Method for automatically acquiring and processing information of electric energy meter

Abstract

The invention discloses an electric energy meter information automatic acquisition and processing method, which comprises the following steps: acquiring the specific position and posture of an electric energy meter in an electric energy meter bin, acquiring the current position information and movement speed of a mechanical arm and the current applied force of a clamping jaw in real time, and inputting the acquired data into a multi-stage control algorithm, a mechanical arm control system is optimized through a multi-level objective function, and optimal control input is obtained; the feature vector of the electric energy meter data is partitioned according to time periods, the difference of adjacent data blocks is calculated, differential data is compressed and coded by using a self-adaptive differential compression algorithm and uploaded to a cloud data center through a wireless communication module, and the original data blocks are decompressed and recovered; and performing real-time analysis on the recovered original data block through a dynamic support vector regression algorithm, extracting key features, performing classification judgment, generating an operation state report of the electric energy meter, detecting abnormal conditions and giving an alarm. According to the invention, centralized management and intelligent analysis of data are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of data collection and processing of electric power systems, and in particular to an automatic collection and processing method for electric energy meter information. Background Art

[0002] With the rapid development of smart grids, higher requirements are placed on the collection, processing and management of electricity meter data. The traditional method of collecting electricity meter data mainly relies on manual meter reading, which is not only inefficient but also easily affected by human factors, resulting in poor data accuracy and consistency. In order to meet the development needs of smart grids, a method that can automatically, efficiently and accurately collect and process electricity meter data is urgently needed.

[0003] The development trend of smart grids has put forward higher requirements for the real-time and accuracy of electricity meter data. Real-time data is crucial for the timeliness and responsiveness of power dispatch and management. At the same time, the surge in the number of power users and the diversification of power-consuming devices have greatly increased the amount and complexity of electricity meter data collection. The traditional manual meter reading method can no longer meet the needs of modern power systems. In addition, as the power system gradually transforms towards digitalization and intelligence, it is necessary to conduct in-depth analysis and intelligent processing of electricity meter data to achieve accurate analysis of power users' electricity consumption behavior and optimal allocation of power resources.

[0004] The existing technology has at least the following technical problems: the lack of intelligent data processing means makes the existing system unable to cope with the complex and dynamically changing power system environment; the lack of advanced machine learning and intelligent analysis algorithms makes it impossible to detect and handle complex abnormal situations in a timely manner, which easily leads to missed faults or false alarms, affecting the stable operation of the power system; it is impossible to achieve full-process automated operation, requiring a lot of manual intervention, and reducing the overall efficiency and intelligence level of the system. Summary of the invention

[0005] In order to solve the defects of the above-mentioned prior art, the present invention provides a method for automatically collecting and processing electric energy meter information to achieve centralized management and intelligent analysis of data and improve the stability and safety of the power system.

[0006] To this end, the present invention adopts the following technical solution: a method for automatically collecting and processing electric energy meter information, which comprises the steps of: S1. Obtain the specific position and posture of the electric energy meter in the electric energy meter bin, obtain the current position information and movement speed of the robot arm and the current force applied by the gripper in real time, input the above-mentioned acquired data into the multi-level control algorithm, optimize the robot arm control system through the multi-level objective function, and obtain the optimal control input; S2, after the robot arm grabs and places the electric energy meter according to the optimal control input, the sensor starts data collection, and the time series and spatial data are modeled and analyzed using the spatiotemporal recursive network algorithm to obtain the feature vector of the electric energy meter data; S3, dividing the characteristic vector of the electric energy meter data into blocks according to time periods, calculating the difference between adjacent data blocks, compressing and encoding the differential data using an adaptive differential compression algorithm, and uploading it to a cloud data center through a wireless communication module, where the cloud data center decompresses and restores the original data blocks; S4. Perform real-time analysis on the recovered original data blocks through the dynamic support vector regression algorithm, extract key features and make classification judgments, generate an operation status report of the electric energy meter, detect abnormal conditions and issue an alarm.

