Data analysis method of smart electric energy meter based on big data

Abstract

The present invention relates to the field of data processing technology, and more specifically, to a data analysis method for a smart electric energy meter based on big data, the method comprising: inputting an electric energy time series sequence into an LSM algorithm for iterative analysis, obtaining a learning rate and an error value of each electric energy data in an iterative analysis process, and calculating the error fluctuation degree; calculating the noise degree of each electric energy data in the electric energy time series sequence, and recording the difference between the error value and the noise degree as a noise quantity fitting deviation; calculating an adjusted learning rate according to the learning rate, the error fluctuation degree, and the noise quantity fitting deviation; using the adjusted learning rate as the learning rate for the next iterative training, continuing to perform iterative analysis until the error fluctuation degree and the noise quantity fitting deviation are both less than the corresponding preset threshold value, the iteration ends, and an iteratively completed LSM algorithm is obtained; and removing the noise in the electric energy time series sequence using the iteratively completed LSM algorithm. Thus, accurate denoising is achieved by setting an adaptive learning rate.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and more specifically, to a data analysis method for a smart electric energy meter based on big data. Background Art

[0002] As a tool for measuring and recording the consumption of electric energy, the electric energy meter is used in all aspects of life. However, the electric energy meter is easily affected by magnetic fields, temperature, humidity, etc. during the measurement process, resulting in noise data in the measured data. In order to eliminate the interference of noise and obtain accurate electric energy data, the collected electric energy data needs to be denoised.

[0003] The patent application document with publication number CN118013191A shows a denoising method for insulators based on LSM filtering, CEEMDAN and wavelet detection. The method in the document is to first perform preliminary filtering using the LSM filtering algorithm, then use the CEEMDAN algorithm to decompose the data after preliminary filtering, and then use the wavelet filtering algorithm to filter each component to achieve data denoising. The method in the document achieves accurate denoising by combining multiple methods. The calculation process of this method is relatively complicated and will reduce the calculation efficiency.

[0004] As a commonly used filtering algorithm, the LSM filtering algorithm estimates the real data by fitting the data, and removes the noise in the data by analyzing the difference between the data and the estimated real data. Therefore, before the LSM filtering algorithm performs filtering analysis on the data, it is necessary to perform iterative analysis on the data to optimize the fitting parameters, so as to remove the noise in the data by fitting the real data according to the optimized fitting parameters. Therefore, how to obtain accurate optimized fitting parameters through iterative analysis in the LSM filtering algorithm becomes the key to accurate filtering.

[0005] LSM filtering algorithm can optimize fitting parameters by gradient descent method in iterative analysis process, wherein there will be a learning rate in the process of utilizing gradient descent method to optimize fitting parameters, for determining the speed of fitting parameter adjustment. Wherein when the learning rate is set larger, it will cause skipping of optimal fitting parameters, if the learning rate is set smaller, it will cause iteration efficiency to be lower, and it is also possible to fall into local optimum, and optimal fitting parameters cannot be obtained. Therefore how to set the learning rate becomes the research focus of the present invention. Summary of the invention

[0006] In order to solve the problem of how to set the learning rate, the present invention proposes a smart electric energy meter data analysis method based on big data, which includes the following steps:

[0007] Obtaining electrical energy timing sequence;

[0008] The electric energy time series sequence is input into the LSM algorithm for iterative analysis. In one iterative analysis process, the learning rate and the error value of each electric energy data in the electric energy time series sequence are obtained. The error value of all the electric energy data in the electric energy time series sequence in the most recent preset number of iterative analysis processes is used to calculate the error fluctuation degree; the noise degree of each electric energy data in the electric energy time series sequence is calculated, and the difference between the error value and the noise degree is recorded as the noise quantity fitting deviation;

[0009] Calculate the adjusted learning rate: , L0 represents the learning rate, Indicates the amount of electric energy data in the electric energy time series. Indicates the first The noise amount fitting deviation of the electric energy data is Indicates the degree of error fluctuation, represents the adjusted learning rate, Indicates the preset parameters. represents the hyperbolic tangent function;

[0010] The adjusted learning rate is used as the learning rate for the next iterative training, and the iterative analysis is continued until the error fluctuation degree and the noise amount fitting deviation are both less than the corresponding preset thresholds. The iteration ends and the iteratively completed LSM algorithm is obtained. The iteratively completed LSM algorithm is used to remove the noise in the electric energy time series.

