Metering error correction method and system based on electric power data acquisition

Abstract

The invention discloses a metering error correction method and system based on electric power data acquisition, and the method comprises the steps: detecting an operation state through a self-inspection and data acquisition module, and obtaining environment data, electrical parameters and error information; key feature data are selected and standardized through a data preprocessing module; an LSTM model and an SVR model are constructed through a model establishment and fusion module, and prediction results are fused to generate error compensation output; real-time error compensation is carried out on a combined prediction result by utilizing an ARIMA model and error time sequence processing through a dynamic error compensation module; the data smoothing and evaluation module is used for smoothing the voltage data and evaluating the error compensation effect; an error compensation effect is monitored through a monitoring and optimizing module, and feature selection and model parameters are optimized regularly; through the fault diagnosis and emergency module, an abnormal condition is detected, and a corresponding standby scheme is triggered. According to the invention, the accuracy and anti-interference capability of electric power metering are effectively improved, and the method is suitable for complex and changeable metering environments.

Description

Technical Field

[0001] The present invention relates to the field of electrical variable measurement and analysis, and particularly to a metering error correction method and system based on power data acquisition. Background Art

[0002] With the continuous development of modern power systems, the accuracy and stability of power metering play a crucial role in the operation and management of power systems. However, since power metering devices are exposed to complex environments for a long time, their measurement results are easily interfered by various factors such as temperature, humidity, and voltage fluctuations, resulting in error accumulation and affecting the accuracy of power data.

[0003] Currently, common error correction methods are mostly static compensations based on single features, which cannot respond to the changes of real-time data in a timely manner and are also difficult to adapt to the dynamic changes of the environment and working conditions. This kind of correction method has limited effect in dealing with error accumulation under complex working conditions, easily leads to data deviation, and affects the metering accuracy of power systems. Especially under the requirement of high-frequency data acquisition, the calculation efficiency of traditional correction methods also has bottlenecks.

[0004] The present invention provides a metering error correction method and system based on power data acquisition, aiming to improve the accuracy and adaptability of power metering. The system obtains the environmental parameters and electrical parameters of the power metering transformer in real time through a data acquisition module, and uses transfer entropy for feature selection to screen out the key features that have a greater impact on errors, ensuring the effectiveness of the input data. Summary of the Invention

[0005] In view of the above problems, the present invention provides a metering error correction method and system based on power data acquisition to solve the problems that the static compensation of single features in the prior art cannot respond to the changes of real-time data in a timely manner, is difficult to adapt to the dynamic changes of the environment and working conditions, has limited effect in dealing with error accumulation under complex working conditions, easily leads to data deviation, and affects the metering accuracy of power systems.

[0006] To solve the above technical problems, the present invention provides the following technical solutions: The present invention provides a metering error correction method based on power data acquisition, which is characterized by including the following steps: Step S1, detecting the operating state through a self-check and data acquisition module, and obtaining environmental data, electrical parameters, ratio error, and angular error information; Wherein in step S1, the following sub-steps are further included: S1-1, automatically running the self-check function at startup to check the connection status of each device and sensor, the power supply situation, and the communication situation; S1-2. Collect environmental data in real time using built-in temperature and humidity sensors, extract electrical parameters using an electrical energy metering transformer, and collect ratio error and angular error information of the metering transformer.

[0007] Step S2. Select key feature data through a data preprocessing module, and perform standardization processing on the data to eliminate dimensional differences. Among them, in step S2, the following sub-steps are also included: S2-1. Feature data selection. Calculate each variable based on the transfer entropy theory, including temperature, humidity, load voltage, secondary voltage, and the error correlation of the electrical energy metering transformer, and screen the key feature data that affects the error. When the transfer entropy value approaches 0, it indicates a weak relationship between variables, and select the variable with a larger entropy value as the feature input. Specifically, as shown in Equation (1): Equation (1) Among them, is the transfer entropy value, is the length of the data time series, and are the dimensions of the feature data respectively. The larger the transfer entropy value, the stronger the influence of the feature data on the error. is the value of the target variable at the future time n+1, is the value sequence of the target variable at the past k time points, is the value sequence of the feature variable y at the past l time points, is the joint probability distribution, is the conditional probability distribution; S2-2. Data standardization. Use the Min-Max standardization formula to perform standardization processing on the feature data to eliminate dimensional differences. Specifically, as shown in Equation (2): Equation (2) Among them, is the original data, and are the minimum and maximum values of the data respectively.