[0007] Furthermore, the S1 specifically includes: The electric energy meter bin, gripper and conveyor are integrated to form an automated electric energy meter grabbing and assembly system; the electric energy meter bin is used to store the electric energy meters to be processed, and the gripper grabs the electric energy meter from the electric energy meter bin and places it on the conveyor through the precise control of the robotic arm. The conveyor is responsible for transporting the electric energy meter to the data collection location; The specific position and posture of the electric energy meter in the electric energy meter compartment are obtained through the camera installed in the electric energy meter compartment and the image recognition algorithm. The current position information and movement speed of the robotic arm are obtained in real time through the position sensor and speed sensor on the robotic arm. The current force applied by the gripper is obtained in real time through the force sensor on the gripper, and the target value of the position and posture where the electric energy meter needs to be placed is preset; Construct a multi-level control algorithm, input the aforementioned acquired and preset data into the multi-level control algorithm, define a multi-level objective function, and optimize the motion path of the robot arm and the gripping force of the gripper through the multi-level objective function.

[0008] Furthermore, in S1, the definition process of the multi-level objective function is: First, the state space of the robotic arm control system for automated energy meter grabbing and data collection is defined, including the position, velocity, and gripper force of the robotic arm. At each time t, the current state of the robotic arm control system is expressed as ,in, is the current position vector of the robot at time t, is the current velocity vector of the robot at time t, is the current force vector of the gripper at time t; Next, define the possible actions in each state, that is, the current control input is expressed as ,in, is the robot position control input, is the robot speed control input, It is the input for gripper force control; To measure the cost of each action, define the stage cost function at each time t , the stage cost function is:

[0009] in, is the target position vector, i.e. the position the robot should reach, is the target velocity vector, i.e. the velocity that the robot should reach, is the target force vector, i.e. the force the gripper should reach, , , is the weight coefficient, is the Euclidean distance; define the state transfer equation of the robotic arm control system, that is, from the current state Through the current control input Transition to next state , the state transfer equation is:

[0010] in, and is the state transfer matrix, Control noise for the robotic arm system; Furthermore, we define the expected value of the cumulative cost from the current state to the target state, namely the value function , the value function is expressed as:

[0011] The value function is obtained by recursive calculation:

[0012] in, is the expected value of the total cost from time t to the target state, i is the index of the time step, from 1 to , is the total time length of the control process, and The robot control system is The state and control input at each moment.

[0013] Furthermore, in S1, the value function and the optimal control input are recursively calculated from back to front, and finally the optimal control input for each time step is obtained to guide the motion path of the robot arm and the gripping force of the gripper.

[0014] Furthermore, in S2, the data collected by the sensor is used as time series data, and the current state of the robot control system is As spatial data, the time series data and spatial data are normalized to ensure that the data are within a uniform scale and are processed in segments. On the basis of the spatiotemporal recursive neural network, spatial convolution layers and temporal convolution layers are added to extract spatiotemporal features and obtain the feature vector of the electric energy meter data.

[0015] Furthermore, in S2, the specific implementation process of the spatiotemporal recurrent neural network is: At each time step t, the current state of the input Perform convolution operation to extract spatial features:

[0016] in, is the spatial feature at time t, extracted by the spatial convolutional layer; represents the convolution process, and They are the weight matrix and bias vector of the spatial convolutional layer respectively; For input time series data Perform convolution operation to extract temporal features:

[0017] in, It is the temporal feature at time t, extracted by the temporal convolutional layer; and They are the weight matrix and bias vector of the temporal convolutional layer respectively; Combine the spatiotemporal features and perform recursive calculations to obtain the hidden state :

[0018] in, is the hidden layer state, which represents the internal state of the spatiotemporal recurrent neural network model at time t; is the weight ratio of time series data; is the weight ratio of the hidden layer state at the previous moment; is the bias parameter in the recursive process; for the hidden state Calculate the activation function and get , the activation function is used to increase nonlinearity, and the output layer is calculated as:

[0019] in, It is the output of the spatiotemporal recurrent neural network, representing the feature vector after the electric energy meter data is processed, including the real-time operating status, historical features, spatial features and comprehensive features of the electric energy meter; is the activation function, , are the output layer weights and bias parameters respectively; the weight distribution is optimized through the attention mechanism:

[0020] in, is the output weight of the attention mechanism, is the weight parameter of the attention mechanism.

[0021] Furthermore, in S3, a fixed time period is determined , as the time interval for block division; initialize an empty data structure to store data blocks of different time periods. Each data block contains two parts: total weight and weighted cumulative value; from the start time of the time series data Start by traversing each time step until the last time step. At each time step, set the current time step The output of the spatiotemporal recurrent neural network and the output weights of the attention mechanism According to their timestamps, they are assigned to corresponding time period blocks, and the data in each time period block is weighted averaged to obtain representative data for the current time period.