[0011] When calculating the learning rate, the present invention introduces the error fluctuation degree and the noise amount fitting deviation index to accurately analyze and judge whether the fitting parameters analyzed in this iteration are close to the optimal solution, so as to set a suitable learning rate to iterate the optimal solution faster; further, by analyzing the difference between the noise degree and the error value, it is judged whether the true value fitted by the fitting parameters in this iteration is accurate, so as to judge whether the fitting parameters are close to the optimal solution, and provide a basis for accurately setting the learning rate; further, by analyzing the error fluctuation degree, it is judged whether the convergence of the fitting error is close to the optimal solution, so as to judge whether the fitting parameters are close to the optimal solution, and provide a basis for accurately setting the learning rate.

[0012] Preferably, the method for obtaining the noise level includes:

[0013] The noise level is calculated, where the noise level is the product of the personalization level and the associated mutation level.

[0014] When calculating the noise level, the present invention takes into account both the personalization level and the correlation mutation level, performs a more comprehensive noise analysis, and reduces the analysis error rate.

[0015] Preferably, the method for obtaining the degree of personalization includes:

[0016] Performing N dimensionality reduction operations on the electric energy time series sequence to obtain a time series sequence after each dimensionality reduction operation, wherein the dimensionality reduction operation includes: a filtering operation and a downsampling operation, and N represents a preset processing number;

[0017] Performing an upsampling operation on the time series sequence after each dimensionality reduction operation to obtain an upsampled time series sequence of the time series sequence after each dimensionality reduction operation;

[0018] The personalization degree is calculated, and the personalization degree is positively correlated with the absolute value of the difference between the data in the up-sampled time series and the corresponding electric energy data in the electric energy time series.

[0019] Preferably, the personalization degree satisfies the relationship:

[0020] ;

[0021] Among them, the The upsampled time series sequence after the dimensionality reduction operation is recorded as the analysis sequence. Indicates the difference between the analysis sequence and the electric energy time series. data, Indicates Electric energy data, Indicates A dimensionality reduction operation, Indicates the first The sum of the absolute values ​​of the differences between the variance of the data within the preset range of the data and the variance of the data within the preset range of all other data. Indicates the first The sum of the absolute values ​​of the differences between the mean of a data within the preset range and the mean of all other data within the preset range. Indicates the analysis sequence with The frequency of occurrence of data with the same value. Indicates The degree of personalization of the power data.

[0022] When analyzing the degree of personalization, the present invention accurately determines the details regularity contained in each electric energy data by introducing the difference between the upsampled time series sequence and the original electric energy time series sequence after each dimensionality reduction operation, thereby accurately determining the personalization of each electric energy data; further, by introducing the distribution difference between the data within a preset range of each data and the data within a preset range of other data in the upsampled time series sequence after each dimensionality reduction operation, the difference between the data and other data is accurately determined, thereby accurately determining the degree of personalization.

[0023] Preferably, the preset range is a data segment consisting of data within a preset sliding window centered on each data.

[0024] Preferably, the method for obtaining the degree of associated mutation includes:

[0025] Arrange the current data at the time corresponding to each electric energy data in the electric energy time series in time series to obtain a current time series;

[0026] Calculate the correlation between the data within the preset range of each electric energy data in the electric energy time series and the data of the corresponding time period in the current time series, and record it as the correlation of each electric energy data; take the absolute value of the difference between the correlation of each electric energy data and the correlation of the previous electric energy data as the correlation change of each electric energy data;

[0027] The degree of association mutation is calculated, and the degree of association mutation is positively correlated with the amount of association change.

[0028] Preferably, the degree of associated mutation satisfies the relationship:

[0029] ;

[0030] in, Indicates The correlation change of electric energy data, Represents the mean value of the correlation change of all electric energy data, Represents the standard deviation of the correlation variation of all electric energy data, Indicates The correlation mutation degree of each electric energy data.

[0031] When analyzing the degree of correlation mutation, the present invention accurately reflects whether there is a mutation in the correlation between each electric energy data and the current data by introducing the difference between the correlation between each electric energy data and the current data compared with the correlation between the previous electric energy data and the current data, thereby providing a basis for accurately analyzing the existence of noise at the location of each electric energy data.

