Electric energy meter metering data fault analysis method based on improved AdaBoost algorithm

Abstract

The invention discloses an electric energy meter metering data fault analysis method based on an improved AdaBoost algorithm, and relates to the technical field of electric energy meter metering detection and analysis. Comprising the following main steps of data acquisition and metering, data information preprocessing, data transmission, data processing, data analysis, statistical comparison of data records, misaligned ammeter marking, data reporting and data application. Compared with a single BP neural network fault prediction method, the BP-AdaBoost fault prediction method has the advantages that the prediction precision can be remarkably improved; according to the electric energy meter measurement data fault analysis method based on the improved AdaBoost algorithm and the BP neural network, the prediction precision is improved, the errors are reduced, certain feasibility is achieved, the correlation between diversity and the test errors is quantitatively calculated through the Pearson's correlation coefficient, the classification performance is further improved by improving the prediction precision, reducing the errors and improving the AdaBoost algorithm, and the analysis accuracy is improved. And the method is more reliable and effective in practical application.

Description

Technical Field

[0001] The invention relates to the technical field of electric energy meter measurement detection and analysis, and in particular to an electric energy meter measurement data fault analysis method based on an improved AdaBoost algorithm. Background Art

[0002] The collection, measurement, storage and transmission of electric energy mainly rely on electric energy meters, and their accuracy is related to the vital interests of power grid companies and every user. Regional anomalies directly affect the accuracy of electric energy meter status evaluation, and there are currently deficiencies in abnormal diagnosis and automated governance mechanisms. With the continuous deepening of data and business governance work, the basic situation of the substation area has been significantly improved, but there are still problems such as household changes, black household electricity theft, line leakage, and misalignment of data collection time, which restrict the accuracy of model calculations and have a great impact on the development of other businesses. Maintaining the stability of the distribution network is imminent. Only when the distribution network is stable can people's normal life and industrial development be guaranteed. Operation and maintenance managers need to pay attention to the causes of metering failures of smart electric energy meters and try to eliminate these failures. Now we need an improved AdaBoost algorithm for electric energy meter metering data fault analysis method to better solve the problem of electric energy meter inaccuracy analysis.

[0003] In conventional technology, the diagnosis of electric energy meter faults is usually limited to the modification of the hardware structure of the electric energy meter. By modifying its appearance structure and wiring method, and making targeted selections of external switches, the fault of the electric energy meter can be controlled. Although this method can reduce the occurrence of electric energy meter faults to a certain extent, it is impossible to comprehensively analyze the fault factors of the electric energy meter due to the many factors that cause the electric energy meter faults. In addition, the traditional electric energy meter measurement data fault analysis method only relies on BP neural network for electric energy meter fault diagnosis. Although BP neural network provides a better structural system, it has the disadvantages of slow training convergence speed and easy to fall into local minimum points, which reduces the work efficiency of operators.

[0004] To solve the above problems, a method for fault analysis of electric energy meter measurement data based on improved AdaBoost algorithm is proposed. Summary of the invention

[0005] The purpose of the present invention is to provide an improved AdaBoost algorithm for analyzing the faults of electric energy meter measurement data, and to adopt the device to work, so as to solve the problem of electric energy meter inaccuracy that is inevitable for electric energy meters that have been installed and used for a long time in the above background. In the prior art, the traditional solution of the current power company to the problem of electric energy meter inaccuracy is to regularly replace the electric energy meter or use instruments for spot check and calibration, and the economic cost of regularly replacing the electric energy meter is extremely high. The method of using instruments for spot check and calibration consumes a lot of manpower and material resources, has high maintenance costs, low efficiency, and a large range of spot checks, which affects people's inability to deal with the faults of electric energy meters in a timely manner, causing losses to people.

