A data correction method and system based on centralized meter reading

Abstract

The present invention relates to the technical field of data correction for meter reading, and discloses a data correction method and system based on centralized meter reading, comprising: collecting different types of meter data through meter reading equipment, and transmitting the meter reading data to a data management platform for centralized meter reading through a collection terminal; making a preliminary abnormality judgment on the meter reading data, and initially screening the data that needs to be corrected; performing a secondary screening of the data that needs to be corrected through a correlation analysis of the data after the initial screening; and using an improved annealing algorithm to perform data correction in the results of the secondary screening. It can effectively reduce manual operations, improve efficiency and data real-time performance. It improves the accuracy of preliminary abnormality detection, and reduces false positives and false negatives. Comprehensively consider the complex correlations between data to further improve the screening accuracy. Through a global optimization strategy, accurate correction of abnormal data is achieved to ensure the reliability and stability of data.

Description

Technical Field

[0001] The present invention relates to the technical field of meter reading data correction, and in particular to a data correction method and system based on centralized meter reading. Background Art

[0002] With the rapid development of smart grid and Internet of Things technologies, data collection and management systems based on centralized meter reading have been widely used in public utilities such as electricity, water, gas and heat. The traditional manual meter reading method has problems such as low efficiency, large errors and high labor costs. The centralized meter reading system collects various meter data through automated means, which can significantly improve meter reading efficiency and data accuracy. Modern centralized meter reading systems usually integrate various types of meter data, including electricity meters, water meters, gas meters and heat meters, and transmit the data to the centralized management platform through the collection terminal for unified processing and analysis. This automated data collection method not only reduces manual intervention, but also improves the real-time and reliability of data, providing important support for the operation and management of public utilities.

[0003] Although the existing centralized meter reading system has made significant progress in data collection and management, it still has many shortcomings in data anomaly detection and correction. First, the traditional statistical analysis method has limited ability to detect data anomalies, especially when faced with complex nonlinear data relationships, which are prone to omissions and false positives. Second, although machine learning and neural network technologies have improved the accuracy of data anomaly detection to a certain extent, these methods usually require a large amount of labeled data for training, and have high requirements for the selection and optimization of model parameters, which increases the complexity of the system and the operation and maintenance costs.

[0004] In addition, most existing abnormal data correction methods focus on the correction of individual data points, lacking comprehensive consideration of the correlation between data. For example, there may be complex correlations between different types of meter data such as electricity meters, water meters, and gas meters. These correlations are often ignored in abnormal data detection and correction, resulting in unsatisfactory correction results. On the other hand, existing methods do not make full use of historical data and cannot effectively use the historical fluctuation range as a correction reference, resulting in a lack of reasonable evaluation criteria and optimization goals in the correction process. Summary of the invention

[0005] In view of the above-mentioned problems, the present invention is proposed.

[0006] Therefore, the technical problems solved by the present invention include: the existing data correction method for meter reading has the following problems: the labor cost is too high, the time consumption is long, the historical data is not fully utilized, etc.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions: a data correction method based on centralized meter reading, comprising:

[0008] Through meter reading equipment, different types of meter data are collected, and the meter reading data is transmitted to the centralized meter reading data management platform through the collection terminal;

[0009] Performing preliminary abnormality judgment on the meter reading data and initially screening the data that needs to be corrected;

[0010] Through the correlation analysis of the data after the initial screening, the secondary screening of the data that needs to be corrected is carried out;

[0011] In the results of the secondary screening, the improved annealing algorithm is used for data correction.

[0012] As a preferred solution of the data correction method based on centralized meter reading described in the present invention, wherein: the different types of meter data include electricity meter data, water meter data, gas meter data, and heat meter data;

[0013] The collection terminal integrates the meter reading data collected by all meter reading devices, inserts three tags into each meter reading data, and transmits the data integrated by the collection terminal to the data management platform through the communication network;

[0014] The three tags include: location tag, user tag, and device tag;

[0015] The location tag includes the geographical location information of the meter, and the location information of the meter can be queried through the Internet to obtain environmental influencing factors;

[0016] The user tag includes, through the matching relationship table between each meter and the user, recording the user matched by the current meter when recording the meter reading data;

[0017] The device tag includes device information for recording current meter data.

