, identify whether there is a fault conduction in the abnormal power cabinet, specifically including: if any associated parameter B of any abnormal state parameter A in the abnormal power cabinet is also an abnormal state parameter, and the deviation amount of A and the deviation amount of B satisfy the fault conduction mode between A and B, then there is a fault conduction in the abnormal power cabinet.

[0028] As an optimal solution of the 5G integrated intelligent power cabinet proximal wireless patrol inspection system described in this application, wherein: the causal association matrix is established as follows:

[0029] Continuously collect each state parameter synchronously and establish the time series of each state parameter;

[0030] Based on the time series, conduct a hypothesis test on any two state parameters and assign values to the corresponding elements in the causal association matrix, specifically including:

[0031] Let represent the time series of the p-th state parameter, and let represent the time series of the q-th state parameter. The value ranges of p and q are both 1, 2, ……, N; based on 、 construct a vector autoregressive model for the p-th state parameter; make assumptions about the vector autoregressive model as follows: calculate for any positive integer t When, the weights of the previous continuous k elements in the vector autoregressive model are all 0;

[0032] Conduct a hypothesis test on the above hypothesis through a chi-square test, calculate the joint significance, denoted as R; based on the joint significance R, assign values to the corresponding elements in the causal association matrix, specifically including: if R is less than 0.05, then the value of is 1 - R; otherwise, the value of is 0.

[0033] As an optimal solution of the 5G integrated intelligent power cabinet proximal wireless patrol inspection system described in this application, wherein: the vector autoregressive model is used to calculate any element in; let the t-th element in be , the t-th element in be , t is a positive integer, then the vector autoregressive model is based on in, the previous continuous k elements and in, the previous continuous k elements are weighted and summed to calculate , k is the lag order.

[0034] As a preferred solution of the 5G integrated intelligent power cabinet proximal wireless inspection system described in this application, wherein: the fault verification module further includes a second correction unit;

[0035] The second correction unit is configured with a second correction strategy for correcting the fault confidence level of the abnormal power cabinet according to the environmental parameters and the status parameters of adjacent power cabinets;

[0036] The second correction unit is also configured with a distance threshold; the second correction unit calculates the straight-line distance between each power cabinet and the abnormal power cabinet, and marks the power cabinets with a straight-line distance less than the distance threshold as adjacent power cabinets of the abnormal power cabinet.

[0037] As a preferred solution of the 5G integrated intelligent power cabinet proximal wireless inspection system described in this application, wherein: the second correction strategy specifically includes:

[0038] Mark the status parameters corresponding to each item of the abnormal status parameters in the adjacent power cabinets as potential abnormal parameters; calculate the deviation amount of each potential abnormal parameter in each adjacent power cabinet, and identify whether there is fault conduction in each adjacent power cabinet; the second correction unit is also configured with an environmental detection model for detecting whether the environmental parameters are abnormal;

[0039] If there is no fault conduction between the abnormal power cabinet and each adjacent power cabinet, and the deviation amount of each abnormal status parameter and each potential abnormal parameter is less than the deviation threshold, and the environmental parameters are detected to be abnormal at the same time, then reduce the fault confidence level of the abnormal power cabinet;

[0040] If at least one adjacent power cabinet has fault conduction, and the deviation amount of each potential abnormal parameter of the adjacent power cabinet with fault conduction has the same positive and negative sign as the deviation amount of the corresponding abnormal status parameter of the abnormal power cabinet, then increase the fault confidence level of the abnormal power cabinet.

[0041] The method for the second correction unit to detect whether the environmental parameters are abnormal based on the environmental detection model is as follows:

[0042] Continuously obtain environmental parameters and form a time series of each environmental parameter;

[0043] Input the time series of each environmental parameter into the environmental detection model;

[0044] The environmental detection model calculates and outputs the detection result, including abnormal environmental parameters and normal environmental parameters.