[0032] Preferably, the error fluctuation degree includes:

[0033] The mean error value of all electric energy data in the electric energy time series in each iterative analysis process is recorded as the comprehensive error in each iterative analysis process, the absolute value of the difference between the comprehensive error in each iterative analysis process and the comprehensive error in the previous iterative process is recorded as the comprehensive error change in each iterative analysis process, and the mean of the comprehensive error change in a preset number of iterative processes is taken as the error fluctuation degree in this iterative analysis process.

[0034] The present invention uses the mean of the error value changes in a preset number of iterations as the error fluctuation degree to prevent the iteration from being stopped due to a temporary small error change, thereby effectively reducing the risk of falling into a local optimum.

[0035] The present invention has the following beneficial effects:

[0036] When calculating the learning rate, the present invention introduces the error fluctuation degree and the noise amount fitting deviation index to accurately analyze and judge whether the fitting parameters analyzed in this iteration are close to the optimal solution, so as to set a suitable learning rate to iterate the optimal solution faster;

[0037] Furthermore, by analyzing the difference between the noise level and the error value, it is determined whether the true value of the fitting parameters of this iteration is accurate, thereby determining whether the fitting parameters are close to the optimal solution, which provides a basis for accurately setting the learning rate; further, by analyzing the error fluctuation degree, it is determined whether the fitting error converges, thereby determining whether the fitting parameters are close to the optimal solution, which provides a basis for accurately setting the learning rate;

[0038] Furthermore, when calculating the noise level, a personalized index is introduced. This index is calculated based on the characteristics that the noise data does not conform to the overall change law. Therefore, the personalized index obtained in this way can accurately analyze the noise content in each electric energy data.

[0039] Furthermore, when calculating the noise level, by introducing the correlation mutation degree index, the index is calculated based on the fact that the noise data does not conform to the correlation characteristics between the electric energy data and the current data. Therefore, the correlation mutation degree index obtained in this way can accurately analyze the noise content in each electric energy data. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0041] Figure 1 It is a flowchart of the steps of a smart electric energy meter data analysis method based on big data in an embodiment of the present invention. DETAILED DESCRIPTION

[0042] 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 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 those skilled in the art without creative work are within the scope of protection of the present invention.

[0043] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0044] See also Figure 1, which shows a flowchart of a method for analyzing smart electric energy meter data based on big data provided by an embodiment of the present invention, the method comprising the following steps:

[0045] S1: Obtain the electric energy timing sequence and the current timing sequence.

[0046] Specifically, the electric energy data of a circuit at each moment is collected by electric energy, and the current data of the same circuit at the corresponding moment is collected by an ammeter, and the electric energy data collected at Q moments are arranged in time sequence to obtain an electric energy time series sequence, and the current data collected at Q moments are arranged in time sequence to obtain a current time series sequence. Q represents a preset time length, and in this embodiment, Q is 200, and other embodiments may take other values, which are not specifically limited in this embodiment.

[0047] S2: Input the electric energy time series sequence into the LSM algorithm for iterative analysis. During one iterative analysis process, obtain the learning rate and the error value of each electric energy data in the electric energy time series sequence, and calculate the error fluctuation degree using the average error value of all electric energy data in the electric energy time series sequence during the most recent preset number of iterative analysis processes.

[0048] It should be noted that the learning rate should be set to be close to the optimal solution. Since it is impossible to know which solution is the optimal solution, it can only be determined based on other factors. Since the optimal solution in the LSM algorithm should be the fitting parameter of the algorithm that fits the true value, when the fitting parameter is close to the optimal solution, the fitted value is closer to the true value. Since the true value is the value after removing the noise, the error value is the difference between the electric energy data and the fitting value. Therefore, when the fitted value is closer to the true value, the error value should be equal to the noise data. Therefore, the noise content in each electric energy data can be analyzed, and the difference between the analyzed noise content and the error value can be used to determine whether the fitting parameter is close to the optimal value.

[0049] It should be further explained that when the fitting parameters are close to the optimal solution, the error value will gradually converge; therefore, the closeness of the fitting parameters to the optimal solution can also be determined by analyzing the convergence of the error value. In this embodiment, the error fluctuation degree reflects the convergence of the error.