[0006] To achieve the above object, the present invention provides the following technical solution: a method for analyzing faults in electric energy meter measurement data based on an improved AdaBoost algorithm, comprising the following main steps;

[0007] Step 1: Data collection and measurement: Obtain the data information of the electric energy meter recorded and detected by the operator; measure and collect the electric energy data of the sub-meter through the low-voltage IoT sensing terminal, combine the voltage and current curves of the power consumption information collection system with Kirchhoff's voltage and current law, and recursively obtain the voltage and current amplitude of the nearest meter box node with the same phase as the electric energy meter to be checked;

[0008] Step 2: Data information preprocessing: screening and sampling the electric energy meter data obtained in step 1, and then outputting pure data information to achieve preprocessing of the electric energy meter data information;

[0009] Step 3: Data transmission: receiving and transmitting the electric energy meter data information processed in step 2 by wired or wireless communication, and transmitting the data information;

[0010] Step 4: Data processing: Pre-judge and process the collected data information related to the energy meter to prepare for subsequent analysis. According to the phase information of the power consumption information collection system, determine the energy meter information to be verified, and calculate, process and transmit the integrated energy meter data information;

[0011] The received electric energy meter data information is fused through big data fusion calculation; the fused electric energy meter data information is classified using the BP neural network algorithm; and then the AdaBoost algorithmThe classified electric energy meter data information is diagnosed and the diagnostic data information is output; through the electronic electric energy meter simulation model built by Simulink, the fault injection point and observation point are selected, and the fault data generated by the simulation operation is used as the processing object of the BP-AdaBoost algorithm; for the output diagnostic electric energy meter data information between weak classifiers under the AdaBoost algorithm, experiments are carried out on the UCI dataset, and the Pearson correlation coefficient is used to quantitatively calculate the correlation between diversity and test error.

[0012] Step 5: Identify and troubleshoot suspected electricity theft meters: Analyze and determine whether there are suspected electricity theft meters, mark them, and eliminate high-loss points to avoid affecting the evaluation of the overall electricity meter collection data;

[0013] Step 6: Data analysis: used to evaluate the accuracy of the energy meter. Combined with the voltage and current curve data of the power consumption information collection system and the number of energy meters with the same phase in the same box, the corresponding formula is used for multivariate linear regression analysis to obtain the operating error of the energy meter to be verified.

[0014] Step 7: Statistical comparison of data records: used to count and record the error data of the electric energy meter to be checked obtained in step 6, so as to facilitate subsequent operators to compare and analyze multiple groups of data and obtain the results of fault analysis of the electric energy meter measurement data;

[0015] Step 8: After determining that there are electric energy meters with inaccurate measurement, select some of the electric energy meters to be inspected from multiple electric energy meters to be inspected in turn according to a preset selection strategy, and re-test the selected electric energy meters for inaccuracy, re-verify, and obtain the test results to avoid detection errors;

[0016] Step 9: Marking and reporting inaccurate meters: The meters with sub-meter errors exceeding the specified threshold are marked as inaccurate meters. When the operating error of the electric energy meter to be verified is greater than the allowable threshold, it is identified as an inaccurate meter, and the data of the inaccurate meter is actively reported to the main station through the acquisition terminal;

[0017] Step 10: Data application: reporting the metering fault data of the electric energy meter re-detected and verified in step 8, for storing, using or transmitting the data processed by the data fusion computing layer, for query, tracing or management by the superior central management computer.

[0018] Preferably, in step 4, the big data fusion calculation method is: filtering data from the received original data; and constructing a data fusion function.

[0019] Preferably, in step 4, the big data classification algorithm module is a BP neural network algorithm, and the principle of BP neural network algorithm classification is: the BP algorithm is composed of two processes: the forward propagation of the signal and the reverse propagation of the error in the learning process;

[0020] Since the training of multi-layer feedforward networks often uses the error back propagation algorithm, people often directly call the multi-layer feedforward network a BP network. The BP algorithm consists of two processes: the forward propagation of the signal and the back propagation of the error.

[0021] Preferably, the BP algorithm is specifically divided into the following steps: (1) Initialization, using small random numbers to assign initial values ​​to each weight and threshold. Note that the initial weights and thresholds in the network cannot be completely equal, otherwise the network cannot run from such a structure to a non-equal weight structure; (2) Reading network parameters and training sample sets; (3) Normalization processing; (4) Calculating each sample in the training set; (5) If the accuracy requirements or other exit conditions are met, the training is terminated, otherwise go to step (4) to continue; (6) Result analysis and output.