[0008] Step S3. Through the model establishment and fusion module, construct LSTM and SVR models, optimize the prediction results through weight optimization, and generate error compensation output. Among them, in step S3, the following sub-steps are also included: S3-1. Construction of LSTM and SVR models. Use improved LSTM and SVR as sub-models for error compensation, input the selected feature data into the LSTM and SVR models for training, and generate prediction output. S3-2. Calculate the weight coefficients of each model, and optimize the weights using the standard deviation and the sum of squares of the lowest prediction errors. Specifically, as shown in Equation (3) - Equation (4): Equation (3) Equation (4) where are the weights of the SVR model are the weights of the LSTM model and are the standard deviation differences of the SVR model and the LSTM model respectively; The prediction results of each sub - model are fused by the average weighted method to form a combined prediction output, as shown in Equation (5) specifically: Equation (5) where is the final prediction result and are the prediction results of the SVR and LSTM models respectively.

[0009] Step S4, through the dynamic error compensation module, use the ARIMA model and error time series processing to perform real - time error compensation on the combined prediction result; Among them, in step S4, the following sub - steps are also included: S4 - 1, using the ARIMA model, taking the output result of the combined prediction model as the input of the ARIMA model, used to model the historical observation data for real - time error compensation, revealing the dynamic relationship between error parameters, and generating a compensation value; S4 - 2, error time series processing, using the AFC method to detect the stationarity of the error time series, selecting the ARIMA model, predicting and compensating the next - step error through linear combination, taking the output of the combined prediction model as the input of the ARIMA model, modeling the historical error sequence to achieve dynamic compensation, as shown in Equation (6) - Equation (7) specifically: Equation (6) Equation (7) where is the current error value is the actual observed value of the power metering transformer is the final output data after compensation is the predicted error compensation value, obtaining the final compensated data output.

[0010] Step S5, through the data smoothing and evaluation module, use low - pass filtering to smooth the voltage data, and use the RMSE and MAE indicators to evaluate the error compensation effect; Among them, in step S5, the following sub - steps are also included: S5-1, output the compensated data of the electrical energy metering transformer, and use the low-pass filtering method to smooth the voltage data and remove high-frequency noise, as shown in Equation (8): Equation (8) where is the current output after filtering, is the current input data, is the output after the previous filtering, is the smoothing factor, and the range is The smaller it is, the stronger the smoothing effect; S5-2, evaluate the error compensation effect, and use the root mean square error and mean absolute error to quantitatively process the compensation effect, as shown in Equation (9) - Equation (10): Equation (9) Equation (10) where is the root mean square error, is the mean absolute error, is the model prediction value, is the actual value, and n is the total number of data points.

[0011] Step S6, through the monitoring and optimization module, monitor the error compensation effect and regularly optimize the feature selection and model parameters; Among them, in Step S6, the following sub-steps are also included: S6-1, monitor the error compensation effect, review the error compensation effect at fixed intervals, regularly draw the time trend chart of the compensation value and the actual value, and observe the consistency of their trends; S6-2, regularly optimize the feature selection and model parameters. By setting the performance threshold, if the real-time error or trend deviation exceeds this threshold, trigger the model update mechanism to re-optimize the feature selection and model parameters.

[0012] Step S7, through the fault diagnosis and emergency module, use the Z-score to detect abnormal conditions and trigger appropriate backup plans according to the abnormal conditions.

[0013] Among them, in Step S7, the following sub-steps are also included: S7-1, use the Z-score detection to judge the outliers of the collected data and the predicted errors, as shown in Equation (11): Equation (11) where is the value of the current data point, is the average value of the data, is the standard deviation of the data. When | When it exceeds the set threshold, this data is regarded as abnormal; S7-2, through the alternative solution module, judges the type of abnormality after detecting the abnormality, triggers different alternative solutions according to different abnormal conditions, and adjusts the system parameters in real time with feedback.