[0050] Preferably, as an example, the electric energy time series sequence is input into the LSM algorithm for iterative analysis. In one iterative analysis process, the learning rate and the error value of each electric energy data in the electric energy time series sequence are obtained, and the error fluctuation degree is calculated using the average error value of all electric energy data in the electric energy time series sequence in the most recent preset number of iterative analysis processes, including:

[0051] The mean error value of all electric energy data in the electric energy time series in each iterative analysis process is recorded as the comprehensive error in each iterative analysis process, the absolute value of the difference between the comprehensive error in each iterative analysis process and the comprehensive error in the previous iterative process is recorded as the comprehensive error change in each iterative analysis process, and the mean of the comprehensive error change in a preset number of iterative processes is taken as the error fluctuation degree in this iterative analysis process.

[0052] It should be noted that inputting the electric energy time series into the LSM algorithm for iterative analysis is a prior art and will not be described in detail here.

[0053] It should be further explained that the learning rate and the error value are data calculated using the LSM algorithm, so the learning rate and the error value can be directly obtained.

[0054] S3: Calculate the noise level of each electric energy data in the electric energy time series, and record the difference between the error value and the noise level as the noise quantity fitting deviation.

[0055] It should be noted that in order to analyze the situation where the fitting parameters are close to the optimum, the difference between the error value and the noise content needs to be analyzed. The situation where each power data contains noise data is first analyzed below. This embodiment uses the noise level to reflect the situation where each power data contains noise data.

[0056] It should be further explained that noise data generally does not conform to the variation law of real data, and the variation of electric energy data will be affected by the variation of current data, so the variation of electric energy data should be correlated with the variation of current data, while the noise data does not conform to this correlation, so the noise degree can be analyzed according to these two characteristics. In this embodiment, the degree of personalization is used to reflect the conformity of each electric energy data to the overall variation law, and the degree of correlation mutation is used to reflect the difference between the correlation between each electric energy data and current data and the correlation between other electric energy data and current data.

[0057] Preferably, as an example, calculating the noise level of each electric energy data in the electric energy time series includes:

[0058] The noise level is the product of the personalization level and the associated mutation level as the noise level of each electric energy data in the electric energy time series.

[0059] The method for obtaining the degree of personalization includes:

[0060] Performing N dimensionality reduction operations on the electric energy time series sequence to obtain a time series sequence after each dimensionality reduction operation, wherein the dimensionality reduction operation includes a filtering operation and a downsampling operation, and N represents a preset processing number;

[0061] The time series after each dimensionality reduction operation is upsampled to obtain an upsampled time series sequence with the same length as the electric energy time series sequence; in this embodiment, a Gaussian filtering algorithm is used for filtering operation, and other embodiments may use other filtering methods, which are not specifically limited in this embodiment.

[0062] It should be noted that the filtering operation, up-sampling and down-sampling operations are all existing technologies and will not be described in detail here.

[0063] The relationship for calculating the degree of personalization is:

[0064] ;

[0065] Among them, the The upsampled time series sequence after the dimensionality reduction operation is recorded as the analysis sequence. Indicates the difference between the analysis sequence and the electric energy time series. data, Indicates The analysis sequence is obtained by filtering and up-sampling and down-sampling the electric energy time series. The filtering and up-sampling and down-sampling processing will remove the details of the electric energy time series. Therefore, the analysis sequence contains the main regularity information. Therefore, the difference between the electric energy time series and the analysis sequence reflects the details. The larger the difference, the greater the degree of personalization. The larger the value, the greater the degree of personalization. Indicates A dimensionality reduction operation, Indicates the first The sum of the absolute values ​​of the differences between the variance of the data within the preset range of the data and the variance of the data within the preset range of all other data. Indicates the first The sum of the absolute values ​​of the differences between the mean of a data within the preset range and the mean of all other data within the preset range. Indicates the analysis sequence with The frequency of occurrence of data with the same value. Reflects the analysis sequence The difference between the data distribution within the preset range of the data and the data distribution within the preset range of other data is larger. The greater the difference between the data and other data within the preset range of the first data, the greater the difference between the first data and the second data. The more personalized the data is, the more filtering and downsampling times are, which means the information contained in the analysis sequence is more likely to be the main regular information. When the data within the preset range of the data is different from the data within the preset range of other data, the higher the degree of personalization of the data, the more times the dimensionality reduction will be. Weight it for weight, Indicates The personalization degree of each electric energy data, | | represents the absolute value symbol.