[0022] Preferably, the method for the AdaBoost algorithm module to perform fault diagnosis of electric energy meter measurement data is:

[0023] (1) Use the AdaBoost algorithm to train weak classifiers to form strong classifiers. AdaBoost adjusts the weight of each weak classifier so that the misclassified samples receive a larger weight in subsequent classifications, allowing the classifier to pay more attention to these samples in subsequent training and improve the overall classification accuracy.

[0024] (2) Training and optimization: Through multiple iterations of training, the weights and parameters of the weak classifier are adjusted to achieve the best classification effect. This process may include steps such as cross-validation and adjusting the learning rate to optimize the performance of the model.

[0025] Preferably, the method for diagnosing faults in electric energy meter measurement data using the AdaBoost algorithm module further includes:

[0026] (3) Fault diagnosis: After training, the model can be used to perform fault diagnosis on new electricity meter data. The model will determine whether a fault exists based on the input data and give the specific fault type.

[0027] (4) Parallel computing and efficiency improvement: In order to improve training efficiency and shorten network running time, the principle of parallel computing can be adopted and this technology can be used in the implementation of software and hardware to accelerate the training process.

[0028] Preferably, the diversity measurement problem and over-adaptation problem between weak classifiers under the AdaBoost algorithm are solved by conducting experiments on the UCI data set and using the Pearson correlation coefficient to quantitatively calculate the correlation between diversity and test error. The specific steps are:

[0029] The Pearson correlation coefficient between two variables is defined as the quotient of the covariance and the standard deviation between the two variables:

[0030] The above formula defines the overall correlation coefficient, which is usually represented by the Greek lowercase letter ρ. By estimating the covariance and standard deviation of the sample, we can get the Pearson correlation coefficient, which is usually represented by the English lowercase letter r:

[0031]

[0032] Preferably, r can also be estimated by the standard score mean of the sample points to obtain an expression equivalent to the above formula:

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] 1. The present invention combines Internet of Things technology, big data technology and data communication technology to construct a detection system architecture, which realizes the sensing, transmission, analysis, calculation and upper-level data monitoring of electric energy meter measurement detection fault data, so that electric energy management users can remotely, online and real-time monitor electric energy meter measurement fault information. Different from conventional technologies, the present invention has a higher level of intelligence and a much stronger ability to process data, which is conducive to users quickly obtaining electric energy meter detection information from complex multi-dimensional data information;

[0035] 2. The present invention can organically integrate the big data fusion calculation module, the big data classification algorithm module and the AdaBoost algorithm module to form an improved AdaBoost algorithm, and can classify and fuse a large amount of data information, so as to facilitate the centralized management of data information of different latitudes and types, thereby effectively classifying data information of different attributes and realizing big data classification, which is conducive to users reading, tracing and managing classified data, and finally realizing data fault analysis through the AdaBoost algorithm model, thereby improving the management ability of electric energy meter measurement data;

[0036] 3. The method of the present invention is applied to the classification and discrimination of typical faults of electronic electric energy meters and rectifier bridge faults, and the fault data generated by simulation operation is used as the processing object of the BP-AdaBoost algorithm. The experimental results show that compared with the single BP neural network fault prediction method, the BP-AdaBoost fault prediction method can significantly improve the prediction accuracy and reduce the error, and has certain feasibility. In addition, by conducting experiments on the UCI data set, the Pearson correlation coefficient is used to quantitatively calculate the correlation between diversity and test error. It is found that the change pattern of the double error metric on different data sets is fixed, and it is basically unchanged in the later iteration period and tends to be stable. The test error of this improved AdaBoost algorithm is reduced by an average of 1.5 percentage points, and can be reduced by up to 4.8 percentage points, thereby further improving the classification performance. This improved AdaBoost algorithm combined with the BP neural network electric energy meter measurement data fault analysis method provides an effective solution for fault detection of electric energy meters by improving prediction accuracy and reducing errors. At the same time, through the improvement of the AdaBoost algorithm, the classification performance is further improved, making it more reliable and effective in practical applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 It is a flow chart of steps 1 to 5 in the method for analyzing faults in electric energy meter measurement data of the present invention;

[0038] Figure 2 It is a flow chart of steps 5 to 7 in the method for analyzing faults in electric energy meter measurement data of the present invention;

[0039] Figure 3 This is a flow chart of steps 7 to 10 in the method for analyzing faults in electric energy meter measurement data of the present invention. DETAILED DESCRIPTION

[0040] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of 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.