[0022] Furthermore, in S3, the difference calculation result of adjacent data blocks is As input for data compression, the differential data is compressed and encoded:

[0023] in, is the kth block of differential data vector The compressed encoding length, is the probability of the jth symbol appearing in Huffman coding, is the total number of symbols.

[0024] Furthermore, in the S3, the cloud data center receives the compressed data After that, first decompress it to restore the original differential data ; Then, restore the original data block by accumulating , the restored original data block It is transmitted as input to the next step and analyzed in real time by a dynamic support vector regression algorithm.

[0025] Furthermore, the S4 specifically includes: Extract key features from the restored original data blocks, use the dynamic support vector regression algorithm to train the regression model, and obtain the classification results of the feature vector; generate an operation status report of the electric energy meter based on the output of the regression model; for detected abnormal operation reports, issue an alarm and send the abnormal operation report to the operation and maintenance personnel.

[0026] The present invention has the following beneficial effects: 1. Combine multi-level control algorithms and precise robotic arm control to achieve efficient and accurate grasping and placement of electricity meters.

[0027] 2. The collected data is processed through the spatiotemporal recursive network algorithm, and the data is modeled and analyzed in combination with the time and space dimensions to extract the real-time operating status and historical characteristics of the electric energy meter; the data is divided and compressed through the adaptive differential compression algorithm to improve the efficiency of data transmission. The cloud data center further analyzes and processes the uploaded data, provides remote monitoring and management functions, and realizes centralized management and intelligent analysis of data.

[0028] 3. Use the dynamic support vector regression algorithm to analyze data in real time, detect abnormal conditions of the electricity meter in a timely manner, issue alarms and generate abnormal reports. Operation and maintenance personnel can perform troubleshooting and repairs based on the reports, improving the stability and safety of the power system; all operating status reports and abnormal operation reports are stored in the cloud database for subsequent analysis and historical record query, providing reliable data support for the long-term maintenance and optimization of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 The present invention is a flow chart of a method for automatically collecting and processing electric energy meter information. DETAILED DESCRIPTION

[0030] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the technical scheme in the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is only a part of the embodiment of the present invention, not all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0031] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0032] The specific scheme of the method for automatic collection and processing of electric energy meter information provided by the present invention is described in detail below with reference to the accompanying drawings.

[0033] Refer to the attached Figure 1 , which shows a flow chart of an automatic collection and processing method of electric energy meter information provided by an embodiment of the present invention, the method comprising the following steps: S1. Obtain the specific position and posture of the electric energy meter in the electric energy meter bin, obtain the current position information and movement speed of the robot arm and the current force applied by the gripper in real time, input the aforementioned acquired data into the multi-level control algorithm, and optimize the robot arm control system through the multi-level objective function to obtain the optimal control input.

[0034] The energy meter bin, gripper and conveyor are integrated to form an automated energy meter grabbing and assembly system. The energy meter bin is used to store the energy meters to be processed. The gripper grabs the energy meter from the bin and places it on the conveyor through the precise control of the robotic arm. The conveyor is responsible for transporting the energy meter to the data collection location.

[0035] The specific position and posture of the electricity meter in the electricity meter compartment are obtained through the camera and image recognition algorithm installed in the electricity meter compartment, the current position information and movement speed of the robotic arm are obtained in real time through the position sensor and speed sensor on the robotic arm, the current force applied by the gripper is obtained in real time through the force sensor on the gripper, and the target value of the position and posture where the electricity meter needs to be placed is preset.

[0036] Construct a multi-level control algorithm, input the above acquired and preset data into the multi-level control algorithm, and optimize the motion path of the robot arm and the gripping force of the gripper through the multi-level objective function by defining the multi-level objective function.

[0037] The definition process of the multi-level objective function is: First, the state space of the robotic arm control system for automated energy meter grabbing and data collection is defined, including the position, velocity, and force of the gripper. At each time t, the current state of the robotic arm control system is expressed as ,in, is the current position vector of the robot at time t, is the current velocity vector of the robot at time t, is the current force vector of the gripper at time t; next, the possible actions in each state are defined, that is, the current control input is expressed as ,in is the position control input, is the speed control input, Input for velocity control.

