Metering box remote control method, storage medium and metering box

Abstract

The invention discloses a metering box remote control method, a storage medium and a metering box. The invention belongs to the technical field of power equipment monitoring. The method comprises the following steps: acquiring operation data and environment data of a metering box through multiple sensors, and constructing a multivariable time sequence data set; taking each metering box as a federated node, training a vector autoregression model by using local data, and performing weighted aggregation to generate a global dynamic model; calculating an environment change factor in real time, and triggering model fine tuning to adapt to a sudden change scene; and generating an abnormal confidence score based on residual mahalanobis distance and fuzzy logic fusion, and calculating a maintenance priority in combination with a topological connection relation matrix. According to the method, the problems that a traditional method depends on manpower, models are rigid and data potential safety hazards exist are solved, and the anomaly detection precision, the dynamic response capability and the system safety are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power equipment monitoring, and specifically relates to a remote control method for a metering box, a storage medium, and a metering box. Background Art

[0002] Traditional metering box monitoring methods rely on manual inspections, which have problems of low efficiency and high missed inspection rates. In the prior art, static models based on fixed thresholds are difficult to adapt to dynamic environmental changes (such as sudden temperature changes and load fluctuations), and the scoring system relies on a single parameter and cannot effectively identify fuzzy anomalies (such as intermittent faults). In addition, centralized training models face the problem of data islands. The historical data of a single metering box is limited, resulting in insufficient generalization ability of the model. The data transmission process mostly uses traditional encryption protocols, which have risks of man-in-the-middle attacks and data tampering. There is an urgent need for a remote monitoring method that can integrate multi-source data, be dynamically adaptive, and be secure and reliable. Summary of the Invention

[0003] (I) Technical Problems to be Solved To solve the above problems, the present invention proposes a remote control method for a metering box, a storage medium, and a metering box, aiming to solve the problems of low efficiency and high missed inspection rates in manual inspections in the prior art.

[0004] (II) Technical Solutions A remote control method for a metering box according to the present invention includes: Obtaining operation data, environmental data, and topological data of the metering box to obtain a multivariate time series data set; Using each metering box within a specific range as a federated node, training a vector autoregressive model with local data to predict the multivariate time series relationship to obtain local model parameters, and weighted aggregating the local model parameters of each node to generate a global model and obtain a first parameter set; Calculating the environmental change factor in real time, dynamically adjusting the update frequency of the global model, and sending the updated global model to each node to replace the local model parameters; Generating a confidence score based on the Mahalanobis distance distribution statistics of historical normal data and historical fault data based on the local model parameters, determining the abnormal state of the metering box based on the confidence score, and generating a maintenance priority and sending it to the end user.

[0005] In the present invention, the topological data includes a connection relationship matrix for marking the physical connection state between nodes, and the environmental data includes a vibration spectrum. Wavelet denoising is performed on the vibration spectrum to filter out high-frequency noise; Dividing the denoised spectrum into frequency bands, calculating the energy of each frequency band, and extracting mechanical fault features.

[0006] In the present invention, the wavelet denoising of the vibration spectrum to filter out high-frequency noise includes: , where is the vibration spectrum, is the inverse discrete wavelet transform, is the threshold function, is the discrete wavelet transform; The energy of each frequency band , .

[0007] In the present invention, the mathematical expression of the vector autoregressive model is: , where, is the observation vector at time is the lag order, is the lag coefficient matrix of order is white noise; The operating data includes voltage , current , power factor and harmonic distortion rate , and the environmental data also includes temperature , humidity .

[0008] In the present invention, the weighted aggregation of the local model parameters of each node includes: , where is the local model parameters of nodes, is the data volume of the th node.

[0009] In the present invention, the method for generating the abnormal confidence score includes: Define the fuzzy membership function , and map the Mahalanobis distance to the confidence value: In the formula, is the maximum Mahalanobis distance of historical normal data, is the maximum Mahalanobis distance of historical fault data; Fuse the weights of multi-source evidence , and generate the comprehensive confidence score , where is the total number of evidence types, is the confidence score of the th evidence.

[0010] In the present invention, based on the topological connection relationship matrix and the confidence score Computing node Maintenance priority of: , where represents the connection status of node and , is the confidence score of node .