[0018] As a preferred solution of the data correction method based on centralized meter reading described in the present invention, wherein: the preliminary abnormality judgment includes classifying the integrated meter reading data according to the user tags to obtain the meter reading data of each user;

[0019] In the meter reading data of each user, abnormality judgment is performed on each meter as a unit, and the abnormal part is identified by analyzing the historical records of meter reading data;

[0020] Features users, the data of each user i is express;

[0021] For each user's data , extract the data of each meter and calculate its change. Suppose the data of each meter is ;

[0022] The formula for calculating the change is:

[0023] ,

[0024] in, represents the change in meter reading data of the jth meter of the i-th user, Indicates User's Meters at time The readings, Indicates the time interval between meter readings. Indicates A meter, Indicates time;

[0025] Calculate the characteristic values ​​of the variation data, including mean, standard deviation and Fourier transform;

[0026] ,

[0027] ,

[0028] ,

[0029] in, represents the average value of the change in the jth meter of the ith user, T represents the time period of the meter reading data, It is expressed as the standard deviation of the change in the jth meter of the ith user, represents the Fourier transform result of the change of the jth meter of the i-th user, ω represents the frequency variable;

[0030] Anomaly detection is performed using a multilayer perceptron neural network with the input being a feature vector , the output is the anomaly score ;

[0031] Eigenvectors:

[0032] ,

[0033] Neural network formula:

[0034] ,

[0035] in, and Represents the weight matrix of the neural network; and Represents the bias vector of the neural network; ReLU is the activation function; Represents the Sigmoid activation function, which converts the output into a probability value between 0 and 1;

[0036] For each meter of each user, if the anomaly score Exceeding the preset threshold , it is judged as abnormal; the data judged as abnormal is determined as the data that needs to be corrected in the initial screening.

[0037] As a preferred solution of the data correction method based on centralized meter reading described in the present invention, wherein: the association analysis includes obtaining each data that needs to be corrected, and establishing a graph neural network according to the position label of each data; the node is each data that needs to be corrected, and the edge is the distance relationship between the positions; according to the position of the abnormal data in the historical records, the DBSCAN algorithm is used to divide multiple historical abnormal areas, each area contains multiple position nodes; by analyzing whether the location where the abnormality occurs is concentrated in the same divided area, if it is concentrated in the same divided area, it is determined to be an abnormality in the area; if it is not concentrated in the same divided area, it is determined that there is no abnormality in the area;

[0038] Specifically:

[0039] Use graph neural networks to calculate the anomaly score of each node and calculate the anomaly probability of each region based on the score;

[0040] Use graph convolutional layers for feature propagation:

[0041] ,

[0042] Calculate the anomaly score for each node:

[0043] ,

[0044] Calculate the anomaly probability for each region:

[0045] ,

[0046] in, Represents the node feature matrix of the lth layer; Represents the node feature matrix of the l+1th layer; represents the nonlinear activation function Sigmoid; D is the degree matrix, which represents the degree of each node in the graph, which is the number of edges connected to the node; represents the square root inverse matrix of the degree matrix, which is used for normalization processing; A is the adjacency matrix, which represents the connection relationship between nodes in the graph; W is the weight matrix, which represents the parameters of the linear transformation; represents the weight matrix of the lth layer; represents the anomaly score of node i; represents the set of neighbor nodes of node i; j represents the index of the neighbor node of node i; Indicates the number of neighbor nodes of node i; Represents the feature vector of node i; represents the feature vector of node j, the neighbor of node i; σ represents the bandwidth parameter of the Gaussian kernel function; Indicates area The probability of abnormality; It is represented as the number of nodes in the kth region; i represents the index of the node;

[0047] Based on the calculated regional anomaly probability, determine whether the current anomaly is concentrated in a certain historical anomaly area.

[0048] As a preferred solution of the data correction method based on centralized meter reading described in the present invention, the secondary screening includes setting a threshold for abnormal nodes in the area if the current abnormal location is concentrated in the historical area. Make a judgment, when , it is judged as abnormal in the region, otherwise it is judged as no abnormality in the region;

[0049] The data to be corrected are classified according to the device labels, and the data abnormality ratio in each device label is analyzed. If the data abnormality ratio exceeds a preset threshold value YC, it is determined that the device is abnormal; otherwise, it is determined that there is no device abnormality;

[0050] The data judged as the device abnormality and the abnormality in the region are taken as the result of the secondary screening and corrected by the improved annealing algorithm; the data judged as having no device abnormality and no abnormality in the region are eliminated from the data that need to be corrected and taken as normal data.