[0038] In order to measure the cost of each action, the stage cost function at each time t is defined The stage cost function is:

[0039] in, is the target position vector, i.e. the position the robot should reach, is the scalar velocity vector, i.e. the velocity that the robot should reach, is the target force vector, i.e. the force the gripper should reach, , , is the weight coefficient, is the Euclidean distance. Define the state transfer equation of the system, that is, from the current state Through the current control input Transition to next state The state transfer equation is:

[0040] in, and is the state transfer matrix, Noise of the robotic arm control system.

[0041] Furthermore, we define the expected value of the cumulative cost from the current state to the target state, namely the value function The value function is expressed as:

[0042] The value function is obtained by recursive calculation:

[0043] in, is the expected value of the total cost from time t to the target state, i is the index of the time step, from 1 to , is the total time length of the control process, and Respectively in The state of the system and the control input at each moment.

[0044] Furthermore, by recursively calculating the value function from back to front, the optimal control strategy is gradually solved. The specific steps are as follows: Initialize the endpoint value function , from time arrive Recursive calculation function and the optimal control input, and finally the optimal control input for each time step is obtained, which is used to guide the motion path of the robot arm and the grasping force of the gripper.

[0045] In the actual operation process, the parameters of the state space and action space are first initialized, including the current position, speed and gripper strength of the robot. Then, the cost of each action is measured by defining the stage cost function to ensure that each action is optimal. At each time step, the system state transfer process is calculated according to the state transfer equation, that is, the state at the next moment is updated by combining the state transfer matrix with the control input. At the same time, the recursive calculation of the cumulative cost function ensures the overall optimality from the current state to the target state.

[0046] Through the recursive process of the dynamic programming algorithm, the optimal control input is calculated at each time step. For example, in actual operation, when the robot arm is at a certain position, the control input is calculated to determine how to adjust the robot arm's motion path to make it closer to the target position; at the same time, the robot arm's speed and the gripper's strength are adjusted to ensure that the gripper can firmly grasp and accurately place the energy meter.

[0047] S2. After the robot arm grabs and places the electric energy meter according to the optimal control input, the sensor starts data collection, and the time series and spatial data are modeled and analyzed using the spatiotemporal recursive network algorithm to obtain the feature vector of the electric energy meter data.

[0048] When the robot arm completes the grabbing and placement of the energy meter according to the optimal control input, the sensor starts to collect data. This ensures that the energy meter is accurately placed at the data collection position, so that the sensor can accurately align with the energy meter and collect data.

[0049] The data collected by the sensor is preprocessed, denoised and normalized. Then, the data is further processed using the spatiotemporal recursive network algorithm, which combines the time and space dimensions to model and analyze the data, improving the accuracy and efficiency of data processing.

[0050] Specifically, the data collected by the sensor is used as time series data, and the state of the system is As spatial data, the time series data and spatial data are normalized to ensure that the data is within a unified scale range, and segmented processing is performed. On the basis of the spatiotemporal recurrent neural network, spatial convolution layers and temporal convolution layers are added to extract spatiotemporal features. The specific implementation process of the spatiotemporal recurrent neural network is as follows: At each time step t, the input spatial data Perform convolution operation to extract spatial features:

[0051] in, is the spatial feature at time t, extracted by the spatial convolution layer, represents the convolution process, and are the weight matrix and bias vector of the spatial convolutional layer, respectively.

[0052] For input time series data Perform convolution operation to extract temporal features:

[0053] in, is the temporal feature at time t, extracted by the temporal convolutional layer, and are the weight matrix and bias vector of the temporal convolutional layer, respectively.

[0054] Combine the spatiotemporal features and perform recursive calculations to obtain the hidden state :

[0055] in, is the hidden layer state, which represents the internal state of the spatiotemporal recurrent neural network model at time t. is the weight ratio of time series data; is the weight ratio of the hidden layer state at the previous moment; is the bias parameter in the recursive process; for the hidden state Perform activation function calculation and get output , the activation function is used to increase nonlinearity, and the output layer is calculated as:

[0056] in, It is the output of the spatiotemporal recurrent neural network, representing the feature vector after the electric energy meter data is processed, including the real-time operating status, historical features, spatial features, comprehensive features, etc. of the electric energy meter. is the activation function, , are the output layer weights and bias parameters. The weight distribution is optimized through the attention mechanism:

[0057] in, is the output weight of the attention mechanism, is the weight parameter of the attention mechanism.

[0058] Processed output data and The processed output data will be automatically uploaded to the cloud data center through the wireless communication module as the input for the next step of data upload and management.