[0045] As a preferred solution of the 5G integrated intelligent power cabinet proximal wireless inspection system described in this application, wherein: the maintenance feedback module includes a maintenance record unit and a feedback optimization unit;

[0046] The maintenance record unit is used to arrange maintenance for the abnormal power supply cabinet and record the maintenance results; the maintenance results include that the abnormal power supply cabinet has a fault and does not have a fault;

[0047] The feedback optimization unit is configured with a feedback optimization strategy; the feedback optimization strategy is used to perform feedback optimization on the identification of the abnormal power supply cabinet, specifically including performing feedback optimization on the confidence threshold;

[0048] The specific feedback optimization strategy is as follows:

[0049] Record the maintenance results of each maintenance and the corresponding confidence thresholds;

[0050] Calculate the true positive rate and false positive rate corresponding to each confidence threshold based on the maintenance results;

[0051] Determine the optimal confidence threshold; the optimal confidence threshold is the confidence threshold that maximizes the difference between the true positive rate and the false positive rate;

[0052] Update the confidence threshold to the optimal confidence threshold.

[0053] Compared with the prior art, the beneficial effects achieved by the present application are as follows:

[0054] By establishing a dynamic health benchmark for each device, the present application effectively distinguishes normal parameter fluctuations from real faults and improves the accuracy of fault identification. By introducing a multi-parameter correlation analysis mechanism, it can identify compound faults caused by internal causal conduction of the device and solve the blind area problem of traditional single-parameter detection.

[0055] Through the collaborative analysis of integrating device status and environmental parameters, the present application can intelligently identify abnormalities caused by the external environment and significantly reduce the false alarm rate. By converting the maintenance results of each time into system knowledge, it realizes the automatic correction and optimization of the accuracy of patrol inspection. Description of the Drawings

[0056] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings. Among them:

[0057] Figure 1 It is a schematic structural diagram of the 5G integrated intelligent power supply cabinet proximal wireless patrol inspection system provided by the present application;

[0058] Figure 2 It is a flowchart of the first correction strategy provided by the present application. Detailed Embodiments

[0059] The technical solution of the present application will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present application and the specific features in the embodiments are detailed descriptions of the technical solution of the present application, rather than limitations on the technical solution of the present application. Without conflict, the technical features in the embodiments of the present application and the embodiments can be combined with each other.

[0060] An embodiment of the present application introduces a proximal wireless inspection system for a 5G integrated intelligent power cabinet. Referring to Figure 1 , the system includes a data acquisition module, a fault detection module, a maintenance feedback module, and a fault verification module; among them:

[0061] The data acquisition module is used to collect the status parameters and environmental parameters of each power cabinet;

[0062] The data acquisition module includes a status parameter unit and an environmental parameter unit;

[0063] The status parameter unit is used to collect each status parameter of each power cabinet; the preferred status parameters in this embodiment include: current harmonic distortion rate, which reflects abnormalities in the power filter circuit and load and has a causal relationship with the capacitor aging rate; capacitance equivalent series resistance, which is the core index of capacitor deterioration; temperature difference between the inside and outside of the power cabinet, which reflects the interference of the environment on the power cabinet status; change rate of the power of the cooling fan, which captures the attenuation of the cooling efficiency; internal temperature gradient, such as the temperature difference between the air inlet, capacitor area, and air outlet, which reflects local temperature abnormalities; and DC output ripple voltage, which reflects the stability of the DC voltage.

[0064] The environmental parameter unit is used to collect environmental parameters; the preferred environmental parameters in this embodiment include: environmental temperature, grid voltage fluctuation data, and wireless signal interference data, such as the bit error rate and signal-to-noise ratio on the 5G base station side. The preferred environmental parameters in this embodiment are all strongly correlated with the accuracy of the power cabinet status parameters and can assist in determining whether the abnormality of the status parameters is caused by a power cabinet fault or a misjudgment caused by environmental interference.

[0065] The fault detection module calculates the fault confidence level of each power cabinet based on the status parameters and identifies abnormal power cabinets based on the fault confidence level;

[0066] The fault detection module includes a data processing unit and a fault identification unit;

[0067] The data processing unit constructs a status feature vector of each power cabinet based on the status parameters; specifically, it includes: performing standardization processing on each status parameter respectively and arranging them in a specified order to form the status feature vector of the corresponding power cabinet; the methods of standardization processing include but are not limited to Z-score standardization, logarithmic transformation to compress the dimension, etc.