[0051] As a preferred solution of the data correction method based on centralized meter reading described in the present invention, wherein: the improved annealing algorithm includes defining each abnormal data as an initial solution, and setting an initial temperature based on the abnormality degree of the initial solution;

[0052] For data determined to be abnormal in a region, the same perturbation strategy is used in each abnormal region;

[0053] For the data determined to be device abnormal, the same perturbation strategy is used in the abnormal data of each device;

[0054] After the perturbation strategy, analyze whether the fitness function value is better than the current solution. If the candidate solution is better than the fitness function value of the current solution, the candidate solution is accepted; otherwise, the candidate solution is not accepted; when the candidate solution is accepted, the abnormal data is automatically updated to the perturbed data;

[0055] When the termination condition is met, the correction of the meter reading data is completed;

[0056] The termination condition includes: if any one of the following three conditions is met, the termination condition is considered to be met:

[0057] The iteration results of abnormal data are all within the historical fluctuation range;

[0058] The iteration results of abnormal data are not updated for z consecutive times;

[0059] The maximum number of iterations has been reached.

[0060] As a preferred solution of the data correction method based on centralized meter reading described in the present invention, the disturbance strategy includes the disturbance strategy within the area:

[0061] ,

[0062] in, Represents a random perturbation, applicable to the region For all data points in the same area are the same value; Indicates the current data. Represents the updated data after disturbance;

[0063] Device disturbance strategy:

[0064] ,

[0065] in, Represents a random perturbation factor, applicable to the device For all data points in the same area are the same value;

[0066] The fitness function includes measuring each candidate solution The difference between the maximum and minimum values ​​of the historical data fluctuation range. If the candidate solution is not within the historical fluctuation range, the distance between the candidate solution and the historical fluctuation range is calculated as the evaluation function value; if the candidate solution is within the historical fluctuation range, the evaluation function value of the candidate solution is 0; the comprehensive result of the statistical evaluation function value is used as the adaptability function value;

[0067] ,

[0068] ,

[0069] ,

[0070] in, Indicates abnormal judgment results. represents the value of the candidate solution, Indicates the minimum value of the historical fluctuation range of meter i, Indicates the maximum value of the historical fluctuation range of meter i, represents the distance between the candidate solution and the historical fluctuation range, represents the fitness function value, M represents the number of abnormal data after secondary screening;

[0071] When the fitness function value of the candidate solution is less than the fitness function value of the current solution, the candidate solution is deemed to be better than the fitness function value of the current solution.

[0072] A data correction system based on centralized meter reading using any method of the present invention, wherein:

[0073] The collection unit collects different types of meter data through meter reading equipment and transmits the meter reading data to the centralized meter reading data management platform through the collection terminal;

[0074] A screening unit performs a preliminary abnormality judgment on the meter reading data and initially screens the data that needs to be corrected; and performs a secondary screening of the data that needs to be corrected by performing a correlation analysis on the data after the initial screening;

[0075] The correction unit uses the improved annealing algorithm to correct the data in the result of the secondary screening.

[0076] A computer device comprises: a memory and a processor; the memory stores a computer program, wherein: the processor implements the steps of any one of the methods of the present invention when executing the computer program.

[0077] A computer-readable storage medium stores a computer program, wherein: when the computer program is executed by a processor, the steps of any one of the methods of the present invention are implemented.

[0078] Beneficial effects of the present invention: The data correction method based on centralized meter reading provided by the present invention can effectively reduce manual operations, improve efficiency and data real-time performance. It improves the accuracy of preliminary anomaly detection and reduces false positives and false negatives. It comprehensively considers the complex correlations between data and further improves the screening accuracy. Through the global optimization strategy, it realizes accurate correction of abnormal data and ensures the reliability and stability of data. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0080] Figure 1 This is an overall flow chart of a data correction method based on centralized meter reading provided in the first embodiment of the present invention. DETAILED DESCRIPTION

[0081] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0082] Example 1

[0083] Reference Figure 1 , is an embodiment of the present invention, and provides a data correction method based on centralized meter reading, comprising:

[0084] S1: Collect different types of meter data through meter reading equipment, and transmit the meter reading data to the centralized meter reading data management platform through the collection terminal.

[0085] The different types of meter data include electricity meter data, water meter data, gas meter data, and heat meter data. The acquisition terminal integrates the meter reading data collected by all meter reading devices, inserts three tags into each meter reading data, and transmits the data integrated by the acquisition terminal to the data management platform through the communication network.

[0086] Furthermore, the three tags include: location tag, user tag, and device tag. The location tag includes the geographical location information of the meter, and the location information of the meter can be queried through the Internet to obtain environmental influencing factors. The user tag includes, through the matching relationship table between each meter and the user, recording the user matched by the current meter when recording meter reading data. The device tag includes recording the device information of the current meter data.