[0059] S3. Divide the characteristic vector of the electric energy meter data into blocks according to time periods, calculate the differences between adjacent data blocks, compress and encode the differential data using an adaptive differential compression algorithm, and upload it to the cloud data center through a wireless communication module. The cloud data center decompresses and restores the original data blocks.

[0060] The cloud data center further analyzes and processes the uploaded data and provides remote monitoring and management functions. The data upload process uses an improved adaptive differential compression algorithm to improve data transmission efficiency.

[0061] Specifically, determine a fixed time period , as the time interval for block division. For example, you can set every minute or every hour as a time period. Initialize an empty data structure to store data blocks of different time periods. Each data block contains two parts: total weight and weighted cumulative value.

[0062] From the start time of the time series data Start, time step by time Traverse until the last time step, at each time step, the current time step Output data and Assign them to the corresponding time period blocks according to their timestamps.

[0063] The specific process is as follows: First, calculate the current time step Relative to start time Time difference , through the time difference and pre-set time periods Calculate the time period block to which the current data belongs, that is:

[0064] The data and Add to the corresponding time period block First, update the time period block Total weight :

[0065] Next, update the time period block The weighted cumulative value of :

[0066] After all the data of all time steps are assigned to the corresponding time period blocks, the data in each time period block is weighted averaged to obtain the representative data of the current time period. , calculate the weighted average :

[0067] Thus, the output data and weight According to the scheduled time period Divide into blocks and pass weights Perform weighted averaging on the data, and each piece of data is represented as .

[0068] As a specific example, if the time period If it is set to one minute, then when traversing the data, each one-minute time period will be assigned a data block. , if the timestamp is seconds, then the data point will be assigned to the first time period block Assume that the output data and weight are [5,10] and [0.5,0.5] respectively, then the total weight of the first time period block is updated to , the weighted cumulative value is updated to .

[0069] At each time step, the time difference is calculated and the time period block of the data point is determined, and then the total weight and weighted cumulative value of the corresponding block are updated. When the traversal reaches the end of the first time period, for example seconds, the weighted average of the blocks in this time period is calculated as , which is the weighted average of the original data points.

[0070] Compute the differences between adjacent blocks of data:

[0071] in, is the kth block of differential data vector. The result of differential calculation is As input for data compression, the differential data is compressed and encoded:

[0072] in, is the kth block of differential data vector The compressed encoding length, is the probability of the jth symbol appearing in Huffman coding, is the total number of symbols.

[0073] The cloud data center receives the compressed data After that, first decompress it to restore the original differential data ; Then, restore the original data block by accumulating . Recovered original data blocks It is transmitted as input to the next step and analyzed in real time by a dynamic support vector regression algorithm.

[0074] S4. Perform real-time analysis on the recovered original data blocks through the dynamic support vector regression algorithm, extract key features and make classification judgments, generate an operation status report of the electric energy meter, detect abnormal conditions and issue an alarm.

[0075] The dynamic support vector regression algorithm generates a report on the operation status of the electric energy meter and issues an alarm when an abnormal situation is detected. First, key features are extracted from the original data block. The feature extraction formula is:

[0076] in, From the original data block The features extracted from , , , is the feature extraction weight coefficient, is the number of block features, It is a data block The dynamic support vector regression algorithm is used to train the regression model. The training process formula is:

[0077] in, is the output of the regression model, which represents the feature vector The classification result is used to determine whether the operation status of the electric energy meter is normal. is the Lagrange multiplier of the i-th training sample, is the label of the i-th training sample, is the kernel function used to calculate the eigenvector and The similarity between is the feature vector of the i-th training sample, is the bias vector of the regression model. The kernel function is defined as:

[0078] in, , These are parameters in the kernel function, which control the width and activation function of the Gaussian kernel.

[0079] According to the output of the regression model , generate the operation status report of the energy meter. If the output is positive, generate a normal operation report; if the output is negative, generate an abnormal operation report. For the detected abnormal situation, the system will issue an alarm and send the abnormal operation report to the operation and maintenance personnel. The operation and maintenance personnel can perform troubleshooting and repair according to the detailed information in the abnormal report (such as abnormal characteristics, time point, etc.).