[0011] In the present invention, the dynamic adjustment of the update frequency of the global model includes: calculating the computing environment change factor , quantifying the mutation rate of environmental parameters: , where in the formula and are the instantaneous change amounts of temperature and humidity respectively, and are mapped to the model update frequency weight through the Sigmoid function . When is greater than a preset threshold, trigger real-time fine-tuning of the global model.

[0012] Another computer-readable storage medium of the present invention stores a computer program thereon, and when the program is executed by a processor, it implements the metering box remote control method described in the above technical solution.

[0013] Another metering box of the present invention, the metering box includes at least one lower-level metering box to form a metering box cluster, and the metering box cluster applies the metering box remote control method described in the above technical solution.

[0014] (III) Beneficial effects Compared with the prior art, the beneficial effects of the present invention are: In the present invention, parameters such as vibration spectrum and harmonic distortion rate are introduced, combined with environmental and operation data, to comprehensively capture mechanical and electrical fault characteristics. Cross-node parameter aggregation solves the data island problem, improves the model generalization ability, and adapts to data distributions in different regions.

[0015] In the present invention, the environmental change factor triggers real-time fine-tuning of the model, and the response speed is increased to the millisecond level, adapting to extreme environmental mutations. Fuzzy logic-evidence theory fuses multi-source evidence to generate a dynamic confidence score, supporting manual review and accurate decision-making. Description of the drawings

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0017] Figure 1Schematic diagram of the remote control method; Figure 2 Schematic diagram of the training process of VAR; Figure 3 Schematic diagram of the topological structure of the metering box node (both N1 and N4 represent the meter box, and the connection lines represent the physical connection relationship); Figure 4 Line graph of the vibration spectrum energy. Specific implementation mode

[0018] Example 1

[0019] As Figures 1-4 shown, a remote control method for a metering box aims to solve the problems of low efficiency and high missed inspection rate in manual inspection in the prior art. This remote control method mainly includes the following steps: S100. Obtain the operation data, environmental data, and topological data of the metering box to obtain a multivariate time series data set; First, the operation data and environmental data are collected in real time through sensors deployed inside and outside the metering box. The operation data includes voltage , current , power factor , and harmonic distortion rate . The environmental data includes temperature , humidity , and vibration spectrum . The voltage can be collected in real time through a voltage sensor to monitor the instantaneous value and fluctuation trend of the input / output voltage. The current can be measured by a Hall effect sensor to capture abnormal information such as load mutation or short circuit. The power factor is calculated by a power analyzer as the ratio of the active power to the apparent power, reflecting the reactive power compensation state of the power grid. The harmonic distortion rate uses FFT (Fast Fourier Transform) to analyze the current / voltage waveform and calculate the total harmonic distortion rate: , where is the fundamental wave amplitude, is the th harmonic amplitude.

[0020] Among the environmental data, the temperature is collected by a thermistor or an infrared sensor on the surface of the box body, the humidity is measured by a capacitive humidity sensor inside the box, and the vibration spectrum is collected by a piezoelectric accelerometer to output the frequency domain energy distribution. The topological data includes the connection relationship matrix , which is obtained from the power grid topology database to define the connection relationship between nodes.

[0021] Furthermore, for the obtained vibration spectrum, the Daubechies wavelet basis function is selected to process the original vibration spectrum signal Perform multi-scale decomposition to filter out high-frequency noise: , where is the inverse discrete wavelet transform, is the threshold function using the soft threshold method, is the discrete wavelet transform. Divide the denoised spectrum into frequency bands and calculate the energy of each frequency band , . The division of frequency bands is based on the characteristics of the device. For example, the low-frequency band (0 - 100 Hz): Detect mechanical looseness; the middle-frequency band (100 - 500 Hz): Identify bearing wear, and the high-frequency band (500 - 2000 Hz): Capture insulation breakdown arcs.

[0022] Normalize heterogeneous data, such as temperature, voltage, current, etc., to eliminate the dimension difference and obtain a multi-variable time series dataset , and the dataset of each metering box is stored in its respective local storage device.