[0041] In order to further understand the content of the present invention, the present invention is described in detail in conjunction with the accompanying drawings.

[0042] Implementation 1:

[0043] Combination Figure 1-Figure 3 The present invention provides a technical solution: a method for analyzing faults in electric energy meter measurement data based on an improved AdaBoost algorithm, comprising the following main steps:

[0044] Step 1: Data collection and measurement: Obtain the data information of the electric energy meter recorded and detected by the operator; measure and collect the electric energy data of the sub-meter through the low-voltage IoT sensing terminal, combine the voltage and current curves of the power consumption information collection system with Kirchhoff's voltage and current law, and recursively obtain the voltage and current amplitude of the nearest meter box node with the same phase as the electric energy meter to be checked;

[0045] Step 2: Data information preprocessing: Screening and sampling the electric energy meter data obtained in step 1, and then outputting pure data information to achieve preprocessing of the electric energy meter data information;

[0046] Step 3: Data transmission: receiving and transmitting the electric energy meter data information processed in step 2 by wired or wireless communication, and transmitting the data information;

[0047] Step 4: Data processing: Pre-judge and process the collected data information related to the energy meter to prepare for subsequent analysis. According to the phase information of the power consumption information collection system, determine the energy meter information to be verified, and calculate, process and transmit the integrated energy meter data information;

[0048] The received electric energy meter data information is fused through big data fusion calculation; the fused electric energy meter data information is classified using the BP neural network algorithm; and then the AdaBoost algorithm The classified electric energy meter data information is diagnosed and the diagnostic data information is output; through the electronic electric energy meter simulation model built by Simulink, the fault injection point and observation point are selected, and the fault data generated by the simulation operation is used as the processing object of the BP-AdaBoost algorithm; for the output diagnostic electric energy meter data information between weak classifiers under the AdaBoost algorithm, experiments are carried out on the UCI dataset, and the Pearson correlation coefficient is used to quantitatively calculate the correlation between diversity and test error.

[0049] Step 5: Identify and troubleshoot suspected electricity theft meters: Analyze and determine whether there are suspected electricity theft meters, mark them, and eliminate high-loss points to avoid affecting the evaluation of the overall electricity meter collection data;

[0050] Step 6: Data analysis: used to evaluate the accuracy of the energy meter. Combined with the voltage and current curve data of the power consumption information collection system and the number of energy meters with the same phase in the same box, the corresponding formula is used for multivariate linear regression analysis to obtain the operating error of the energy meter to be verified.

[0051] Step 7: Statistical comparison of data records: used to count and record the error data of the electric energy meter to be checked obtained in step 6, so as to facilitate the subsequent operator to compare and analyze multiple groups of data and obtain the results of the fault analysis of the electric energy meter measurement data;

[0052] Step 8: After determining that there are electric energy meters with inaccurate measurement, select some of the electric energy meters to be inspected from multiple electric energy meters to be inspected in turn according to a preset selection strategy, and re-test the selected electric energy meters for inaccuracy, re-verify, and obtain the test results to avoid detection errors;

[0053] Step 9: Marking and reporting inaccurate meters: The meters with sub-meter errors exceeding the specified threshold are marked as inaccurate meters. When the operating error of the electric energy meter to be verified is greater than the allowable threshold, it is identified as an inaccurate meter, and the data of the inaccurate meter is actively reported to the main station through the acquisition terminal;

[0054] Step 10: Data application: Report the metering fault data of the electric energy meter re-detected and verified in step 8, which is used to store, use or transmit the data processed by the data fusion calculation layer for query, tracing or management by the superior center management computer.