[0080] The above embodiments have the following beneficial effects: 1. By integrating the electric energy meter bin, gripper and conveying device, an automated electric energy meter grabbing and assembly system is formed. Combined with multi-level control algorithms and precise robotic arm control, efficient and accurate grabbing and placement of electric energy meters can be achieved. The specific position and posture of the electric energy meter in the bin are obtained through the camera installed in the bin and the image recognition algorithm, and the position sensor, speed sensor on the robotic arm and the force sensor on the gripper are used to obtain relevant information in real time to ensure the accuracy of data collection position and posture. 2. The collected data is processed through the spatiotemporal recursive network algorithm, and the data is modeled and analyzed in combination with the time and space dimensions to extract the real-time operating status and historical characteristics of the electric energy meter; the dynamic support vector regression algorithm is further combined to classify and detect anomalies in the data, generate an electric energy meter operating status report, monitor the operating status of the electric energy meter in real time, and promptly detect and handle abnormal situations; the data is divided into blocks and compressed through the improved adaptive differential compression algorithm, which improves the efficiency of data transmission. The cloud data center further analyzes and processes the uploaded data, provides remote monitoring and management functions, and realizes centralized management and intelligent analysis of data; 3. Use the dynamic support vector regression algorithm to analyze data in real time, detect abnormal conditions of the electricity meter in a timely manner, issue alarms and generate abnormal reports. Operation and maintenance personnel can perform troubleshooting and repairs based on the reports, improving the stability and safety of the power system; all operating status reports and abnormal reports are stored in the cloud database for subsequent analysis and historical record query, providing reliable data support for the long-term maintenance and optimization of the power system.

[0081] The sequence of the embodiments of the present invention is for description only and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0082] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

[0083] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention.

Claims

1. A method for automatic collection and processing of electric energy meter information, characterized in that: include: S1. Obtain the specific position and posture of the electric energy meter in the electric energy meter bin, obtain the current position information and movement speed of the robot arm and the current force applied by the gripper in real time, input the above-mentioned acquired data into the multi-level control algorithm, optimize the robot arm control system through the multi-level objective function, and obtain the optimal control input; S2, after the robot arm grabs and places the electric energy meter according to the optimal control input, the sensor starts data collection, and the time series and spatial data are modeled and analyzed using the spatiotemporal recursive network algorithm to obtain the feature vector of the electric energy meter data; S3, dividing the characteristic vector of the electric energy meter data into blocks according to time periods, calculating the difference between adjacent data blocks, compressing and encoding the differential data using an adaptive differential compression algorithm, and uploading it to a cloud data center through a wireless communication module, where the cloud data center decompresses and restores the original data blocks; S4. Perform real-time analysis on the recovered original data blocks through the dynamic support vector regression algorithm, extract key features and make classification judgments, generate an operation status report of the electric energy meter, detect abnormal conditions and issue an alarm.

2. The method for automatic collection and processing of electric energy meter information according to claim 1, characterized in that: The S1 specifically includes: The electric energy meter bin, gripper and conveyor are integrated to form an automated electric energy meter grabbing and assembly system; the electric energy meter bin is used to store the electric energy meters to be processed, and the gripper grabs the electric energy meter from the electric energy meter bin and places it on the conveyor through the precise control of the robotic arm. The conveyor is responsible for transporting the electric energy meter to the data collection location; The specific position and posture of the electric energy meter in the electric energy meter compartment are obtained through the camera installed in the electric energy meter compartment and the image recognition algorithm. The current position information and movement speed of the robotic arm are obtained in real time through the position sensor and speed sensor on the robotic arm. The current force applied by the gripper is obtained in real time through the force sensor on the gripper, and the target value of the position and posture where the electric energy meter needs to be placed is preset; Construct a multi-level control algorithm, input the aforementioned acquired and preset data into the multi-level control algorithm, define a multi-level objective function, and optimize the motion path of the robot arm and the gripping force of the gripper through the multi-level objective function.

3. The method for automatic collection and processing of electric energy meter information according to claim 2, characterized in that: In S1, the definition process of the multi-level objective function is: First, the state space of the robotic arm control system for automated energy meter grabbing and data collection is defined, including the position, velocity, and gripper force of the robotic arm. At each time t, the current state of the robotic arm control system is expressed as ,in, is the current position vector of the robot at time t, is the current velocity vector of the robot at time t, is the current force vector of the gripper at time t; Next, define the possible actions in each state, that is, the current control input is expressed as ,in, is the robot position control input, is the robot speed control input, It is the input for gripper force control; To measure the cost of each action, define the stage cost function at each time t , the stage cost function is: ， in, is the target position vector, i.e. the position the robot should reach, is the target velocity vector, i.e. the velocity that the robot should reach, is the target force vector, i.e. the force the gripper should reach, , , is the weight coefficient, is the Euclidean distance; define the state transfer equation of the robotic arm control system, that is, from the current state Through the current control input Transition to next state , the state transfer equation is: ， in, and is the state transfer matrix, Control noise for the robotic arm system; Furthermore, we define the expected value of the cumulative cost from the current state to the target state, namely the value function , the value function is expressed as: ， The value function is obtained by recursive calculation: ， in, is the expected value of the total cost from time t to the target state, i is the index of the time step, from 1 to , is the total time length of the control process, and The robot control system is The state and control input at each moment.

