A basic error estimation method and system for smart electric energy meter under data missing condition

Abstract

The application discloses a kind of data missing condition under intelligent electric energy meter basic error estimation method and system, the application includes collecting the original day data of intelligent electric energy meter, after data cleaning and normalization, day data is fitted using trigonometric function interpolation method FFI to complete missing value filling, calculate monthly average basic error according to the day data after completing missing value filling;Time and environmental stress data are used as input, and monthly average basic error is used as output to divide data set, and least square support vector machine LSSVM is constructed;Improved sparrow search algorithm ISSA is used to optimize parameters of least square support vector machine LSSVM;Data set is used to train least square support vector machine LSSVM.The application aims at the intelligent electric energy meter under data missing condition to realize intelligent electric energy meter basic error estimation, to realize higher prediction accuracy and operating efficiency.

Description

Technical Field

[0001] This invention relates to the field of reliability prediction for electricity metering equipment, and specifically to a method and system for predicting the basic error of smart meters under conditions of missing data. Background Technology

[0002] With the development of smart grids, new types of electricity metering equipment are increasingly being used as the basic units for industrial and residential energy metering. Smart meters are not only a link connecting households but also a key node reflecting the State Grid's high-quality service and ensuring fair and impartial metering principles. Before leaving the factory, smart meters and data acquisition terminals undergo various rigorous routine and extended tests, possessing a relatively complete inspection standard system to ensure product quality meets technical requirements and improves product reliability. However, in complex natural and electrical environments, smart meters and data acquisition terminals still lack corresponding testing methods and evaluation standards, resulting in weaknesses in operational reliability and stability.

[0003] To address the various extreme natural climates that may exist for electricity metering equipment, and based on natural environments characterized by extreme cold, high dry heat, high humidity, and high smog, the first real-time operation of smart meters under full load in all climates can be achieved through outdoor smart meter test bases established in multiple provinces. Based on the operating data of smart meters under different stress conditions, a reasonable and effective basic error prediction model for smart meters can be established. In recent decades, data-driven methods have been widely used in equipment degradation modeling and reliability analysis. Data-driven methods can be divided into two main categories: probabilistic methods and deterministic methods. Among them, deterministic methods can effectively solve the problem of multi-source information fusion, providing accurate evaluation results for equipment reliability analysis. Currently, deterministic methods have achieved their expected goals, but some issues still need to be considered. For example, in the extreme natural environments of the test bases, smart meters are prone to data acquisition terminal failures, communication equipment failures, interference, etc., resulting in missing basic error data, which in turn affects the accuracy of the basic error prediction model. Neural network models and least squares support vector machines do not rely on prior selection and are suitable for predicting equipment performance degradation under multiple stress conditions. Least squares support vector machines (LSVMs) have a significant advantage under conditions of small sample sizes and nonlinear regression. However, in extreme environments, smart meters suffer from severe data gaps in their original degraded data. This not only fails to guarantee the continuity of data over time but also reduces the model's prediction accuracy, making it impossible to further explore the degradation trend of the basic error data of smart meters over time.

[0004] In conclusion, to further optimize the smart meter rotation mechanism, it is imperative to analyze the degradation trend of smart meters under extreme natural environments over time. Therefore, for smart meters in extreme natural environments with severe data gaps, further exploration and research are needed on relevant data filling algorithms and basic error prediction methods for smart meters. Summary of the Invention

[0005] The technical problem to be solved by this invention is to provide a method and system for predicting the basic error of smart meters under data loss conditions, in order to achieve higher prediction accuracy and operating efficiency.

[0006] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: A method for predicting the basic error of a smart meter under conditions of missing data includes: Step S101: Collect the original daily data of the smart energy meter, including time and feature data and basic errors; Step S102: Clean and normalize the original daily data of the smart energy meter; Step S103: Use the trigonometric interpolation method FFI to fit the daily data to complete the missing value filling, and calculate the monthly average basic error based on the daily data after missing value filling. Step S104: Divide the dataset using time and environmental stress data as input and monthly average basic error as output, and construct a least squares support vector machine (LSSVM). Step S105: The improved Sparrow Search Algorithm (ISSA) is used to optimize the parameters of the Least Squares Support Vector Machine (LSSVM). Step S106: Train the least squares support vector machine (LSSVM) using the dataset; Step S107: Determine whether the Least Squares Support Vector Machine (LSSVM) meets the requirements. If it does, end and exit; otherwise, jump to step S105 to continue iterating.