[0055] Implementation 2:

[0056] In step 4, the big data fusion calculation method is: filtering data from the received raw data; and constructing a data fusion function;

[0057] In step 4, the big data classification algorithm module is the BP neural network algorithm. The principle of BP neural network algorithm classification is: the BP algorithm is composed of two processes: the forward propagation of the signal and the reverse propagation of the error;

[0058] Since the training of multi-layer feedforward networks often uses the error back propagation algorithm, people often directly call the multi-layer feedforward network a BP network. The BP algorithm consists of two processes: the forward propagation of the signal and the back propagation of the error.

[0059] The BP algorithm is specifically divided into the following steps: (1) Initialization, using small random numbers to assign initial values ​​to each weight and threshold. Note that the initial weights and thresholds in the network cannot be completely equal, otherwise the network cannot run from such a structure to a non-equal weight structure; (2) Reading network parameters and training sample sets; (3) Normalization processing; (4) Calculating each sample in the training set; (5) If the accuracy requirements or other exit conditions are met, the training ends, otherwise go to step (4) to continue; (6) Result analysis and output.

[0060] The method for AdaBoost algorithm module to perform fault diagnosis on electric energy meter measurement data is as follows:

[0061] (1) Use the AdaBoost algorithm to train weak classifiers to form strong classifiers. AdaBoost adjusts the weight of each weak classifier so that the misclassified samples receive a larger weight in subsequent classifications, allowing the classifier to pay more attention to these samples in subsequent training and improve the overall classification accuracy.

[0062] (2) Training and optimization: Through multiple iterations of training, the weights and parameters of the weak classifier are adjusted to achieve the best classification effect. This process may include steps such as cross-validation and adjusting the learning rate to optimize the performance of the model.

[0063] The method for diagnosing the fault of the electric energy meter measurement data by the AdaBoost algorithm module also includes:

[0064] (3) Fault diagnosis: After training, the model can be used to perform fault diagnosis on new electricity meter data. The model will determine whether a fault exists based on the input data and give the specific fault type.

[0065] (4) Parallel computing and efficiency improvement: In order to improve training efficiency and shorten network running time, the principle of parallel computing can be adopted and this technology can be used in the implementation of software and hardware to accelerate the training process.

[0066] Aiming at the diversity measurement problem and over-adaptation problem between weak classifiers under the AdaBoost algorithm, experiments were conducted on the UCI dataset and the Pearson correlation coefficient was used to quantitatively calculate the correlation between diversity and test error. The specific steps are as follows:

[0067] The Pearson correlation coefficient between two variables is defined as the quotient of the covariance and the standard deviation between the two variables:

[0068] The above formula defines the overall correlation coefficient, which is usually represented by the Greek lowercase letter ρ. By estimating the covariance and standard deviation of the sample, we can get the Pearson correlation coefficient, which is usually represented by the English lowercase letter r:

[0069]

[0070] r can also be estimated by the mean of the standard scores of the sample points, and the equivalent expression is obtained as follows:

[0071] In summary, the invention proposes an improved AdaBoost algorithm for analyzing the fault of electric energy meter measurement data: the invention combines the Internet of Things technology, big data technology and data communication technology together to construct a detection system architecture system, realizes the sensing, transmission, analysis, calculation of the fault data of electric energy meter measurement detection and the monitoring of the upper data, so that the electric energy management user can monitor the fault information of electric energy meter measurement remotely, online and in real time. Different from the conventional technology, the invention has a relatively high level of intelligence and a much stronger ability to process data, which is conducive to users to quickly obtain the detection information of electric energy meter from the complex multi-dimensional data information;

[0072] The present invention can organically integrate the big data fusion calculation module, the big data classification algorithm module and the AdaBoost algorithm module to form an improved AdaBoost algorithm, and can classify and integrate a large amount of data information, so as to facilitate the centralized management of data information of different latitudes and types, thereby effectively classifying data information of different attributes and realizing big data classification, which is beneficial for users to read, trace and manage classified data, and finally realize data fault analysis through the AdaBoost algorithm model, thereby improving the management capability of electric energy meter measurement data;