[0068] The fault identification unit is configured with a reference feature vector for each power cabinet; the reference feature vector is established as follows: during the stable operation of the power cabinet, each status parameter is continuously collected; for each status parameter, the outliers are removed and the mean value is calculated to obtain the reference status parameter corresponding to the power cabinet; each reference status parameter of each power cabinet is standardized and arranged into the reference feature vector of each power cabinet.

[0069] The fault identification unit is also configured with a fault identification strategy; the fault identification strategy is used to calculate the fault confidence level according to the status feature vector and the reference feature vector of each power cabinet, specifically including: for any power cabinet, calculate the Mahalanobis distance between the status feature vector and the reference feature vector; map the Mahalanobis distance to the fault confidence level through a non-linear activation function; the value range of the fault confidence level is . In this embodiment, an extended form of the Sigmoid function is preferably used as the non-linear activation function to map the Mahalanobis distance between the status feature vector and the reference feature vector to the fault confidence level of the corresponding power cabinet. The formula is as follows:

[0070] ;

[0071] where C is the fault confidence level, D is the Mahalanobis distance, is the exponential function. The conventional form of the Sigmoid function is . In this embodiment, adjustable parameters and are introduced to improve the accuracy of the fault confidence level in actually representing the fault probability. The adjustable parameters and are determined by fitting historical fault data.

[0072] The fault identification unit is also configured with a confidence threshold; the fault identification strategy further includes: if the fault confidence level of any power cabinet is higher than the confidence threshold, the corresponding power cabinet is an abnormal power cabinet.

[0073] The fault verification module corrects the fault confidence level of the abnormal power cabinet based on the environmental parameters and the status parameters;

[0074] The fault verification module includes a first correction unit and a second correction unit;

[0075] The first correction unit is configured with a first correction strategy for correcting the fault confidence level of the abnormal power cabinet according to the status parameters;

[0076] Referring to Figure 2 , the first correction strategy specifically includes:

[0077] Establish a causal association matrix for describing the causal association between status parameters;

[0078] The causal association matrix is a matrix with N rows and N columns, where N is the number of state parameters; the element in the i-th row and j-th column of the causal association matrix is denoted as , the value range of i, j is 1, 2, ..., N; represents the causal strength coefficient of the j-th state parameter to the i-th state parameter; that is, Indicates the probability that the i-th state parameter will be triggered when the j-th state parameter is abnormal.

[0079] Identify abnormal status parameters in an abnormal power supply cabinet; the method is as follows: calculate the difference between each status parameter of the abnormal power supply cabinet and the corresponding reference status parameter item by item, as the deviation of each status parameter; the first correction unit is also configured with a deviation threshold value for each status parameter; a status parameter with a deviation greater than the deviation threshold value is an abnormal status parameter of the abnormal power supply cabinet;

[0080] Based on the causal association matrix, the associated parameters of each abnormal state parameter are determined; the first correction unit is further configured with a causal strength threshold; in the causal association matrix, the causal strength coefficient in the corresponding row of any abnormal state parameter is the associated causal strength coefficient of the abnormal state parameter; the associated parameters of any abnormal state parameter include all state parameters corresponding to the associated causal strength coefficients greater than the causal strength threshold; for example, the fifth state parameter is the abnormal state parameter, and If it is greater than the causal strength threshold, the first state parameter is the associated parameter of the fifth state parameter.

[0081] Identify whether fault conduction exists in the abnormal power cabinet based on the associated parameters of each abnormal state parameter; if fault conduction exists in the abnormal power cabinet, set the fault confidence level to 1;

[0082] If any associated parameter B of any abnormal state parameter A in the abnormal power supply cabinet is also an abnormal state parameter, and the deviation of A and the deviation of B satisfy the fault conduction mode between A and B, then fault conduction exists in the abnormal power supply cabinet.