[0007] Optionally, the feature data in step S101 may include some or all of the ambient temperature, humidity, air pressure, wind speed, ultraviolet radiation, and light intensity.

[0008] Optionally, when using the trigonometric interpolation method FFI to fit the daily data and complete the missing value imputation in step S103, it includes first fitting the daily data using the sine superposition function shown in the following formula: f ( x )= a 1*sin( b 1* x+ c 1)+ a 2*sin( b 2* x + c 2)+ a 3*sin( b 3* x + c 3)+ a 4*sin( b 4* x + c 4)+ a 5*sin( b 5* x + c 5)+ a 6*sin( b 6* x + c 6)+ a 7*sin( b 7*x+ c 7)+ a 8*sin( b 8*x+ c 8), In the above formula, f ( x () is a superposition of sinusoidal functions. a 1~ a 8. b 1~ b 8 and c 1~ c 8 is the fitting coefficient. x This refers to the input data.

[0009] Optionally, in step S104, when constructing the Least Squares Support Vector Machine (LSSVM), the decision function used by the LSSVM has the following expression: , , In the above formula, For input data The output mapped to a higher-dimensional space, For Lagrange multipliers, For kernel function, For the first in the dataset i One input data, b The hyperplane intercept is... For kernel function The width factor; and the optimization objective and constraints adopted by the least squares support vector machine (LSSVM) are: , In the above formula, ω Let be the normal vector of the hyperplane. is the regularization parameter, and c> 0; For the first in the dataset i The relaxation factor for each data point For the first in the dataset i One output data, For the first in the dataset i One input data, The number of data points in the dataset. It is a nonlinear function.

[0010] Optionally, when step S105 uses the Improved Sparrow Search Algorithm (ISSA) to optimize the parameters of the Least Squares Support Vector Machine (LSSVM), the optimized parameters include the regularization parameters of the LSSVM. c and the width factor of the kernel function σ .

[0011] Optionally, step S105 includes: Step S201, regularization parameters c and the width factor of the kernel function σ Consider the population as the improved sparrow search algorithm ISSA, and initialize the number of iterations t, the maximum number of iterations Max_t, and the number of sparrows pop. Step S202: Initialize the sparrow population using the Piecewise chaotic mapping based on the following formula: , In the above formula, P ∈(0,1) is a control parameter; ∈(0,1) and ∈(0,1) is a chaotic variable. This indicates the updated position of the sparrow. This indicates the position of the sparrows before the update; Step S203: Calculate the fitness value of individual sparrows and sort them according to the fitness value; Step S204: Update the positions of discoverers, joiners, and watchers in the sparrow population; Step S205: Update the position of the optimal sparrow individual using Gaussian and chaotic perturbations: , In the above formula, The updated sparrow position. The position of the sparrow before the update. It follows a Gaussian distribution; Step S206: Determine whether the iteration number t equals the maximum iteration number Max_t. If it does, output the regularization parameters of the Least Squares Support Vector Machine (LSSVM). c and the width factor of the kernel function σ If the condition is met, proceed to step S106; otherwise, proceed to step S203.

[0012] Optionally, in step S107, determining whether the least squares support vector machine (LSSVM) meets the requirements means whether the number of iterations is equal to the preset maximum number of iterations, and when jumping to step S105 to continue iterating, it also includes incrementing the number of iterations by 1.

[0013] Furthermore, the present invention also provides a basic error prediction system for smart meters under data missing conditions, including a microprocessor and a memory interconnected, wherein the microprocessor is programmed or configured to execute the basic error prediction method for smart meters under data missing conditions.

[0014] Furthermore, the present invention also provides a computer-readable storage medium storing a computer program / instructions that are programmed or configured to execute, via a processor, the basic error prediction method for smart meters under data loss conditions.

[0015] In addition, the present invention also provides a computer program product, including a computer program / instruction, which is programmed or configured to execute the basic error prediction method for smart meters under data loss conditions by a processor.