[0087] It is important to collect different types of meter data such as electricity meters, water meters, gas meters, and heat meters through meter reading equipment. Comprehensive coverage of various types of utility data is achieved to ensure the comprehensiveness and diversity of data. The collected meter data is transmitted to the centralized meter reading data management platform through the collection terminal. Through a unified collection terminal, the integration and transmission of multiple meter data is achieved to ensure the real-time and consistency of data.

[0088] It realizes automated data collection, reduces manual intervention, reduces error rates, and improves data accuracy and real-time performance. It ensures the timeliness and consistency of data, which facilitates subsequent centralized processing and analysis. By integrating and transmitting multiple meter data and combining location, user, and device tags, it can generate higher-value data reports and analysis results, support utilities' operational decisions, and improve service quality and management efficiency.

[0089] S2: Perform a preliminary abnormality judgment on the meter reading data and initially screen the data that needs to be corrected.

[0090] The integrated meter reading data is classified according to the user tags to obtain the meter reading data of each user. In the meter reading data of each user, abnormality judgment is performed on each meter as a unit, and the abnormal part is identified by analyzing the historical records of the meter reading data.

[0091] Features users, the data of each user i is express.

[0092] For each user's data , extract the data of each meter and calculate its change. Suppose the data of each meter is .

[0093] The formula for calculating the change is:

[0094] ,

[0095] in, represents the change in meter reading data of the jth meter of the i-th user, Indicates User's Meters at time The readings, Indicates the time interval between meter readings. Indicates A meter, Indicates time.

[0096] Calculate the characteristic values ​​of the variation data, including the mean, standard deviation, and Fourier transform.

[0097] ,

[0098] ,

[0099] ,

[0100] in, represents the average value of the change in the jth meter of the ith user, T represents the time period of the meter reading data, It is expressed as the standard deviation of the change in the jth meter of the ith user, represents the Fourier transform result of the change of the jth meter of the ith user, and ω represents the frequency variable.

[0101] Anomaly detection is performed using a multilayer perceptron neural network with the input being a feature vector , the output is the anomaly score .

[0102] Eigenvectors:

[0103] ,

[0104] Neural network formula:

[0105] ,

[0106] in, and Represents the weight matrix of the neural network; and Represents the bias vector of the neural network; ReLU is the activation function; represents the Sigmoid activation function, which converts the output into a probability value between 0 and 1. For each meter of each user, if the abnormal score Exceeding the preset threshold , it is judged as abnormal; the data judged as abnormal is determined as the data that needs to be corrected in the initial screening.

[0107] It is important to know that through the initial screening, all data that may contain anomalies can be obtained, and secondary screening of these data will be more accurate and easier. Through user label classification, data can be grouped by user, which is convenient for refined management and analysis of each user's data. After classification by user, abnormal judgment is made on a per-meter basis, which can more accurately identify the abnormal part of the data and avoid misjudgment caused by overall analysis. Neural networks can efficiently process data containing multiple features and quickly output anomaly detection results. By utilizing the nonlinear characteristics of neural networks, abnormal data can be identified more accurately and false positives and false negatives can be reduced.

[0108] S3: Perform secondary screening of data that need correction through correlation analysis of the data after the initial screening.

[0109] It should also be said that the association analysis includes obtaining each data that needs to be corrected, and establishing a graph neural network based on the location label of each data; the node is each data that needs to be corrected, and the edge is the distance relationship between the positions; according to the location of the abnormal data in the historical records, the DBSCAN algorithm is used to divide multiple historical abnormality areas, and each area contains multiple location nodes; by analyzing whether the location where the abnormality occurs is concentrated in the same divided area, if it is concentrated in the same divided area, it is determined to be an abnormality within the area; if it is not concentrated in the same divided area, it is determined to be no abnormality within the area.

[0110] Specifically:

[0111] Use graph neural networks to calculate the anomaly score of each node and calculate the anomaly probability of each region based on the score. Use graph convolutional layers for feature propagation:

[0112] ,

[0113] Calculate the anomaly score for each node:

[0114] ,

[0115] Calculate the anomaly probability for each region:

[0116] ,

[0117] in, Represents the node feature matrix of the lth layer; Represents the node feature matrix of the l+1th layer; represents the nonlinear activation function Sigmoid; D is the degree matrix, which represents the degree of each node in the graph, which is the number of edges connected to the node; represents the square root inverse matrix of the degree matrix, which is used for normalization processing; A is the adjacency matrix, which represents the connection relationship between nodes in the graph; W is the weight matrix, which represents the parameters of the linear transformation; represents the weight matrix of the lth layer; represents the anomaly score of node i; represents the set of neighbor nodes of node i; j represents the index of the neighbor node of node i; Indicates the number of neighbor nodes of node i; Represents the feature vector of node i; represents the feature vector of node j, the neighbor of node i; σ represents the bandwidth parameter of the Gaussian kernel function; Indicates area The probability of abnormality; It is represented as the number of nodes in the kth region; i represents the index of the node. According to the calculated regional anomaly probability, it is determined whether the current anomaly is concentrated in a certain historical anomaly region.