[0083] In this embodiment, the fault conduction mode between any state parameter A and its associated parameter B is preferably determined as follows: based on historical fault data, the positive and negative relationship between the deviations of state parameters A and B when both are abnormal state parameters is analyzed, and the relationship is used as the fault conduction mode between A and B. For example, the fault conduction mode is such that when the deviation of A is positive, the deviation of B is also positive.

[0084] The fault conduction path among various state parameters can be found through the causal correlation matrix. If B is the correlation parameter of A, there exists a fault conduction path where the abnormality of A is caused by the abnormality of B. For example, if A represents the current harmonic distortion rate and B represents the equivalent series resistance of the capacitor, the aging of the capacitor causes the abnormality of the equivalent series resistance of the capacitor, which in turn leads to the abnormality of the current harmonic distortion rate. If the deviation of A and the deviation of B satisfy the fault conduction mode between A and B, it indicates that the fault propagates along this path. The abnormalities of A and B are not caused by reasons such as sensor system errors or communication error codes, but due to typical faults in the power cabinet, that is, there is fault conduction. Therefore, the fault confidence level is set to 1.

[0085] The establishment method of the causal correlation matrix is as follows:

[0086] When the power cabinet is operating stably, continuously collect each state parameter synchronously and establish the time series of each state parameter; the timestamps of the time series of all state parameters are strictly aligned;

[0087] Based on the time series, perform hypothesis testing on any two state parameters and assign values to the corresponding elements in the causal correlation matrix, specifically including:

[0088] Let represent the time series of the p-th state parameter, and let represent the time series of the q-th state parameter. The value ranges of p and q are both 1, 2, ……, N; based on 、 construct the vector autoregressive model of the p-th state parameter; the vector autoregressive model is used to calculate any element in; let the t-th element in be , the t-th element in be ,t is a positive integer. Then the vector autoregressive model is based on in, the previous continuous k elements before and in, the previous continuous k elements before are weighted and summed to calculate ,k is the lag order; determine the value of k through methods such as the information criterion method;

[0089] Make assumptions about the vector autoregressive model as follows: when calculating for any positive integer t, the weights of the previous continuous k elements before in the vector autoregressive model are all 0;

[0090] Perform hypothesis testing on the assumption through the chi-square test, calculate the joint significance, denoted as R; based on the joint significance R, for the corresponding element in the causal correlation matrix Perform assignment, specifically including: if R is less than 0.05, then the value of is 1 - R; otherwise, the value of is 0.

[0091] In this embodiment, when calculating for any positive integer t in the vector autoregressive model, the weights of the previous continuous k elements are all 0; if the hypothesis holds, it means that the time series For calculating any element value in is of no help, that is the p-th state parameter corresponding to and there is no causal association between the q-th state parameter corresponding to. If the hypothesis does not hold, it means that the q-th state parameter is the associated parameter of the p-th state parameter, and the smaller the joint significance R, the stronger the causal association, the greater the value of.

[0092] The second correction unit is configured with a second correction strategy for correcting the fault confidence of the abnormal power cabinet according to the environmental parameters and the state parameters of the adjacent power cabinets;

[0093] The second correction unit is also configured with a distance threshold; the second correction unit calculates the straight-line distance between each power cabinet and the abnormal power cabinet, and marks the power cabinets with the straight-line distance less than the distance threshold as the adjacent power cabinets of the abnormal power cabinet.

[0094] The second correction strategy specifically includes:

[0095] Mark the state parameters corresponding to each item of the abnormal state parameter in the adjacent power cabinets as potential abnormal parameters; calculate the deviation amount of each potential abnormal parameter in each adjacent power cabinet, and identify whether there is fault conduction in each adjacent power cabinet; the second correction unit is also configured with an environment detection model for detecting whether the environmental parameters are abnormal;

[0096] If there is no fault conduction between the abnormal power cabinet and each adjacent power cabinet, and the deviation amount of each abnormal state parameter and each potential abnormal parameter is less than the deviation amount threshold, and at the same time the environmental parameters are detected to be abnormal, then reduce the fault confidence of the abnormal power cabinet; for example, reduce the fault confidence by 50%. At this time, the fluctuations of the state parameters in the adjacent power cabinets are all within the normal threshold range, and at this time the environmental parameters are abnormal, it is determined that the fluctuations of the state parameters are caused by environmental interference. At the same time, no fault conduction is identified in the adjacent power cabinets, further corroborating that the probability of the power cabinet having a fault is low.