[0016] Compared with existing technologies, this invention has the following main advantages: This invention proposes a basic error prediction method for smart meters based on an improved sparrow search optimization least squares support vector machine using trigonometric interpolation. First, it combines smart meter performance data with on-site operating environment data, and uses function fitting interpolation (FFI) to fill in missing daily data, thereby obtaining better monthly average data, which is then used as the input feature of the least squares support vector machine (LSSVM). Subsequently, an improved sparrow search algorithm (ISSA) is established by combining chaotic mapping and Gaussian mutation methods to achieve adaptive optimization of the LSSVM prediction model parameters. Finally, an example analysis of real-time basic error data from smart meters in a typical high-temperature and dry environment test base in Xinjiang shows that the prediction model proposed in this invention has higher prediction accuracy and operating efficiency than currently popular prediction methods. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the basic process of the method in an embodiment of the present invention.

[0018] Figure 2 This refers to the raw daily data collected in this embodiment of the invention.

[0019] Figure 3 This refers to the daily data after missing value imputation in this embodiment of the invention.

[0020] Figure 4 This is a schematic diagram illustrating the basic process of optimizing parameters using the improved Sparrow Search Algorithm (ISSA) in an embodiment of the present invention.

[0021] Figure 5 This example compares the evaluation metrics of different models. Detailed Implementation

[0022] like Figure 1 As shown, the basic error prediction method for smart meters under data loss conditions in this embodiment includes: Step S101: Collect the original daily data of the smart energy meter, including time and feature data and basic errors; Step S102: Clean and normalize the original daily data of the smart energy meter; Step S103: Use the trigonometric interpolation method FFI to fit the daily data to complete the missing value filling, and calculate the monthly average basic error based on the daily data after missing value filling. Step S104: Divide the dataset using time and environmental stress data as input and monthly average basic error as output, and construct a least squares support vector machine (LSSVM). Step S105: The improved Sparrow Search Algorithm (ISSA) is used to optimize the parameters of the Least Squares Support Vector Machine (LSSVM). Step S106: Train the least squares support vector machine (LSSVM) using the dataset; Step S107: Determine whether the Least Squares Support Vector Machine (LSSVM) meets the requirements. If it does, end and exit; otherwise, jump to step S105 to continue iterating.

[0023] Step S101 is used to collect the original daily data of the smart energy meter, including time, feature data, and basic error. In this embodiment, the high-dry-heat outdoor smart energy meter laboratory established by the State Grid in Xinjiang is used as an example to explore and study the operating characteristics and reliability of the smart energy meter under high-dry-heat conditions. The typical operating environment laboratory collects and records various environmental stress data and equipment operating data through an environmental monitoring system and a smart energy meter monitoring system, respectively. The data is then compared and calculated with indoor standard metering equipment data to obtain the metering error. Subsequently, the laboratory uploads the collected and processed data to the base station server and database for storage via a network switch or GPRS cloud. The basic error of a company's smart energy meter from October 1, 2020 to March 31, 2022 is selected from the experimental base for visualization to observe data loss. The time data in the original data is stored in year / month / day format, quantified with the day as the minimum time interval, with the processed time data as the horizontal axis and the basic error value of the smart energy meter as the vertical axis. As an optional implementation, the feature data in step S101 of this embodiment includes ambient temperature (…). T / ℃), humidity ( R / %RH), air pressure ( P / %hPa), wind speed ( W / (m / s)), ultraviolet ( U / (mW / m 2 )),illumination( V / Lux). Taking the State Grid's high-temperature and hot natural outdoor smart meter laboratory in Xinjiang as an example, basic error data of smart meters from Company A were selected from October 1, 2020 to March 31, 2022, along with the corresponding temperature data ( T / ℃), humidity ( R / %RH), air pressure ( P / %hPa), wind speed ( W / (m / s)), ultraviolet ( U / (mW / m 2 )),illumination( V Environmental stress data such as / Lux were cleaned and normalized, and then missing values ​​in the basic error and environmental stress were filled using the FFI method to obtain a better monthly average dataset. D’ ={ x i , y i}( i =1,2,···, n ),in x =[ T , R , P , W ,U , V , t The stress combination that affects the basic error of a smart energy meter. y = [ E ] This refers to the basic error of a smart energy meter.