[0118] Furthermore, the secondary screening includes setting a threshold for abnormal nodes in the area if the current abnormal location is concentrated in the historical area. Make a judgment, when , it is judged as an abnormality within the region, otherwise it is judged as no abnormality within the region.

[0119] The data that needs to be corrected is classified according to the device label, and the data abnormality ratio in each device label is analyzed. If the data abnormality ratio exceeds the preset threshold value YC, it is judged as device abnormality; otherwise, it is judged as no device abnormality. The data judged as the device abnormality and the abnormality in the region are used as the results of secondary screening and corrected through the improved annealing algorithm; the data judged as no device abnormality and no abnormality in the region are eliminated from the data that needs to be corrected and regarded as normal data.

[0120] Through secondary screening, we can find out whether the data anomaly is caused by abnormal environmental factors or abnormal equipment. Through historical records, we can find areas where abnormalities may be caused by environmental factors (because the impact of environmental factors is generally regional). If the current anomaly is also concentrated in these areas, it means that this anomaly is likely to be caused by environmental factors. Similarly, it is also possible that there is a problem with the equipment, resulting in widespread abnormalities in the data collected by this device. If it is not the above two factors, then it is most likely due to the user's personal behavior (caused by changes in behavior or changes in the user). This situation cannot be modified on a large scale through artificial intelligence, and generally this situation is not a data anomaly.

[0121] By building a graph neural network through location tags and using the relationship between nodes and edges, we can build the spatial correlation of data and help understand the geographical relationship between data. Using the DBSCAN algorithm to divide historical anomaly areas helps to identify and understand the spatial aggregation of data and clarify the spatial distribution of abnormal data. By analyzing the location distribution of abnormal data, we can more accurately determine whether the anomaly is concentrated in a certain area and improve the spatial sensitivity of anomaly detection.

[0122] S4: In the results of the secondary screening, the improved annealing algorithm is used to perform data correction.

[0123] The improved annealing algorithm includes defining each abnormal data as an initial solution and setting an initial temperature based on the abnormality degree of the initial solution.

[0124] For data determined to be abnormal within a region, the same perturbation strategy is used in each abnormal region.

[0125] For data determined to be device abnormalities, the same perturbation strategy is used in the abnormal data of each device.

[0126] After the perturbation strategy, analyze whether the fitness function value is better than the current solution. If the candidate solution is better than the fitness function value of the current solution, the candidate solution is accepted; otherwise, the candidate solution is not accepted; when the candidate solution is accepted, the abnormal data is automatically updated to the perturbed data.

[0127] When the termination condition is met, the correction of the meter reading data is completed.

[0128] The termination condition includes: if any one of the following three conditions is met, the termination condition is considered to be met: the iteration results of the abnormal data are all within the historical fluctuation range; the iteration results of the abnormal data are not updated for z consecutive times; the maximum number of iterations is reached.

[0129] The disturbance strategy includes the disturbance strategy within the region:

[0130] ,

[0131] in, Represents a random perturbation, applicable to the region For all data points in the same area are the same values, Indicates the current data. Represents the updated data after disturbance.

[0132] Device disturbance strategy:

[0133] ,

[0134] in, Represents a random disturbance factor, applicable to the device For all data points in the same area are the same values.

[0135] The fitness function includes measuring each candidate solution The difference between the maximum and minimum values ​​of the historical data fluctuation range. If the candidate solution is not within the historical fluctuation range, the distance between the candidate solution and the historical fluctuation range is calculated as the evaluation function value; if the candidate solution is within the historical fluctuation range, the evaluation function value of the candidate solution is 0; the comprehensive result of the statistical evaluation function value is used as the adaptability function value;

[0136] ,

[0137] ,

[0138] ,

[0139] in, Indicates abnormal judgment results. represents the value of the candidate solution, Indicates the minimum value of the historical fluctuation range of meter i, Indicates the maximum value of the historical fluctuation range of meter i, represents the distance between the candidate solution and the historical fluctuation range, represents the fitness function value, and M represents the number of abnormal data after secondary screening. When the fitness function value of the candidate solution is less than the fitness function of the current solution, the candidate solution is considered to be better than the fitness function value of the current solution.