[0014] The present invention provides a metering error correction system based on power data acquisition, which is characterized by including: A self-check and data acquisition module, a data preprocessing module, a model establishment and fusion module, a dynamic error compensation module, a data smoothing and evaluation module, a monitoring and optimization module, and a fault diagnosis and emergency module; The self-check and data acquisition module checks the connection status of the device and the sensor, the power supply situation, and the communication status at startup, collects environmental parameters (and electrical parameters) in real time, and records metering error data; The data preprocessing module, based on the transfer entropy theory, screens out the key feature data closely related to the error, and uses the Min-Max normalization method to normalize the data to eliminate the dimensional difference of different data; The model establishment and fusion module uses the LSTM and SVR models to construct an error prediction model to deal with different time series data patterns, and weights and optimizes the weights of the prediction results of LSTM and SVR; The dynamic error compensation module inputs the fused prediction result into the ARIMA model for dynamic error correction, and uses time series analysis to smooth the error data; The data smoothing and evaluation module uses a low-pass filtering method to smooth the voltage data to remove high-frequency noise, and uses indicators such as RMSE and MAE to evaluate the compensation effect and quantify the error correction accuracy; The monitoring and optimization module regularly checks the compensation effect, draws a trend chart of the compensation value and the actual value, observes the error change, and automatically triggers the update of the model parameters and feature selection when the error deviation exceeds the set threshold; The fault diagnosis and emergency module detects abnormal data through the Z-score, timely identifies potential faults in the system, automatically triggers alternative solutions according to the fault type, and adjusts the system parameters in real time.

[0015] Compared with the prior art, the beneficial effects of the present invention are: Detect the operating status through the self-check and data acquisition module, and obtain environmental data, electrical parameters, ratio difference, and angular difference error information; through the data preprocessing module, select key feature data and perform data standardization processing to eliminate dimensional differences; through the model establishment and fusion module, construct LSTM and SVR models, optimize the prediction results through weight optimization, and generate error compensation output; through the dynamic error compensation module, use the ARIMA model and error time series processing to perform real-time error compensation on the combined prediction results; through the data smoothing and evaluation module, use low-pass filtering to smooth the voltage data, and use RMSE and MAE indicators to evaluate the error compensation effect; through the monitoring and optimization module, monitor the error compensation effect and regularly optimize feature selection and model parameters; through the fault diagnosis and emergency module, use the Z-score to detect abnormal conditions and trigger appropriate backup plans according to the abnormal conditions.

[0016] The present invention screens key feature data through the transfer entropy method, ensuring that the input data of the error compensation model is more accurate, avoiding the interference of irrelevant data on prediction, and thus improving the accuracy of metering error correction.

[0017] The present invention combines the LSTM and SVR models, and realizes the complementary advantages of multiple models through weight optimization, making the error prediction more accurate, effectively adapting to complex non-linear features, and improving the reliability of the compensation effect.

[0018] The present invention introduces the ARIMA model for dynamic error compensation, which can respond to data changes in real time, perform rapid correction for different environments and working conditions, reduce the problem that traditional static compensation methods cannot adapt to environmental changes, and thus maintain high-precision metering in a dynamic environment.

[0019] Through the data standardization and low-pass filtering processing methods of the present invention, the compensated data is kept stable. At the same time, the RMSE and MAE evaluation mechanisms provide accurate feedback for error compensation, ensuring the stability of the system during long-term operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required to be used in the embodiments. It should be understood that the following drawings only show some embodiments of the present invention, and therefore should not be regarded as limiting the scope. For those of ordinary skill in the art, without creative efforts, other related drawings can also be obtained based on these drawings.

[0021] Figure 1 is the method flow chart of the present invention; [[ID=2**]] Figure 2 is the system architecture diagram of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0022] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed present invention, but is merely for the selected embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0023] Please refer to Figure 1 which is a flowchart of a metering error correction method based on power data acquisition provided by an embodiment of the present invention, including the following steps: Step S1, detect the operating state through the self-check and data acquisition module, and obtain environmental data, electrical parameters, ratio error, and angular error information; Among them, in step S1, the following sub-steps are further included: S1-1, automatically run the self-check function at startup to check the connection status of each device and sensor, the power supply situation, and the communication situation; S1-2, use the built-in temperature and humidity sensors to collect environmental data in real time, use the power metering transformer to extract electrical parameters, and use the metering transformer to collect ratio error and angular error information.

[0024] It should be noted that for the device connection status check, the connection status of all key hardware devices is detected in sequence at startup to ensure that all devices are in an online state; Detect the stability and voltage value of the power supply to ensure that the power supply is within the normal range; if the voltage deviates from the preset range, record the abnormality or issue an alarm to prevent power fluctuations from affecting subsequent data collection; Check the communication link and signal quality between devices to ensure that data transmission is smooth and error-free; if there is a communication failure, the system will activate the backup communication link or automatically switch to the local cache mode to avoid data loss.