[0024] In extreme natural environments, severe data gaps exist. It should be noted that "data gaps" here refers to a relative insufficiency of data, which can also be understood as the equivalent of a small sample size within a large dataset. Therefore, this embodiment proposes a trigonometric function interpolation (FFI) method to fit daily data, complete missing value imputation, and obtain monthly average basic error data. Specifically, in step S103, when using the FFI method to fit daily data and complete missing value imputation, it includes first fitting the daily data using the following sine superposition function: f ( x )= a 1*sin( b 1* x + c 1)+ a 2*sin( b 2* x + c 2)+ a 3*sin( b 3* x + c 3)+ a 4*sin( b 4* x + c 4)+ a 5*sin( b 5* x + c 5)+ a 6*sin( b 6* x + c 6)+ a 7*sin( b 7*x+ c 7)+ a 8*sin( b 8*x+ c 8), In the above formula, f ( x () is a superposition of sinusoidal functions. a 1~ a 8. b 1~ b 8 andc 1~ c 8 is the fitting coefficient. x The input data is denoted as . The above function expression can also be represented as: , In the above formula, M The number of terms in the sine function is determined by the fitting effect of the original daily data; The amplitude is determined by the maximum and minimum values ​​of the original daily data; It is the angular frequency, determined by the harmonic components contained in the original daily data; The initial phase is determined by the relative position of the original daily data. In this embodiment, the coefficient with a 95% confidence interval is: a1 = 1.238 (1.18, 1.296) b1 = 0.03441 (0.03396, 0.03486) c1 = 1.389 (1.291, 1.486) a2 = 0.2633(-0.2005, 0.727) b2 = 0.1164(0.1062, 0.1266) c2 = -0.7318 (-3.152, 1.689) a3 = 0.247(-0.2188, 0.7128) b3 = 0.1072 (0.09637, 0.118) c3 = -1.671(-4.219, 0.8767) a4 = 0.1708(0.1156, 0.226) b4 = 0.4208 (0.4178, 0.4238) c4 = -0.7655 (-1.41, -0.1212) a5 = 0.1546 (0.09843, 0.2107) b5 = 0.2069 (0.2033, 0.2105) c5 = 2.195 (1.428, 2.961) a6 = 0.1645(0.1062, 0.2227) b6 = 0.05609 (0.05282, 0.05936) c6 = -4.156(-4.872, -3.441) a7 = 0.1273 (0.07222, 0.1823) b7 = 0.7836 (0.7795, 0.7876) c7 = -1.276 (-2.14, -0.4118) a8 = 0.1434 (0.0844, 0.2025) b8 = 0.1487 (0.1445, 0.153) c8 = 1.314 (0.3901, 2.237) The goodness of fit for the above coefficients is as follows: Sum of Squared Errors (SSE): 49.98; R-squared: 0.8646; Adjusted R-squared: 0.8556; Root Mean Square Error (RMSE): 0.3801. The function value at the missing value is calculated. To simulate an actual data acquisition system, appropriate noise is added to this function value, and the resulting value is the imputed value. However, the magnitude of the added noise directly affects the quality of the missing value imputation and requires repeated adjustments to achieve the best results. The maximum deviation between the missing values ​​obtained according to the above steps and their corresponding original daily data does not exceed 10%, indicating that the proposed method is feasible.

[0025] This embodiment visualizes the basic errors of a company's smart meters from October 1, 2020 to March 31, 2022, selected from an experimental base, to observe data gaps. The original time data is stored in year / month / day format, quantized with the day as the minimum time interval. The processed time data is plotted on the x-axis, and the basic operational error of the smart meters is plotted on the y-axis. The original sample data is shown below. Figure 2 As shown. From Figure 2 As can be seen, from day 1 to day 150, the smart energy meter data acquisition system can collect sample data relatively well; around day 300 and day 400, the smart energy meter data acquisition system fails to collect sample data for a month, thus making it impossible to obtain the monthly average basic error data. Therefore, this embodiment proposes a trigonometric function interpolation (FFI) method to fit the daily data, complete the missing value filling, and obtain the monthly average basic error data. The function value at the missing value is calculated, and to simulate an actual data acquisition system, appropriate noise is added to this function value; the resulting value is the filled value. However, the magnitude of the added noise directly affects the quality of the missing value filling, requiring repeated adjustments to achieve the best effect. The maximum deviation between the missing value obtained according to the above steps and its corresponding original daily data does not exceed 10%, indicating that the proposed method is feasible, and the obtained interpolation filling effect is as follows: Figure 3 As shown.