[0066] In addition, the method for obtaining the degree of associated mutation includes:

[0067] Calculate the correlation between the data within the preset range of each electric energy data in the electric energy time series and the data of the corresponding time period in the current time series, and record it as the correlation of each electric energy data; take the absolute value of the difference between the correlation of each electric energy data and the correlation of the previous electric energy data as the correlation change of each electric energy data;

[0068] The relationship for calculating the degree of associated mutation is:

[0069] ;

[0070] in, Indicates The correlation change of electric energy data, Represents the mean value of the correlation change of all electric energy data, Represents the standard deviation of the correlation variation of all electric energy data, Reflects the The difference between the correlation change of the first electric energy data and the correlation change of other electric energy data. The larger the value, the greater the difference. The change of the first power data is greater than that of other power data. The power data has a large mutation rate. Indicates The correlation mutation degree of each electric energy data.

[0071] The above embodiments also involve preset ranges and correlations, and the method for determining the preset ranges and correlations needs to be described below.

[0072] The preset range is the data segment consisting of the data in the preset sliding window centered on each data; the preset window is Window, Indicates the preset window length. The value 11 is taken as an example for description, and other embodiments may take other values, and this implementation does not impose any specific limitation.

[0073] The correlation is the absolute value of the Pearson correlation coefficient between the current time series and the electric energy time series.

[0074] S4: Calculate the adjusted learning rate.

[0075] It should be noted that since the noise fitting deviation and error fluctuation can both reflect whether the fitting parameters are close to the optimal solution, the learning rate needs to be adjusted based on these two indicators.

[0076] Preferably, as an example, the adjusted learning rate satisfies the relationship:

[0077]

[0078] in, represents the learning rate, Indicates the amount of electric energy data in the electric energy time series. Indicates the first The larger the value is, the less close the data fitted by the fitting parameters are to the true value, indicating that the fitting parameters are not accurate enough, and thus the fitting parameters are far from the optimal solution. It reflects the accuracy of fitting the true values ​​of all electric energy data in the entire electric energy time series. The larger the value, the farther away from the optimal solution, the learning rate should be increased. At this time, the adjustment value of the learning rate is positive. The smaller the value, the closer to the optimal solution, the learning rate should be reduced. At this time, the adjustment value of the learning rate is negative. In this embodiment, the function The output rate adjustment value is normalized to between -1 and 1. Indicates the degree of error fluctuation. The larger the value, the more the error value does not converge and the fitting parameters are far from the optimal solution. The smaller the value, the more the error value converges and the fitting parameters may be close to the optimal solution or may fall into the local optimum. The noise content fitting deviation is combined to make a judgment. When the noise content fitting deviation is also small, it means that the fitting parameters are close to the optimal solution and the learning rate should be reduced. When the noise content fitting parameters are large, it means that the fitting parameters may fall into the local optimum and the learning rate needs to be increased to jump out of the local optimum. represents the adjusted learning rate, Indicates the preset parameters. In this embodiment, Take 2, The description is made by taking 0.2 as an example, and other embodiments may take other values, which are not specifically limited in this embodiment. Represents the hyperbolic tangent function, which is used for normalization.

[0079] S5: Use the adjusted learning rate as the learning rate for the next iterative training, and continue to perform iterative analysis until the error fluctuation degree and the noise amount fitting deviation are both less than the corresponding preset thresholds. The iteration ends and an iteratively completed LSM algorithm is obtained; the iteratively completed LSM algorithm is used to remove the noise in the electric energy time series.

[0080] In this embodiment, the preset threshold corresponding to the error fluctuation degree is 1.1, and the preset threshold corresponding to the noise amount fitting deviation is 0.8. Other embodiments may take other values, and this embodiment does not make specific limitations.

[0081] It should be noted that the use of an iterative LSM algorithm to remove noise from an electric energy time series is a prior art and will not be described in detail here.

[0082] At this point, this embodiment is completed.