[0140] By perturbing and correcting abnormal data in the region and equipment in a targeted manner, the true value of the data can be restored more accurately and the overall data quality can be improved. By using the improved annealing algorithm, the temperature and acceptance criteria can be dynamically adjusted to achieve global optimization of abnormal data and avoid falling into the local optimal solution. The design of multiple termination conditions can ensure that the algorithm converges within a reasonable time, avoid invalid iterations, and improve correction efficiency.

[0141] By setting a reasonable adaptability function, the correction result is ensured to be within the historical fluctuation range, the robustness of the algorithm is improved, and it can adapt to different types of abnormal data.

[0142] By using the same perturbation strategy for data judged to be abnormal in the region, common anomalies in the region can be captured and collectively corrected, thereby improving the overall consistency and accuracy of the correction. By using the same perturbation strategy for data judged to be abnormal in the device, it is possible to identify and correct the abnormal patterns unique to the device and improve the ability to handle device anomalies.

[0143] By measuring the difference between the candidate solution and the historical data fluctuation range, dynamic adjustment and global optimization are achieved to ensure that the data correction results are closer to the actual situation. This includes the iteration results of abnormal data are within the historical fluctuation range, the iteration results of abnormal data are not updated for z consecutive times, and the maximum number of iterations is reached, ensuring that the algorithm can end in time and improve computing efficiency. Among them, z is the preset number of iterations.

[0144] On the other hand, this embodiment also provides a data correction system based on centralized meter reading, which includes:

[0145] The collection unit collects different types of meter data through meter reading equipment, and transmits the meter reading data to the centralized meter reading data management platform through the collection terminal.

[0146] The screening unit performs a preliminary abnormality judgment on the meter reading data and initially screens the data that needs to be corrected; and performs a secondary screening of the data that needs to be corrected by performing a correlation analysis on the data after the initial screening.

[0147] The correction unit uses the improved annealing algorithm to correct the data in the result of the secondary screening.

[0148] If the above functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program code.

[0149] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0150] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0151] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0152] Example 2

[0153] The following is an embodiment of the present invention, which provides a data correction method based on centralized meter reading. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0154] This method was verified through simulation experiment 1, and three situations were designed as the experimental group, the situation of one area abnormality, the situation of one device abnormality, and the situation of one device abnormality plus area abnormality; a situation without abnormality was designed as the control group.

[0155] The data before and after correction were recorded and compared to obtain Table 1.

[0156] Table 1 Data record table

[0157] ,

[0158] The data correction effects of the existing technology and this method are compared through the data record table, and the significant advantages of this method in data correction are clarified. By comparing the abnormality rate of the original data and the abnormality rate of the corrected data, it can be seen that this method has obvious advantages in correcting data anomalies.

[0159] No abnormality: The original data values ​​are 230, 245, 255, 270, and 280, and the data values ​​do not change after correction. Since there is no abnormality, the data is not biased, and the correction success rate is 100%, indicating that this method will not cause miscalibration of non-abnormal data under normal circumstances.

[0160] Regional anomalies: The original data values ​​are 230, 245, 400, 270, 280, and the data values ​​after correction by this method are adjusted to 232, 246, 255, 269, 278. Strong magnetic field interference causes a certain data point to be significantly higher than normal. Through the regional disturbance strategy, the corrected data value is significantly restored to normal, and the correction success rate is 92.5%, indicating that this method can effectively correct regional anomalies.

[0161] Equipment anomaly: The original data values ​​are 230, 245, 500, 270, 280. After correction by this method, the data values ​​are adjusted to 231, 244, 256, 268, 277. Equipment failure causes a single data point to be abnormally high. Through the equipment anomaly disturbance strategy, the corrected data value returns to normal, and the correction success rate is 93.2%, indicating that this method can effectively correct anomalies caused by equipment failure.

[0162] Equipment anomaly + regional anomaly: The original data values ​​are 230, 245, 600, 270, 280. After correction by this method, the data values ​​are adjusted to 233, 246, 254, 269, 278. Strong magnetic field interference and equipment failure together lead to data anomalies. By combining the disturbance strategy of the region and the equipment, the corrected data values ​​return to normal, and the correction success rate is 91.8%, indicating that this method can cope with complex multiple abnormal situations and effectively correct abnormal data.

[0163] The superiority of the data correction method based on centralized meter reading proposed by the present invention in terms of correction speed and accuracy is verified by Experiment 2. Experimental subjects: 1000 users are selected, including data of electricity meters, water meters, gas meters and heat meters. The experimental data are recorded to obtain the average data results shown in Table 2.