[0026] Support Vector Machines (SVMs) have unique advantages in handling small sample sizes and nonlinear problems. This embodiment proposes using a Least Squares Support Vector Machine (LSSVM) model to analyze the degradation trend of smart energy meters. The loss function is a least squares linear system, and the inequality constraints of SVM are changed to equality constraints, simplifying the computational complexity. In step S104 of this embodiment, when constructing the LSSVM, the decision function used by the LSSVM is expressed as follows: , , In the above formula, For input data The output mapped to a higher-dimensional space, For Lagrange multipliers, For kernel function, For the first in the dataset i One input data, b The hyperplane intercept is... For kernel function The width factor; and the optimization objective and constraints adopted by the least squares support vector machine (LSSVM) are: , In the above formula, ω Let be the normal vector of the hyperplane. is the regularization parameter, and c> 0; For the first in the dataset i The relaxation factor for each data point For the first in the dataset i One output data, For the first in the dataset i One input data, The number of data points in the dataset. It is a nonlinear function.

[0027] In this embodiment, the padded monthly average data is used as the input to the least squares support vector machine (LSVM). The loss function is a least squares linear system, and the inequality constraints of the SVM are changed to equality constraints, simplifying the computational complexity. LSSVM utilizes nonlinear functions... φ ( x Input sample set D Mapping to a higher-dimensional space to build a regression model: , In the above formula, For input data The output mapped to a higher-dimensional space, ω Let be the normal vector of the hyperplane. It is a nonlinear function. b Let be the hyperplane intercept. The optimization objective and constraints, based on the principle of minimizing structural risk, are as follows: , In the above formula, is the regularization parameter, and c> 0; For the first in the dataset i The relaxation factor for each data point For the first in the dataset i One output data, For the first in the dataset i One input data, The number of data points in the dataset; in this embodiment, Lagrange multipliers are introduced. The Karush-Kuhn-Tucker constraints and the least squares method are used to... a i and b By solving this problem, the decision function in the high-dimensional space can be reconstructed as follows:

[0028] In the above formula, For input data The output mapped to a higher-dimensional space, For Lagrange multipliers, The kernel function (representing a nonlinear mapping from the input space to a high-dimensional feature space) For the first in the dataset i One input data, b is the hyperplane intercept.

[0029] While ensuring parameter consistency, a cross-validation method was used with the root mean square error (RMSE) as the evaluation criterion. Combined with the sample dataset, the RMSE value of the Gaussian kernel was found to be 0.0197, slightly lower than that of the polynomial kernel (0.0209) and the linear kernel (0.0225). Therefore, the model proposed in this invention selects the Gaussian kernel as the kernel function, and its expression is: .

[0030] Parameters in LSSVM c , σRelying on empirical reference values ​​often makes it difficult to achieve optimal performance for predicting the basic error of smart energy meters under multi-source input. Therefore, in this embodiment, when using the improved Sparrow Search Algorithm (ISSA) to optimize the parameters of the Least Squares Support Vector Machine (LSSVM) in step S105, the optimized parameters include the regularization parameters of the LSSVM. c and the width factor of the kernel function σ .