[0023] S200. Take each metering box within a specific range as a federated node, use local data to train a vector autoregressive model to predict the multi-variable time series relationship to obtain local model parameters, and weighted aggregate the local model parameters of each node to generate a global model and obtain a first parameter set; Furthermore, the expression of the vector autoregressive model is: , where is the observation vector at time, is the lag order, is the order lag coefficient matrix, is white noise and . Use maximum likelihood estimation (MLE) or least squares method to solve the coefficient matrix , and the selection of the lag order is determined by the Akaike information criterion (AIC) to find the optimal value: , where is the number of variables, is the number of samples, and select the value that minimizes AIC.

[0024] Each local node outputs the trained parameter set . After the model training is completed, aggregate the local parameters by weighted average, and the weights are determined by the proportion of the data volume of each node: , where is the local model parameter of nodes, is the th node's data volume. For example, if a node has 1000 historical data, then , i.e., the total data volume of all nodes. Nodes with a large data volume have a greater impact on the global model because their parameters are trained based on richer local data, and the coefficient matrix for each lag order is the weighted average of each node .

[0025] The coefficient matrix of the VAR model encodes the dynamic coupling relationship between multiple variables. For example represents the influence coefficient of the current at lag order 1 on the current voltage , represents the influence coefficient of the humidity at lag order 2 on the current temperature . Local data of different nodes may have different parameters due to environmental or load differences. For example, coastal nodes have a high humidity (the influence coefficient of humidity on harmonic distortion) is large, and industrial area nodes have large load fluctuations, (the influence coefficient of current on voltage) is more significant.

[0026] S300. Calculate the environmental change factor in real time and dynamically adjust the update frequency of the global model. Send the updated global model to each node to replace the local model parameters; Calculate the environmental change factor in real time , such as the temperature / humidity mutation rate, and dynamically adjust the model update frequency through the Sigmoid function. When is greater than the preset threshold of 0.8, trigger real-time fine-tuning of the global model. Send the updated global parameters to each node to replace the local model parameters. That is, dynamically optimize the global VAR model to adapt to environmental and load fluctuations.

[0027] S400. Generate a confidence score based on the Mahalanobis distance distribution statistics of historical normal data and historical fault data obtained from the local model parameters. Determine the abnormal state of the metering box based on the confidence score and generate a maintenance priority and send it to the end user.

[0028] When the global VAR model has been deployed to each metering box and the Mahalanobis distance distribution of historical normal and historical fault data among each metering box has been statistically analyzed, then start multi-dimensional anomaly monitoring and confidence evaluation.

[0029] Specifically, when performing anomaly detection, calculate the residual between the actual observed value and the model predicted value Quantify the degree of deviation through Mahalanobis distance , is the covariance matrix of historical normal data. If where is the significance level with a value of 0.01, it is marked as abnormal.

[0030] The estimation method of the confidence score is as follows: Define the fuzzy membership function , and map the Mahalanobis distance to the confidence value: In the formula, is the maximum Mahalanobis distance of historical normal data, is the maximum Mahalanobis distance of historical fault data. Generate the comprehensive confidence by fusing multi-source evidences such as environmental mutation and topological association. Among them, the weight of environmental mutation , and the weight of topological association .

[0031] Fuse the weights of multi-source evidences , and generate the comprehensive confidence score , where is the total number of evidence types, is the confidence score of the th evidence. The abnormal confidence score , and add the fault type label according to the value of the confidence score, such as harmonic over-limit, mechanical looseness, etc.

[0032] Based on the topological connection relationship matrix and the confidence score calculate the maintenance priority of node : , where represents the connection status between node and , is the confidence score of node . If node is connected to multiple high-confidence abnormal nodes, the priority is doubled. Match the engineer with the closest distance and suitable skills according to the real-time position of the maintenance personnel and send the real-time dispatch instruction, so as to realize the remote control and maintenance of the metering box.

[0033] Embodiment 2

[0034] The embodiment of the present invention provides a computer-readable storage medium.

[0035] The computer program stored on the computer-readable storage medium provided by the embodiment of the present invention can implement the steps of any of the above-mentioned metering box remote control methods when executed by a processor.

[0036] The computer-readable storage medium may include: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs.

[0037] For the introduction of the computer-readable storage medium provided in the embodiments of the present invention, please refer to the above method embodiments, and the present invention will not be elaborated herein.