[0097] If there is a fault conduction in at least one adjacent power supply cabinet, and the deviation of each potential abnormal parameter of the adjacent power supply cabinet with fault conduction has the same positive or negative sign as the deviation of the corresponding abnormal state parameter of the abnormal power supply cabinet, the fault confidence level of the abnormal power supply cabinet is increased. For example, the fault confidence level is set to 1. At this time, there is a fault conduction in the adjacent power supply cabinet, and there is a similar fluctuation of the state parameters as that of the abnormal power supply cabinet. It is determined that there is a potential fault in the abnormal power supply cabinet and the adjacent power supply cabinet with fault conduction due to environmental factors or other reasons (such as batch power supply defects, etc.).

[0098] The method for the second correction unit to detect whether the environmental parameters are abnormal based on the environmental detection model is as follows:

[0099] Continuously obtain environmental parameters and form a time series of each environmental parameter;

[0100] Input the time series of each environmental parameter into the environmental detection model;

[0101] The environmental detection model calculates and outputs the detection result, including abnormal environmental parameters and normal environmental parameters.

[0102] In this embodiment, the LSTM-Autoencoder model is preferably used as the environmental detection model. This model combines a long short-term memory network and an autoencoder, and can automatically learn and identify abnormal situations from the time series of environmental parameters.

[0103] The maintenance feedback module is used to arrange maintenance for the abnormal power supply cabinet and optimize the identification of the abnormal power supply cabinet based on the maintenance result.

[0104] The maintenance feedback module includes a maintenance record unit and a feedback optimization unit;

[0105] The maintenance record unit is used to arrange maintenance for the abnormal power supply cabinet and record the maintenance result; the maintenance result includes that the abnormal power supply cabinet has a fault and does not have a fault;

[0106] The feedback optimization unit is configured with a feedback optimization strategy; the feedback optimization strategy is used to optimize the identification of the abnormal power supply cabinet by feedback, and specifically includes optimizing the confidence threshold by feedback;

[0107] The feedback optimization strategy is specifically as follows:

[0108] Record the maintenance result of each maintenance and the corresponding confidence threshold;

[0109] Calculate the true positive rate and false positive rate corresponding to each confidence threshold based on the maintenance result;

[0110] Determine the optimal confidence threshold; the optimal confidence threshold is the confidence threshold that maximizes the difference between the true positive rate and the false positive rate;

[0111] Update the confidence threshold to the optimal confidence threshold.

[0112] In this embodiment, preferably, the confidence threshold that maximizes the difference between the true positive rate and the false positive rate is the optimal confidence threshold. The difference between the true positive rate and the false positive rate is the Youden index; the Youden index can measure the discriminative ability of the confidence threshold for the power cabinet failure, and the larger the Youden index, the better the discriminative ability of the corresponding confidence threshold. The true positive rate is the ratio of the number of maintenance records of the power cabinet with faults and identified as an abnormal power cabinet to the total number of maintenance records of the power cabinet with faults; the false positive rate is the ratio of the number of maintenance records of the power cabinet without faults but identified as an abnormal power cabinet to the total number of maintenance records of the power cabinet without faults.

[0113] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0114] The embodiments of the present application have been described above in conjunction with the accompanying drawings. However, the present application is not limited to the above specific embodiments. The above specific embodiments are merely illustrative and not restrictive. Under the inspiration of the present application, those of ordinary skill in the art can also make many forms without departing from the purpose and scope of the present application. These all fall within the protection scope of the present application.