[0031] In step S105 of this embodiment, a method based on the Improved Sparrow Search Algorithm (ISSA) is proposed to implement the parameters. c , σ The optimization settings are as follows: Chaotic mapping and Gaussian mutation are introduced in the population initialization stage and the later stage of iteration, respectively, to improve the performance of SSA and form the ISSA algorithm to achieve adaptive optimization of LSSVM parameters. Figure 2 As shown, step S105 includes: Step S201, regularization parameters c and the width factor of the kernel function σ Consider the population as the improved sparrow search algorithm ISSA, and initialize the number of iterations t, the maximum number of iterations Max_t, and the number of sparrows pop. Step S202: Initialize the sparrow population using the Piecewise chaotic mapping based on the following formula: , In the above formula, P ∈(0,1) is a control parameter; ∈(0,1) and ∈(0,1) is a chaotic variable. This indicates the updated position of the sparrow. This indicates the position of the sparrows before the update; Step S203: Calculate the fitness value of individual sparrows and sort them according to the fitness value; Step S204: Update the positions of discoverers, joiners, and watchers in the sparrow population; Step S205: Update the position of the optimal sparrow individual using Gaussian and chaotic perturbations: , In the above formula, The updated sparrow position. The position of the sparrow before the update. It follows a Gaussian distribution; Step S206: Determine whether the iteration number t equals the maximum iteration number Max_t. If it does, output the regularization parameters of the Least Squares Support Vector Machine (LSSVM). cand the width factor of the kernel function σ If the condition is met, proceed to step S106; otherwise, proceed to step S203.

[0032] In this embodiment, step S107, determining whether the Least Squares Support Vector Machine (LSSVM) meets the requirements, refers to whether the number of iterations is equal to the preset maximum number of iterations. Furthermore, when jumping to step S105 to continue iterations, the iteration count is incremented by 1. Alternatively, a prediction accuracy less than or equal to a set value can be used as a condition for meeting the requirements.

[0033] To verify the basic error prediction method for smart meters under data shortage conditions in this embodiment, 10 smart meters from a certain batch of Company A were randomly selected from a typical high-temperature and dry environment test base in Xinjiang between October 1, 2020 and March 31, 2022. This included 18 months of basic error data and environmental stress data, with a sampling interval of 1 day. Each sample contained a total of 524 data points. To verify the model's generalization ability, the first 60% of 8 sets of data were randomly selected as the training sample set, and the remaining 40% of the data from the other two sets were used as the validation and test sample sets to predict the future degradation trend of smart meters. Before model training, to prevent large differences in the order of magnitude of the columns in the smart meter error dataset from affecting the model's accuracy and generalization, the sample data were normalized. To verify the accuracy of the ISSA-LSSVM combined prediction model based on FFI data imputation, popular prediction methods were selected for comparison, including the composite kernel SVM model, the multi-kernel GPR model, and the PSO-ELMAN neural network. The performance of the ISSA-LSSVM model without data imputation was also compared. In the composite kernel SVM model, the number of FCM clusters was set to 3, determined using grid search and cross-validation. c , σ The kernel function parameters were set as follows: the inertia factor of PSO was 0.8, the learning factors c1 and c2 were 1.5, the number of hidden layer nodes of the Elman network was set to 10, the learning rate was set to 0.01, and the final result was obtained by averaging the results five times. The maximum number of iterations for each model was Max_t=30, and the population size of the optimization algorithm was POP=20, which satisfied both the search range and the execution efficiency of the algorithm. To more intuitively compare the prediction performance of different models on the basic error of smart meters, RMSE, MAE, and R2 were selected as model performance evaluation indicators to quantitatively compare the performance of each model. The results are as follows. Figure 5 As shown. By Figure 5As can be seen, the model proposed in this embodiment significantly outperforms the composite kernel SVM, PSO-Elman, and multi-kernel GPR in prediction performance. The prediction results show an RMSE of 1.82%, a MAE of 1.53%, and an R² of 0.988. Compared to the composite kernel SVM model, the RMSE is reduced by 0.35%, the MAE by 0.26%, and the R² is increased by 0.4%, verifying the accuracy of the model proposed in this embodiment. The prediction results demonstrate that the ISSA-LSSVM combined model based on FFI filling proposed in this embodiment exhibits better data tracking ability, higher adaptability, and more accurate prediction performance. The test times for the five prediction models in this embodiment are shown in Table 1.

[0034] Table 1 Comparison of test times for five prediction models

[0035] Table 1 lists multi-core GPR, PSO-Elman, composite-core SVM, and ISSA-LSSVM as well-known existing models. As shown in Table 1, the average test time of the proposed model is 39.4 ms, significantly faster than PSO-ELMAN and multi-core GPR, and the fastest among the five prediction models. However, compared to ISSA-LSSVM, its running speed is slower due to the data padding process. Overall, the results demonstrate that the proposed FFI-filled ISSA-LSSVM combined weighted model (the proposed model) effectively improves the accuracy of basic error prediction for smart meters, exhibits higher operating efficiency, and demonstrates significant advantages in basic error prediction for smart meters.