[0038] The various embodiments in the specification are described in a progressive manner. The key point of each embodiment is to illustrate the differences from other embodiments. For the same or similar parts among the various embodiments, reference can be made to each other. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the description in the method part.

[0039] Those skilled in the art can further realize that the units and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of the two. To clearly illustrate the interchangeability of hardware and software, the components and steps of the examples have been generally described according to functions in the above description. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0040] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be directly implemented by hardware, software modules executed by a processor, or a combination of the two. The software modules can be placed in a random access memory (RAM), internal memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium well-known in the technical field.

[0041] The above-described embodiments are only used to describe the preferred embodiments of the present invention, and do not limit the concept and scope of the present invention. Without departing from the design concept of the present invention, various modifications and improvements made by those of ordinary skill in the art to the technical solution of the present invention should fall within the protection scope of the present invention. The technical content claimed by the present invention has been fully recorded in the claims.

Claims

1. A remote control method for a metering box, characterized in that: include: Obtain the operation data, environmental data and topological data of the meter box to obtain a multivariate time series data set; Taking each meter box within a specific range as a federated node, using local data to train a vector autoregressive model to predict a multivariate time series relationship to obtain local model parameters, weightedly aggregating the local model parameters of each node to generate a global model and obtain a first parameter set; Calculate environmental change factors in real time, dynamically adjust the update frequency of the global model, and send the updated global model to each node to replace the local model parameters; A confidence score is generated by obtaining Mahalanobis distance distribution statistics of historical normal data and historical fault data based on the local model parameters, an abnormal state of the meter box is determined based on the confidence score, and a maintenance priority is generated and sent to the end user.

2. The remote control method of the metering box according to claim 1, characterized in that: The topological data includes a connection relationship matrix for marking the physical connection status between nodes, and the environmental data includes a vibration spectrum, and wavelet denoising is performed on the vibration spectrum to filter out high-frequency noise; The denoised spectrum is divided into frequency bands, calculate the energy of each frequency band, and extract the mechanical fault characteristics.

3. The remote control method of the metering box according to claim 2, characterized in that: The performing wavelet denoising on the vibration spectrum to filter out high-frequency noise comprises: ,in is the vibration spectrum, is the inverse discrete wavelet transform, is the threshold function, is discrete wavelet transform; Energy in each frequency band , .

4. The remote control method of the metering box according to claim 3 is characterized in that: The mathematical expression of the vector autoregression model is: ,in for The observation vector at time, is the lag order, for The lag coefficient matrix of order, is white noise; The operating data includes voltage , Current , Power Factor And harmonic distortion , the environmental data also includes temperature ,humidity .

5. The remote control method of the metering box according to claim 4, characterized in that: The local model parameters of the weighted aggregation nodes include: ,in for The local model parameters of each node, For the The amount of data per node, .

6. The remote control method of the metering box according to claim 5, characterized in that: The method for generating the anomaly confidence score includes: Define the fuzzy membership function , the Mahalanobis distance Mapped to confidence values: In the formula, is the maximum Mahalanobis distance of historical normal data, is the maximum Mahalanobis distance of historical fault data; Fusion of multi-source evidence weights , generating a composite confidence score ,in is the total number of evidence types, For the Confidence rating of the evidence.

7. The remote control method of the metering box according to claim 6, characterized in that: Based on the topological connection relationship matrix and confidence score Compute Node Maintenance priorities: ,in Representation Node and The connection status, For Node Confidence score of .

8. The remote control method of the metering box according to claim 7, characterized in that: The dynamic adjustment of the update frequency of the global model includes: calculating the environmental change factor , quantifying the mutation rate of environmental parameters: , where and They are the instantaneous changes of temperature and humidity, respectively, and are mapped to the model update frequency weight through the Sigmoid function ,when When it is greater than the preset threshold, real-time fine-tuning of the global model is triggered.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the meter box remote control method described in any one of claims 1 to 8 is implemented.

10. A metering box, characterized in that: The meter box includes at least one lower-level meter box to form a meter box cluster, and the meter box cluster applies the meter box remote control method according to any one of claims 1 to 8.

Images