[0164] Table 2 Data comparison table

[0165] ,

[0166] It can be seen from Table 2 that the present invention can greatly shorten the calibration time, greatly saving manpower and time costs. At the same time, the present method can further improve the accuracy after calibration.

[0167] It is proved that the method of the present invention has significant advantages in both efficiency and accuracy, which reflects the innovation and practical value of the present invention.

[0168] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A data correction method based on centralized meter reading, characterized in that: include: Through meter reading equipment, different types of meter data are collected, and the meter reading data is transmitted to the centralized meter reading data management platform through the collection terminal; Performing preliminary abnormality judgment on the meter reading data and initially screening the data that needs to be corrected; Through the correlation analysis of the data after the initial screening, the secondary screening of the data that needs to be corrected is carried out; In the results of the secondary screening, the improved annealing algorithm was used to correct the data; The association analysis includes obtaining each data that needs to be corrected, and establishing a graph neural network according to the location label of each data; the node is each data that needs to be corrected, and the edge is the distance relationship between the locations; according to the location of the abnormal data in the historical records, the DBSCAN algorithm is used to divide multiple historical abnormal areas, each area contains multiple location nodes; by analyzing whether the location where the abnormality occurs is concentrated in the same divided area, if it is concentrated in the same divided area, it is determined to be an abnormality in the area; If they are not concentrated in the same divided area, it is determined that there is no abnormality in the area; Specifically: Use graph neural networks to calculate the anomaly score of each node and calculate the anomaly probability of each region based on the score; Use graph convolutional layers for feature propagation: H (l+1) =ρ(D -1 / 2 AD -1 / 2 H (l) W (l) ) Calculate the anomaly score for each node: Calculate the anomaly probability for each region: Among them, H( l ) represents the node feature matrix of the lth layer; H( l+1 ) represents the node feature matrix of the l+1th layer; ρ represents the nonlinear activation function Sigmoid; D is the degree matrix, which represents the degree of each node in the graph, which is the number of edges connected to the node; D -1 / 2 represents the square root inverse matrix of the degree matrix, which is used for normalization; A is the adjacency matrix, which represents the connection relationship between nodes in the graph; W is the weight matrix, which represents the parameters of the linear transformation; W( l ) represents the weight matrix of the lth layer; S i represents the abnormality score of node i; N(i) represents the set of neighbor nodes of node i; j represents the index of the neighbor node of node i; |N(i)| represents the number of neighbor nodes of node i; H i represents the feature vector of node i; H j represents the feature vector of node j, the neighbor of node i; σ represents the bandwidth parameter of the Gaussian kernel function; P(R k ) represents region R k The abnormal probability of |R k | represents the number of nodes in the kth region; i represents the index of the node; According to the calculated regional anomaly probability, determine whether the current anomaly is concentrated in a certain historical anomaly area; The secondary screening includes setting a threshold θ2 for abnormal nodes in the region to make a judgment if the current abnormal location is concentrated in the historical region. k )≥θ2, it is judged as an abnormality within the region, otherwise it is judged as an abnormality without the region; The data to be corrected are classified according to the device labels, and the data abnormality ratio in each device label is analyzed. If the data abnormality ratio exceeds a preset threshold value YC, it is determined that the device is abnormal; otherwise, it is determined that there is no device abnormality; The data judged as the device abnormality and the abnormality in the region are taken as the result of the secondary screening and corrected by the improved annealing algorithm; the data judged as having no device abnormality and no abnormality in the region are eliminated from the data that need to be corrected and taken as normal data.

2. The data correction method based on centralized meter reading according to claim 1, characterized in that: The different types of meter data include electricity meter data, water meter data, gas meter data, and heat meter data; The collection terminal integrates the meter reading data collected by all meter reading devices, inserts three tags into each meter reading data, and transmits the data integrated by the collection terminal to the data management platform through the communication network; The three tags include: location tag, user tag, and device tag; The location tag includes the geographical location information of the meter, and the location information of the meter can be queried through the Internet to obtain environmental influencing factors; The user tag includes, through the matching relationship table between each meter and the user, recording the user matched by the current meter when recording the meter reading data; The device tag includes device information for recording current meter data.