[0036] In summary, addressing the issue of decreased accuracy in prediction models when basic error data for smart meters is missing under extreme natural environments, this embodiment proposes an ISSA-LSSVM combined prediction model based on FFI imputation, using operational data from the Xinjiang high-temperature and dry-heat test base in my country. First, the missing daily data is imputed using the FFI method to obtain better monthly average data, which is then used as the input feature of the Least Squares Support Vector Machine (LSSVM). Subsequently, an improved Sparrow Search Algorithm (ISSA) is established by combining chaotic mapping and Gaussian mutation methods to achieve adaptive optimization of the LSSVM prediction model parameters. Finally, a case study analysis of real-time basic error data from smart meters at the Xinjiang high-temperature and dry-heat test base demonstrates that the prediction model proposed in this embodiment has higher prediction accuracy and operational efficiency than currently popular prediction methods. The research results of this embodiment can provide guidance for the performance degradation analysis of smart meters under extreme natural environments and offer theoretical support for the State Grid Corporation's operation and maintenance of smart meters. Furthermore, this embodiment is applicable not only to smart meters but also to the degradation trend assessment of other metering equipment under different typical environments, exhibiting strong versatility. To improve the accuracy of basic error prediction for smart meters under hot and dry conditions, this embodiment proposes an ISSA-LSSVM combined prediction model filled with FFI. In the parameter selection stage of LSSVM, this embodiment proposes an ISSA optimization algorithm. Comparison with existing SSA algorithms shows that ISSA not only solves the problem of traditional optimization algorithms easily getting trapped in local optima, but also improves the parameter optimization speed, effectively optimizing the LSSVM model. It comprehensively improves model performance evaluation indicators such as RMSE, MAE, and R², demonstrating the model's superiority. The prediction model proposed in this embodiment can realize smart meter monitoring and early warning, providing accurate metering basis for reducing equipment failure rates and maintenance costs, ensuring fair and accurate metering, and helping to improve the economy and reliability of the power system.

[0037] Furthermore, this embodiment also provides a basic error prediction system for smart meters under data loss conditions, including a microprocessor and a memory interconnected, wherein the microprocessor is programmed or configured to execute the basic error prediction method for smart meters under data loss conditions.

[0038] Furthermore, this embodiment also provides a computer-readable storage medium storing a computer program / instruction that is programmed or configured to execute the basic error prediction method for smart meters under data loss conditions via a processor.

[0039] In addition, this embodiment also provides a computer program product, including a computer program / instruction, which is programmed or configured to execute the basic error prediction method for smart meters under data loss conditions via a processor.

[0040] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-readable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create a machine for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The functions specified in one or more boxes. These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable apparatus for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0041] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A method for predicting the basic error of a smart meter under conditions of missing data, characterized in that, include: Step S101: Collect the original daily data of the smart energy meter, including time and feature data and basic errors; Step S102: Clean and normalize the original daily data of the smart energy meter; Step S103: Use the trigonometric interpolation method FFI to fit the daily data to complete the missing value filling, and calculate the monthly average basic error based on the daily data after missing value filling. Step S104: Divide the dataset using time and environmental stress data as input and monthly average basic error as output, and construct a least squares support vector machine (LSSVM). Step S105: The improved sparrow search algorithm ISSA is used as the optimization parameters for the least squares support vector machine LSSVM. When using the improved sparrow search algorithm ISSA as the optimization parameters for the least squares support vector machine LSSVM, the sparrow population is initialized using Piecewise chaotic mapping, and the position of the best sparrow individual is updated using Gaussian and chaotic perturbations. Step S106: Train the least squares support vector machine (LSSVM) using the dataset; Step S107: Determine whether the Least Squares Support Vector Machine (LSSVM) meets the requirements. If it does, end and exit; otherwise, jump to step S105 to continue iterating. In step S103, when using the trigonometric interpolation method FFI to fit the daily data and complete the missing value imputation, the process includes first fitting the daily data using the sine superposition function shown in the following formula: f ( x )= a 1*sin( b 1* x + c 1)+ a 2*sin( b 2* x + c 2)+ a 3*sin( b 3* x + c 3)+ a 4*sin( b 4* x + c 4)+ a 5*sin( b 5* x + c 5)+ a 6*sin( b 6* x + c 6)+ a 7*sin( b 7*x+ c 7)+ a 8*sin( b 8*x+ c 8), In the above formula, f ( x () is a superposition of sinusoidal functions. a 1~ a 8. b 1~ b 8 and c 1~ c 8 is the fitting coefficient. x This refers to the input data.