[0025] Step S2, through the data preprocessing module, select key feature data and perform standardized processing on the data to eliminate dimensional differences; Among them, in step S2, the following sub-steps are further included: S2-1, Feature data selection: Calculate each variable based on transfer entropy theory, including temperature, humidity, load voltage, secondary voltage, and the correlation of the error of the electrical energy metering transformer. Screen the key feature data that affect the error. When the transfer entropy value approaches 0, it indicates a weak relationship between variables. Select the variable with a larger entropy value as the feature input, as shown in Equation (1): Equation (1) Where is the transfer entropy value, is the length of the data time series, and are the dimensions of the feature data respectively. The larger the transfer entropy value, the stronger the influence of the feature data on the error. is the value of the target variable at the future time n+1, is the value sequence of the target variable at the past k time points, is the value sequence of the feature variable y at the past l time points, is the joint probability distribution, is the conditional probability distribution; S2-2, Data standardization: Use the Min-Max standardization formula to standardize the feature data to eliminate the dimension difference, as shown in Equation (2): Equation (2) Where is the original data, and are the minimum and maximum values of the data respectively.

[0026] It should be noted that the feature data with high transfer entropy is considered to be highly correlated with the target variable and is preferentially selected as the input of the model to enhance the prediction ability of the model; there may be a large number of features in the dataset, but not every feature is helpful for error compensation. Transfer entropy helps the system filter out the feature variables with weak relationship with the error, thus simplifying the data input, reducing the calculation amount, and improving the system efficiency; by screening out the most useful feature data, the interference of irrelevant data can be avoided, making the model more guaranteed in the quality of the input data.

[0027] In addition, the screened feature data can provide more valuable input information for the model, thus improving the accuracy of error prediction. Transfer entropy ensures that the model focuses on the data most relevant to the electrical energy metering error and avoids the influence of irrelevant data; in the power metering environment, factors such as temperature, humidity, and load voltage may affect the error.

[0028] Transfer entropy enables the model to flexibly select key features in different environments by dynamically analyzing the influence of each feature, ensuring the adaptability and stability of the model in complex environments; transfer entropy helps the system automatically filter out redundant features, simplify the data structure, reduce the computational burden of the model, and enable the system to operate efficiently while ensuring the compensation accuracy, meeting the requirements of real-time dynamic compensation.

[0029] Step S3: Through the model establishment and fusion module, construct the LSTM and SVR models, optimize the prediction results through weight optimization, and generate the error compensation output. Among them, in step S3, the following sub-steps are also included: S3-1: Construction of LSTM and SVR models. Use the improved LSTM and SVR as sub-models for error compensation. Input the selected feature data into the LSTM and SVR models for training, and generate the prediction output. S3-2: Calculate the weight coefficients of each model. Optimize the weights using the standard deviation and the sum of squares of the lowest prediction errors, as shown in equations (3)-(4): Equation (3) Equation (4) Among them, is the weight of the SVR model, is the weight of the LSTM model, and are the standard deviation differences of the SVR model and the LSTM model respectively; Fuse the prediction results of each sub-model through the average weighting method to form the combined prediction output, as shown in equation (5): Equation (5) Among them, is the final prediction result, and are the prediction results of the SVR and LSTM models respectively.

[0030] It should be noted that the key features selected by using transfer entropy are used as the input of the LSTM. By constructing a multi-layer LSTM network, the ability of the model to capture non-linear and complex patterns is enhanced to adapt to the complex dynamic characteristics of the power metering error; combined with the key feature data selected by transfer entropy, multi-dimensional input data is introduced into the SVR to enhance the prediction ability of the model in the high-dimensional space, enabling it to better adapt to the influence of multiple factors on the error.

[0031] Step S4: Through the dynamic error compensation module, use the ARIMA model and error time series processing to perform real-time error compensation on the combined prediction results. Among them, in step S4, the following sub-steps are also included: S4-1. Using the ARIMA model, take the output result of the combined prediction model as the input of the ARIMA model, which is used to model the historical observation data by compensating the error in real time, reveal the dynamic relationship between error parameters, and generate compensation values. S4-2. Error time series processing. Use the AFC method to detect the stationarity of the error time series, select the ARIMA model, predict and compensate the next error through linear combination, and take the output of the combined prediction model as the input of the ARIMA model to model the historical error series and achieve dynamic compensation, as shown in Equations (6)-(7) specifically: Equation (6) Equation (7) Wherein, is the current error value, is the actual observed value of the electric energy metering transformer, is the final output data after compensation, is the predicted error compensation value, and the finally compensated data output is obtained.