[0073] The method of the present invention is applied to the classification and discrimination of typical faults of electronic electric energy meters and rectifier bridge faults, and the fault data generated by simulation operation is used as the processing object of the BP-AdaBoost algorithm. The experimental results show that compared with the single BP neural network fault prediction method, the BP-AdaBoost fault prediction method can significantly improve the prediction accuracy and reduce the error, and has certain feasibility. In addition, by conducting experiments on the UCI data set, the Pearson correlation coefficient is used to quantitatively calculate the correlation between diversity and test error. It is found that the change pattern of the double error metric on different data sets is fixed, and it is basically unchanged in the later iteration period and tends to be stable. The test error of this improved AdaBoost algorithm is reduced by an average of 1.5 percentage points, and can be reduced by up to 4.8 percentage points, thereby further improving the classification performance. This improved AdaBoost algorithm combined with the BP neural network electric energy meter measurement data fault analysis method provides an effective solution for fault detection of electric energy meters by improving prediction accuracy and reducing errors. At the same time, through the improvement of the AdaBoost algorithm, the classification performance is further improved, making it more reliable and effective in practical applications.

[0074] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0075] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for analyzing faults in electric energy meter measurement data based on an improved AdaBoost algorithm, characterized in that: The main steps include: Step 1: Data collection and measurement: Obtain the data information of the electric energy meter recorded and detected by the operator; measure and collect the electric energy data of the sub-meter through the low-voltage IoT sensing terminal, combine the voltage and current curves of the power consumption information collection system with Kirchhoff's voltage and current law, and recursively obtain the voltage and current amplitude of the nearest meter box node with the same phase as the electric energy meter to be checked; Step 2: Data information preprocessing: screening and sampling the electric energy meter data obtained in step 1, and then outputting pure data information to achieve preprocessing of the electric energy meter data information; Step 3: Data transmission: receiving and transmitting the electric energy meter data information processed in step 2 by wired or wireless communication, and transmitting the data information; Step 4: Data processing: Process the collected data information related to the energy meter to prepare for subsequent analysis. According to the phase information of the power consumption information collection system, determine the energy meter information to be verified, and calculate, process and transmit the integrated energy meter data information; The received electric energy meter data information is fused through big data fusion calculation; the fused electric energy meter data information is classified using the BP neural network algorithm; and then the AdaBoost algorithm The classified electric energy meter data information is diagnosed and the diagnostic data information is output; the fault injection point and observation point are selected through the electronic electric energy meter simulation model built by Simulink, and the fault data generated by the simulation operation is used as the processing object of the BP-AdaBoost algorithm; Aiming at the output diagnosis of electric energy meter data information between weak classifiers under AdaBoost algorithm, experiments were conducted on UCI dataset, and the correlation between diversity and test error was quantitatively calculated using Pearson correlation coefficient. Step 5: Identify and troubleshoot suspected electricity theft meters: Analyze and determine whether there are suspected electricity theft meters, mark them, and eliminate high-loss points to avoid affecting the evaluation of the overall electricity meter collection data; Step 6: Data analysis: used to evaluate the accuracy of the energy meter. Combined with the voltage and current curve data of the power consumption information collection system and the number of energy meters with the same phase in the same box, the corresponding formula is used for multivariate linear regression analysis to obtain the operating error of the energy meter to be verified. Step 7: Statistical comparison of data records: used to count and record the error data of the electric energy meter to be checked obtained in step 6, so as to facilitate subsequent operators to compare and analyze multiple groups of data and obtain the results of fault analysis of the electric energy meter measurement data; Step 8: After determining that there are electric energy meters with inaccurate measurement, select some of the electric energy meters to be inspected from multiple electric energy meters to be inspected in turn according to a preset selection strategy, and re-test the selected electric energy meters for inaccuracy, re-verify, and obtain the test results to avoid detection errors; Step 9: Marking and reporting inaccurate meters: The meters with sub-meter errors exceeding the specified threshold are marked as inaccurate meters. When the operating error of the electric energy meter to be verified is greater than the allowable threshold, it is identified as an inaccurate meter, and the data of the inaccurate meter is actively reported to the main station through the acquisition terminal; Step 10: Data application: reporting the metering fault data of the electric energy meter re-detected and verified in step 8, for storing, using or transmitting the data processed by the data fusion computing layer, for query, tracing or management by the superior central management computer.