3. The data correction method based on centralized meter reading according to claim 2, characterized in that: The preliminary abnormality judgment includes classifying the integrated meter reading data according to the user tags to obtain the meter reading data of each user; In the meter reading data of each user, abnormality judgment is performed on each meter as a unit, and the abnormal part is identified by analyzing the historical records of meter reading data; Suppose there are N users, and the data of each user i is represented by D i express; For each user’s data D i , extract the data of each meter and calculate its change. Suppose the data of each meter is x ij (t); The formula for calculating the change is: Δx ij (t)=x ij (t)-x ij (t-Δt) Where Δx ij (t) represents the change in meter reading data of the jth meter of the i-th user, x ij (t) represents the reading of the jth meter of the i-th user at time t, Δt represents the time interval between meter readings, j represents the jth meter, and t represents time; Calculate the characteristic values ​​of the variation data, including mean, standard deviation and Fourier transform; Among them, μ ij represents the average value of the change in the jth meter of the ith user, T represents the time period of the meter reading data, σ ij It is expressed as the standard deviation of the change in the jth meter of the ith user, represents the Fourier transform result of the change of the jth meter of the i-th user, ω represents the frequency variable; Anomaly detection is performed using a multilayer perceptron neural network with the input being the feature vector f ij , the output is the anomaly score s ij ; Eigenvectors: Neural network formula: s ij =Ε(W2·ReLU(W1·f ij +b1)+b2) Where W1 and W2 represent the weight matrices of the neural network; b1 and b2 represent the bias vectors of the neural network; ReLU(x)=max(0,x) is the activation function; Represents the Sigmoid activation function, which converts the output into a probability value between 0 and 1; For each meter of each user, if the anomaly score s ij If it exceeds the preset threshold value θ1, it is determined to be abnormal; the data determined to be abnormal is determined as data that needs to be corrected in the initial screening.

4. The data correction method based on centralized meter reading as claimed in claim 3, characterized in that: The improved annealing algorithm includes defining each abnormal data as an initial solution and setting an initial temperature based on the abnormality degree of the initial solution; For data determined to be abnormal in a region, the same perturbation strategy is used in each abnormal region; For the data determined to be device abnormal, the same perturbation strategy is used in the abnormal data of each device; After the perturbation strategy, analyze whether the fitness function value is better than the current solution. If the candidate solution is better than the fitness function value of the current solution, the candidate solution is accepted; otherwise, the candidate solution is not accepted; when the candidate solution is accepted, the abnormal data is automatically updated to the perturbed data; When the termination condition is met, the correction of the meter reading data is completed; The termination condition includes: if any one of the following three conditions is met, the termination condition is considered to be met: The iteration results of abnormal data are all within the historical fluctuation range; The iteration results of abnormal data are not updated for z consecutive times; The maximum number of iterations has been reached.

5. The data correction method based on centralized meter reading as claimed in claim 4, characterized in that: The disturbance strategy includes the disturbance strategy within the region: X new =X current +δ Among them, δ represents a random perturbation, applicable to the region R k For all data points in the same area, δ has the same value; X current Describe the current data, X new Represents the updated data after disturbance; Device disturbance strategy: X new =X current ·(1+n) Where η represents a random perturbation factor, applicable to device E j For all data points in the same region, η has the same value; The fitness function includes measuring each candidate solution The difference between the maximum and minimum values ​​of the historical data fluctuation range. If the candidate solution is not within the historical fluctuation range, the distance between the candidate solution and the historical fluctuation range is calculated as the evaluation function value; if the candidate solution is within the historical fluctuation range, the evaluation function value of the candidate solution is 0; the comprehensive result of the statistical evaluation function value is used as the adaptability function value; Among them, FW i Indicates abnormal judgment results. represents the value of the candidate solution, Indicates the minimum value of the historical fluctuation range of meter i, Indicates the maximum value of the historical fluctuation range of meter i, SZ i represents the distance between the candidate solution and the historical fluctuation range, represents the fitness function value, M represents the number of abnormal data after secondary screening; When the fitness function value of the candidate solution is less than the fitness function value of the current solution, the candidate solution is deemed to be better than the fitness function value of the current solution.

6. A data correction system based on centralized meter reading using the method according to any one of claims 1 to 5, characterized in that: The collection unit collects different types of meter data through meter reading equipment and transmits the meter reading data to the centralized meter reading data management platform through the collection terminal; A screening unit, for making a preliminary abnormality judgment on the meter reading data and initially screening the data that needs to be corrected; Through the correlation analysis of the data after the initial screening, the secondary screening of the data that needs to be corrected is carried out; The correction unit uses the improved annealing algorithm to correct the data in the result of the secondary screening.

7. A computer device comprising: Memory and processor; The memory stores a computer program, wherein the processor implements the steps of any one of the methods of claims 1-5 when executing the computer program.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

Images