2. The method for predicting the basic error of a smart energy meter under data loss conditions according to claim 1, characterized in that, The feature data in step S101 includes some or all of the ambient temperature, humidity, air pressure, wind speed, ultraviolet radiation, and light intensity.

3. The method for predicting the basic error of a smart energy meter under data loss conditions according to claim 1, characterized in that, In step S104, when constructing the Least Squares Support Vector Machine (LSSVM), the decision function used by the LSSVM is expressed as follows: ， ， In the above formula, For input data The output mapped to a higher-dimensional space, For Lagrange multipliers, For kernel function, For the first in the dataset i One input data, b The hyperplane intercept is... For kernel function The width factor; and the optimization objective and constraints adopted by the least squares support vector machine (LSSVM) are: ， In the above formula, ω Let be the normal vector of the hyperplane. Here is the regularization parameter, and c> 0; For the first in the dataset i The relaxation factor for each data point For the first in the dataset i One output data, For the first in the dataset i One input data, The number of data points in the dataset. It is a nonlinear function.

4. The method for predicting the basic error of a smart energy meter under data loss conditions according to claim 3, characterized in that, In step S105, when the improved Sparrow Search Algorithm (ISSA) is used to optimize the parameters of the Least Squares Support Vector Machine (LSSVM), the optimized parameters include the regularization parameters of the LSSVM. c and kernel function Width factor σ .

5. The method for predicting the basic error of a smart energy meter under data loss conditions according to claim 4, characterized in that, Step S105 includes: Step S201, regularization parameters c and the width factor of the kernel function σ Consider the population as the improved sparrow search algorithm ISSA, and initialize the number of iterations t, the maximum number of iterations Max_t, and the number of sparrows pop. Step S202: Initialize the sparrow population using the Piecewise chaotic mapping based on the following formula: ， In the above formula, P ∈(0,1) is a control parameter; ∈(0,1) and ∈(0,1) is a chaotic variable. This indicates the updated position of the sparrow. This indicates the position of the sparrow before the update; Step S203: Calculate the fitness value of individual sparrows and sort them according to the fitness value; Step S204: Update the positions of discoverers, joiners, and watchers in the sparrow population; Step S205: Update the position of the optimal sparrow individual using Gaussian and chaotic perturbations: ， In the above formula, The updated sparrow position. The position of the sparrow before the update. It follows a Gaussian distribution; Step S206: Determine whether the iteration number t equals the maximum iteration number Max_t. If it does, output the regularization parameters of the Least Squares Support Vector Machine (LSSVM). c and the width factor of the kernel function σ If the condition is met, proceed to step S106; otherwise, proceed to step S203.

6. The method for predicting the basic error of a smart energy meter under data loss conditions according to claim 1, characterized in that, In step S107, determining whether the least squares support vector machine (LSSVM) meets the requirements means whether the number of iterations is equal to the preset maximum number of iterations. When jumping to step S105 to continue iterating, it also includes incrementing the number of iterations by 1.

7. A basic error prediction system for smart energy meters under data loss conditions, comprising a microprocessor and a memory interconnected, characterized in that, The microprocessor is programmed or configured to execute the basic error prediction method for smart meters under data loss conditions as described in any one of claims 1 to 6.

8. A computer-readable storage medium storing a computer program / instructions, characterized in that, The computer program / instructions are programmed or configured to execute, via a processor, the basic error prediction method for smart meters under data loss conditions as described in any one of claims 1 to 6.

9. A computer program product comprising a computer program / instructions, characterized in that, The computer program / instructions are programmed or configured to execute, via a processor, the basic error prediction method for smart meters under data loss conditions as described in any one of claims 1 to 6.

Images