2. The method for analyzing faults in electric energy meter measurement data based on an improved AdaBoost algorithm according to claim 1 is characterized in that: In step 4, the big data fusion calculation method is: filtering data from the received original data; and constructing a data fusion function.

3. The method for analyzing faults in electric energy meter measurement data based on an improved AdaBoost algorithm according to claim 2 is characterized in that: In step 4, the big data classification algorithm module is a BP neural network algorithm. The principle of BP neural network algorithm classification is: the BP algorithm is composed of two processes: the forward propagation of the signal and the reverse propagation of the error in the learning process; Since the training of multi-layer feedforward networks often uses the error back propagation algorithm, people often directly call the multi-layer feedforward network a BP network. The BP algorithm consists of two processes: the forward propagation of the signal and the back propagation of the error.

4. The method for analyzing faults in electric energy meter measurement data based on an improved AdaBoost algorithm according to claim 3 is characterized in that: The BP algorithm is specifically divided into the following steps: (1) Initialization, using small random numbers to assign initial values ​​to each weight and threshold. Note that the initial weights and thresholds in the network cannot be completely equal, otherwise the network cannot run from such a structure to a non-equal weight structure; (2) Reading network parameters and training sample sets; (3) Normalization processing; (4) Calculating each sample in the training set; (5) If the accuracy requirements or other exit conditions are met, the training is terminated, otherwise go to step (4) to continue; (6) Result analysis and output.

5. The method for analyzing faults in electric energy meter measurement data based on an improved AdaBoost algorithm according to claim 4 is characterized in that: The method for the AdaBoost algorithm module to perform fault diagnosis on the electric energy meter measurement data is as follows: (1) Use the AdaBoost algorithm to train weak classifiers to form strong classifiers. AdaBoost adjusts the weight of each weak classifier so that the misclassified samples receive a larger weight in subsequent classifications, allowing the classifier to pay more attention to these samples in subsequent training and improve the overall classification accuracy. (2) Training and optimization: Through multiple iterations of training, the weights and parameters of the weak classifier are adjusted to achieve the best classification effect. This process may include steps such as cross-validation and adjusting the learning rate to optimize the performance of the model.

6. The method for analyzing faults in electric energy meter measurement data based on an improved AdaBoost algorithm according to claim 5 is characterized in that: The method for diagnosing faults in electric energy meter measurement data using the AdaBoost algorithm module further includes: (3) Fault diagnosis: After training, the model can be used to perform fault diagnosis on new electricity meter data. The model will determine whether there is a fault based on the input data and give the specific fault type. (4) Parallel computing and efficiency improvement: In order to improve training efficiency and shorten network running time, the principle of parallel computing can be adopted and this technology can be used in the implementation of software and hardware to accelerate the training process.

7. The method for analyzing faults in electric energy meter measurement data based on an improved AdaBoost algorithm according to claim 6 is characterized in that: Aiming at the diversity measurement problem and over-adaptation problem between weak classifiers under the AdaBoost algorithm, experiments were conducted on the UCI dataset, and the Pearson correlation coefficient was used to quantitatively calculate the correlation between diversity and test error. The specific steps are as follows: The Pearson correlation coefficient between two variables is defined as the quotient of the covariance and the standard deviation between the two variables: The above formula defines the overall correlation coefficient, which is usually represented by the Greek lowercase letter ρ. By estimating the covariance and standard deviation of the sample, we can get the Pearson correlation coefficient, which is usually represented by the English lowercase letter r:

8. The method for analyzing faults in electric energy meter measurement data based on an improved AdaBoost algorithm according to claim 7, characterized in that: The r can also be estimated by the standard score mean of the sample points, and the equivalent expression to the above formula is obtained:

Images