Claims

1. 5G integrated intelligent power cabinet proximal wireless inspection system, characterized in that: It includes a data acquisition module, a fault detection module, a maintenance feedback module, and a fault verification module; among them: The data acquisition module is used to acquire the status parameters and environmental parameters of each power cabinet; The fault detection module calculates the fault confidence level of each power cabinet based on the status parameters, and identifies the abnormal power cabinets based on the fault confidence level; The fault verification module corrects the fault confidence level of the abnormal power cabinets based on the environmental parameters and status parameters; The maintenance feedback module is used to arrange maintenance for the abnormal power cabinets, and optimize the identification of the abnormal power cabinets based on the maintenance results.

2. The 5G integrated intelligent power cabinet proximal wireless inspection system according to claim 1, characterized in that: The fault detection module includes a data processing unit and a fault identification unit; The data processing unit constructs a status feature vector of each power cabinet based on the status parameters; specifically, it includes: performing standardization processing on each status parameter respectively, and arranging them in a specified order to form the status feature vector of the corresponding power cabinet; The fault identification unit is configured with a reference feature vector for each power cabinet; the reference feature vector is established as follows: continuously acquire each status parameter during the stable operation of the power cabinet; for each status parameter, calculate the mean value after removing the outliers to obtain the reference status parameter corresponding to the power cabinet; perform standardization processing on each reference status parameter of each power cabinet and arrange them into the reference feature vector of each power cabinet.

3. The 5G integrated intelligent power cabinet proximal wireless inspection system according to claim 2, characterized in that: The fault identification unit is further configured with a fault identification strategy; the fault identification strategy is used to calculate the fault confidence according to the state feature vector and the reference feature vector of each power cabinet, and specifically includes: for any power cabinet, calculating the Mahalanobis distance between the state feature vector and the reference feature vector; mapping the Mahalanobis distance to the fault confidence through a non-linear activation function; the value range of the fault confidence is ; The fault identification unit is also configured with a confidence threshold; the fault identification strategy also includes: if the fault confidence level of any power cabinet is higher than the confidence threshold, the corresponding power cabinet is an abnormal power cabinet.

4. The 5G integrated intelligent power cabinet proximal wireless inspection system according to claim 3, characterized in that: The fault verification module includes a first correction unit; the first correction unit is configured with a first correction strategy for correcting the fault confidence level of the abnormal power cabinets according to the status parameters; The first correction strategy specifically includes: Establish a causal association matrix for describing the causal correlation between status parameters; The causal association matrix is an N-row and N-column matrix, where N is the number of items of the state parameters; the element in the i-th row and j-th column of the causal association matrix is denoted as , and the value ranges of both i and j are 1, 2, ……, N; represents the causal intensity coefficient of the j-th state parameter on the i-th state parameter; Identify the abnormal status parameters in the abnormal power cabinets; Based on the causal association matrix, determine the associated parameters of each abnormal status parameter; Based on the associated parameters of each abnormal status parameter, identify whether there is fault conduction in the abnormal power cabinets; if there is fault conduction in the abnormal power cabinets, set the fault confidence level to 1.

5. The 5G integrated smart power cabinet proximal wireless inspection system according to claim 4, characterized in that: The method for identifying the abnormal status parameters in the abnormal power cabinets is as follows: calculate the difference between each status parameter of the abnormal power cabinet and the corresponding reference status parameter item by item as the deviation amount of each status parameter; the first correction unit is also configured with a deviation amount threshold for each status parameter; the status parameter with a deviation amount greater than the deviation amount threshold is the abnormal status parameter in the abnormal power cabinet; The first correction unit is also configured with a causal strength threshold; Based on the causal association matrix, determine the associated parameters of each abnormal status parameter, specifically including: in the causal association matrix, the causal strength coefficient in the corresponding row of any abnormal status parameter is the associated causal strength coefficient of the abnormal status parameter; the associated parameters of any abnormal status parameter include all status parameters corresponding to the associated causal strength coefficients greater than the causal strength threshold; Based on the associated parameters of each abnormal state parameter, identify whether there is a fault conduction in the abnormal power cabinet, specifically including: if any associated parameter B of any abnormal state parameter A in the abnormal power cabinet is also an abnormal state parameter, and the deviation of A and the deviation of B satisfy the fault conduction mode between A and B, then there is a fault conduction in the abnormal power cabinet.