[0083] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A data analysis method for smart electric energy meters based on big data, characterized in that: include: Obtaining electrical energy timing sequence; The electric energy time series sequence is input into the LSM algorithm for iterative analysis. In one iterative analysis process, the learning rate and the error value of each electric energy data in the electric energy time series sequence are obtained, and the error fluctuation degree is calculated using the error values ​​of all electric energy data in the electric energy time series sequence in the most recent preset number of iterative analysis processes; Calculate the noise level of each electric energy data in the electric energy time series, and record the difference between the error value and the noise level as the noise quantity fitting deviation; Calculate the adjusted learning rate: , represents the learning rate, Indicates the amount of electric energy data in the electric energy time series. Indicates the first The noise amount fitting deviation of the electric energy data is Indicates the degree of error fluctuation, represents the adjusted learning rate, All represent preset parameters. represents the hyperbolic tangent function; The adjusted learning rate is used as the learning rate for the next iterative training, and the iterative analysis is continued until the error fluctuation degree and the noise amount fitting deviation are both less than the corresponding preset thresholds. The iteration ends and the iteratively completed LSM algorithm is obtained. The iteratively completed LSM algorithm is used to remove the noise in the electric energy time series.

2. The method for analyzing smart electric energy meter data based on big data according to claim 1, characterized in that: The method for obtaining the noise level includes: The noise level is calculated, where the noise level is the product of the personalization level and the associated mutation level.

3. The method for analyzing data of smart electric energy meters based on big data according to claim 2 is characterized in that: The method for obtaining the personalization degree includes: Performing N dimensionality reduction operations on the electric energy time series sequence to obtain a time series sequence after each dimensionality reduction operation, wherein the dimensionality reduction operation includes: a filtering operation and a downsampling operation, and N represents a preset processing number; Performing an upsampling operation on the time series sequence after each dimensionality reduction operation to obtain an upsampled time series sequence of the time series sequence after each dimensionality reduction operation; The personalization degree is calculated, and the personalization degree is positively correlated with the absolute value of the difference between the data in the up-sampled time series and the corresponding electric energy data in the electric energy time series.

4. The method for analyzing data of smart electric energy meters based on big data according to claim 3 is characterized in that: The degree of personalization satisfies the relationship: ； Among them, the The upsampled time series sequence after the dimensionality reduction operation is recorded as the analysis sequence. Indicates the difference between the analysis sequence and the electric energy time series. data, Indicates Electric energy data, Indicates A dimensionality reduction operation, Indicates the first The sum of the absolute values ​​of the differences between the variance of the data within the preset range of the data and the variance of the data within the preset range of all other data. Indicates the first The sum of the absolute values ​​of the differences between the mean of a data within the preset range and the mean of all other data within the preset range. Indicates the analysis sequence with The frequency of occurrence of data with the same value. Indicates The degree of personalization of the power data.

5. The method for analyzing data of smart electric energy meters based on big data according to claim 4 is characterized in that: The preset range is a data segment consisting of data within a preset sliding window centered on each data.

6. The method for analyzing smart electric energy meter data based on big data according to claim 2, characterized in that: The method for obtaining the degree of associated mutation includes: Arrange the current data at the time corresponding to each electric energy data in the electric energy time series in time series to obtain a current time series; Calculate the correlation between the data within the preset range of each electric energy data in the electric energy time series and the data of the corresponding time period in the current time series, and record it as the correlation of each electric energy data; take the absolute value of the difference between the correlation of each electric energy data and the correlation of the previous electric energy data as the correlation change of each electric energy data; The degree of association mutation is calculated, and the degree of association mutation is positively correlated with the amount of association change.

7. The method for analyzing smart electric energy meter data based on big data according to claim 6, characterized in that: The degree of associated mutation satisfies the relationship: ； in, Indicates The correlation change of electric energy data, Represents the mean value of the correlation change of all electric energy data, Represents the standard deviation of the correlation variation of all electric energy data, Indicates The correlation mutation degree of each electric energy data.

8. The method for analyzing smart electric energy meter data based on big data according to claim 1, characterized in that: The error fluctuation degree includes: The mean error value of all electric energy data in the electric energy time series in each iterative analysis process is recorded as the comprehensive error in each iterative analysis process, the absolute value of the difference between the comprehensive error in each iterative analysis process and the comprehensive error in the previous iterative analysis process is recorded as the comprehensive error change in each iterative analysis process, and the mean of the comprehensive error change in a preset number of iterative processes is taken as the error fluctuation degree in this iterative analysis process.

Images