4. The method for automatic collection and processing of electric energy meter information according to claim 3 is characterized in that: In S1, the value function and the optimal control input are recursively calculated from back to front, and finally the optimal control input for each time step is obtained to guide the motion path of the robot arm and the gripping force of the gripper.

5. The method for automatic collection and processing of electric energy meter information according to claim 1, characterized in that: In S2, the data collected by the sensor is used as time series data, and the current state of the robot control system is used as the time series data. As spatial data, the time series data and spatial data are normalized to ensure that the data are within a uniform scale and are processed in segments. On the basis of the spatiotemporal recursive neural network, spatial convolution layers and temporal convolution layers are added to extract spatiotemporal features and obtain the feature vector of the electric energy meter data.

6. The method for automatic collection and processing of electric energy meter information according to claim 1, characterized in that: In S2, the specific implementation process of the spatiotemporal recurrent neural network is: At each time step t, the current state of the input Perform convolution operation to extract spatial features: ， in, is the spatial feature at time t, extracted by the spatial convolutional layer; represents the convolution process, and They are the weight matrix and bias vector of the spatial convolutional layer respectively; For input time series data Perform convolution operation to extract temporal features: ， in, It is the temporal feature at time t, extracted by the temporal convolutional layer; and They are the weight matrix and bias vector of the temporal convolutional layer respectively; Combine the spatiotemporal features and perform recursive calculations to obtain the hidden state : ， in, is the hidden layer state, which represents the internal state of the spatiotemporal recurrent neural network model at time t; is the weight ratio of time series data; is the weight ratio of the hidden layer state at the previous moment; is the bias parameter in the recursive process; for the hidden state Calculate the activation function and get , the activation function is used to increase nonlinearity, and the output layer is calculated as: ， in, It is the output of the spatiotemporal recurrent neural network, representing the feature vector after the electric energy meter data is processed, including the real-time operating status, historical features, spatial features and comprehensive features of the electric energy meter; is the activation function, , are the output layer weights and bias parameters respectively; the weight distribution is optimized through the attention mechanism: ， in, is the output weight of the attention mechanism, is the weight parameter of the attention mechanism.

7. The method for automatic collection and processing of electric energy meter information according to claim 6, characterized in that: In S3, a fixed time period is determined , as the time interval for block division; initialize an empty data structure to store data blocks of different time periods. Each data block contains two parts: total weight and weighted cumulative value; from the start time of the time series data Start by traversing each time step until the last time step. At each time step, set the current time step The output of the spatiotemporal recurrent neural network and the output weights of the attention mechanism According to the timestamp, the data is assigned to the corresponding time period block, and the weighted average calculation is performed on the data in each time period block to obtain the representative data of the current time period.

8. The method for automatic collection and processing of electric energy meter information according to claim 7, characterized in that: In S3, the difference calculation result of adjacent data blocks As input for data compression, the differential data is compressed and encoded: ， in, is the kth block of differential data vector The compressed encoding length, is the probability of the jth symbol appearing in Huffman coding, is the total number of symbols.

9. The method for automatic collection and processing of electric energy meter information according to claim 8, characterized in that: In the S3, the cloud data center receives the compressed data After that, first decompress it to restore the original differential data ; Then, restore the original data block by accumulating , the restored original data block It is transmitted as input to the next step and analyzed in real time by a dynamic support vector regression algorithm.

10. The method for automatic collection and processing of electric energy meter information according to claim 1, characterized in that: The S4 specifically includes: Key features are extracted from the restored original data blocks, and the regression model is trained using the dynamic support vector regression algorithm to obtain the classification results of the feature vector; based on the output of the regression model, an operation status report of the electric energy meter is generated; for any abnormal operation report detected, an alarm is issued and the abnormal operation report is sent to the operation and maintenance personnel.

Images