[0032] It should be noted that through autoregressive and differencing operations, the ARIMA model can accurately capture the dynamic change trend of errors, especially having good modeling ability for short-term fluctuations and long-term trends in errors.

[0033] The differencing process of ARIMA can eliminate non-stationarity and make the time series data more stable, thus providing a more reliable basis for error compensation. In a complex power metering environment, errors may change dynamically due to environmental changes (such as temperature and humidity fluctuations) or equipment conditions (such as load fluctuations). The ARIMA model can adapt to these changes by updating parameters in real time and improve the accuracy of compensation results.

[0034] Fusing the output of ARIMA with LSTM and SVR can form a multi-level compensation mechanism, which is suitable for both long-term trends and dynamic fluctuations. During the dynamic compensation process, if the error is not corrected in real time, it will lead to the cumulative effect of errors. The moving average part of ARIMA can effectively correct the prediction error and avoid the influence of error accumulation on the accuracy of compensation results.

[0035] Step S5. Through the data smoothing and evaluation module, use low-pass filtering to smooth the voltage data, and use the RMSE and MAE indicators to evaluate the error compensation effect. Among them, in step S5, the following sub-steps are also included: S5-1. Output the data of the electric energy metering transformer after compensation, and use the low-pass filtering method to smooth the voltage data to remove high-frequency noise, as shown in Equation (8) specifically: Equation (8) where is the output after current filtering, is the current input data, is the output after previous filtering, is the smoothing factor, with a range of The smaller it is, the stronger the smoothing effect; S5-2, Error compensation effect evaluation, using root mean square error and mean absolute error to quantify the compensation effect, specifically as shown in Equation (9) - Equation (10): Equation (9) Equation (10) where is the root mean square error, is the mean absolute error, is the model prediction value, is the actual value, and n is the total number of data points.

[0036] It should be noted that the selection of the smoothing factor will also be different in different scenarios: In scenarios with stable environment and low noise, the data changes slowly, the noise is less, and the error trend is stable. A higher smoothing factor should be selected, which can quickly respond to data changes while retaining less smoothing effect.

[0037] In scenarios with frequent data fluctuations and large noise, the data contains more high-frequency noise or sudden changes, and the error signal is not stable. A lower smoothing factor should be selected. This helps to suppress noise and make the output data more stable, suitable for dealing with situations with more high-frequency noise.

[0038] In scenarios with high real-time response requirements, such as dynamic compensation scenarios, where it is necessary to quickly capture data mutations, a higher smoothing factor should be selected to make the filter more sensitive to new data, which can quickly reflect real-time changes, but the smoothing effect is relatively weak.

[0039] In scenarios of data trend analysis or long-term evaluation, more attention is paid to the long-term trend of data rather than short-term fluctuations. A medium to low smoothing factor should be selected to make the output more stable, suitable for analyzing the long-term trend of errors.

[0040] Step S6, through the monitoring and optimization module, monitor the error compensation effect and optimize the feature selection and model parameters regularly; Among them, in step S6, the following sub-steps are also included: S6-1, Monitor the error compensation effect, review the error compensation effect at fixed intervals, regularly plot the time trend graph of the compensation value and the actual value, and observe the consistency of their trends; S6-2. Regularly optimize feature selection and model parameters. By setting a performance threshold, if the real-time error or trend deviation exceeds this threshold, trigger the model update mechanism to re-optimize the feature selection and model parameters.

[0041] It should be noted that for the situation where the frequency is unstable or fluctuates greatly in actual applications, a strategy of dynamically adjusting the update frequency can be considered: Update triggered based on the error threshold. An error threshold can be set, and when the system error continuously exceeds the preset range, an update is triggered; for example, when the RMSE or MAE exceeds the set value, immediately reselect features or update model parameters.

[0042] Rolling window update. Use a rolling time window to evaluate the changes in features and model parameters. When there are significant changes in the data distribution within the window, automatically adjust the update frequency to ensure that the model has stronger adaptability in an unstable data environment.

[0043] Adaptive update based on environmental parameters. Incorporate external environmental parameters into the update logic. When the environmental change exceeds the set threshold, the system automatically increases the update frequency to ensure that the model keeps in sync with the environmental change.

[0044] Step S7. Through the fault diagnosis and emergency module, use the Z-score to detect abnormal conditions and trigger appropriate backup plans according to the abnormal conditions.