6. The 5G integrated intelligent power cabinet proximal wireless inspection system according to claim 5, characterized in that: The establishment method of the causal association matrix is as follows: Synchronously and continuously collect each state parameter and establish the time series of each state parameter; Based on the time series, perform hypothesis testing on any two state parameters and assign values to the corresponding elements in the causal association matrix, specifically including: Let represent the time series of the p-th state parameter. Let represent the time series of the q-th state parameter, where the value ranges of p and q are both 1, 2, ……, N; Based on and construct a vector autoregressive model for the p-th state parameter; Make assumptions about the vector autoregressive model as follows: When calculating for any positive integer t, the weights of the consecutive k elements before in the vector autoregressive model are all 0; Perform a hypothesis test on the said hypothesis through a chi-square test, calculate the joint significance, denoted as R; assign values to the corresponding elements in the causal association matrix based on the joint significance R, specifically including: if R is less than 0.05, then the value of is 1 - R; otherwise, the value of is 0.

7. The proximal wireless inspection system for the 5G integrated intelligent power cabinet according to claim 6, wherein: The vector autoregressive model is used to calculate Any element in The tth element in , The tth element in , t is a positive integer, then the vector autoregressive model is based on middle, The previous k consecutive elements and middle, Calculate the weighted sum of the previous k consecutive elements , k is the lag order.

8. The 5G integrated intelligent power cabinet proximal wireless inspection system according to claim 7, characterized in that: The fault verification module further includes a second correction unit; The second correction unit is configured with a second correction strategy for correcting the fault confidence of the abnormal power cabinet according to the environmental parameters and the state parameters of the adjacent power cabinets; The second correction unit is further configured with a distance threshold; The second correction unit calculates the straight-line distance between each power cabinet and the abnormal power cabinet, and marks the power cabinets with the straight-line distance less than the distance threshold as the adjacent power cabinets of the abnormal power cabinet.

9. The 5G integrated intelligent power cabinet proximal wireless inspection system according to claim 8, characterized in that: The second correction strategy specifically includes: Mark the state parameters corresponding to the abnormal state parameters in each adjacent power cabinet as potential abnormal parameters; calculate the deviation of each potential abnormal parameter in each adjacent power cabinet, and identify whether there is a fault conduction in each adjacent power cabinet; the second correction unit is further configured with an environmental detection model for detecting whether the environmental parameters are abnormal; If there is no fault conduction in the abnormal power cabinet and each adjacent power cabinet, and the deviation of each abnormal state parameter and each potential abnormal parameter is less than the deviation threshold, and the environmental parameters are detected to be abnormal at the same time, then reduce the fault confidence of the abnormal power cabinet; If there is a fault conduction in at least one adjacent power cabinet, and the deviation of each potential abnormal parameter in the adjacent power cabinet with fault conduction has the same positive or negative sign as the deviation of the corresponding abnormal state parameter of the abnormal power cabinet, then increase the fault confidence of the abnormal power cabinet.

10. The 5G integrated intelligent power cabinet proximal wireless inspection system according to claim 9, characterized in that: The maintenance feedback module includes a maintenance record unit and a feedback optimization unit; The maintenance record unit is used to arrange maintenance for the abnormal power cabinet and record the maintenance results; the maintenance results include that the abnormal power cabinet has a fault and does not have a fault; The feedback optimization unit is configured with a feedback optimization strategy; the feedback optimization strategy is used to perform feedback optimization on the identification of the abnormal power cabinet, specifically including performing feedback optimization on the confidence threshold; The feedback optimization strategy is specifically as follows: Record the maintenance results of each maintenance and the corresponding confidence thresholds; Calculate the true positive rate and false positive rate corresponding to each confidence threshold based on the maintenance results; Determine the optimal confidence threshold; the optimal confidence threshold is the confidence threshold that maximizes the difference between the true positive rate and the false positive rate; Update the confidence threshold to the optimal confidence threshold.

Images