[0045] Among them, in step S7, the following sub-steps are also included: S7-1. Use the Z-score detection to judge outliers for the collected data and prediction errors, specifically as shown in Equation (11): Equation (11) Where, is the value of the current data point, is the average value of the data, is the standard deviation of the data. When | | exceeds the set threshold, this data is regarded as abnormal; S7-2. Through the backup plan module, judge the type of abnormality after detecting the abnormality, trigger different backup plans according to different abnormal conditions, and adjust the system parameters in real-time feedback.

[0046] It should be noted that when the Z-score exceeds the set threshold, this data is regarded as an outlier, and the outlier may indicate sensor abnormality, electrical parameter fluctuation or communication failure situation.

[0047] Multiple consecutive abnormal Z - scores, with persistent data deviation, usually indicate sensor damage or failure; Z - scores deviate within a short time but return to normal, usually indicating temporary electrical interference or load fluctuations; if the data intervals are long or there are obvious data gaps, it may be an anomaly caused by network or communication problems.

[0048] According to the judged abnormal situations, different backup plans can be corresponding: Sensor failure backup plan: Switch to a backup sensor, or enable a backup sensing device of the same type, and recalibrate the data acquisition parameters to ensure data continuity and accuracy.

[0049] Electrical fluctuation backup plan: Introduce a short - term moving average in the system as a supplement to the current data to reduce the impact of fluctuations, and at the same time adjust the smoothing factor of the model input to reduce the sensitivity of the system to short - term fluctuations.

[0050] Communication failure backup plan: Enable a data caching mechanism to temporarily store the current data locally and upload it after the communication is restored. In addition, add a resume - interrupted - transfer mechanism to ensure that the data can be uploaded completely after the system connection is restored.

[0051] Please refer to Figure 2 It is the architecture diagram of the metering error correction system based on power data acquisition provided by the embodiment of the present invention, including: Self - inspection and data acquisition module, data pre - processing module, model establishment and fusion module, dynamic error compensation module, data smoothing and evaluation module, monitoring and optimization module, fault diagnosis and emergency module; The self - inspection and data acquisition module checks the connection status of the device and sensor, power supply situation and communication status during startup, and real - time collects environmental parameters (and electrical parameters, and records metering error data; The data pre - processing module screens out key feature data closely related to the error based on the transfer entropy theory, and normalizes the data using the Min - Max normalization method to eliminate the dimensional differences of different data; The model establishment and fusion module constructs an error prediction model using LSTM and SVR models to handle different time - series data patterns, and performs weighted fusion on the prediction results of LSTM and SVR to optimize the weights; The dynamic error compensation module inputs the fusion prediction result into the ARIMA model for dynamic error correction, and uses time - series analysis to smooth the error data; The data smoothing and evaluation module smooths the voltage data using a low - pass filtering method to remove high - frequency noise, and evaluates the compensation effect using indicators such as RMSE and MAE to quantify the error correction accuracy; The monitoring and optimization module regularly checks the compensation effect, draws a trend chart of the compensation value and the actual value, observes the error change, and automatically triggers the update of the model parameters and feature selection when the error deviation exceeds the set threshold. The fault diagnosis and emergency module detects abnormal data through the Z-score, timely identifies potential faults in the system, automatically triggers a backup plan according to the fault type, and adjusts the system parameters in real time.

[0052] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, various changes and modifications can be made to the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A metering error correction method based on power data acquisition, characterized in that It includes the following steps: Step S1, detect the operating status through the self-check and data acquisition module, and obtain environmental data, electrical parameters, ratio difference, and angular difference error information; Step S2, through the data preprocessing module, select key feature data and perform standardization processing on the data to eliminate dimensional differences; Step S3, through the model establishment and fusion module, construct LSTM and SVR models, optimize the prediction results through weight optimization, and generate error compensation output; Step S4, through the dynamic error compensation module, use the ARIMA model and error time series processing to perform real-time error compensation on the combined prediction results; Step S5, through the data smoothing and evaluation module, use low-pass filtering to smooth the voltage data, and use RMSE and MAE indicators to evaluate the error compensation effect; Step S6, through the monitoring and optimization module, monitor the error compensation effect, and regularly optimize feature selection and model parameters; Step S7, through the fault diagnosis and emergency module, use the Z-score to detect abnormal conditions, and trigger appropriate backup plans according to the abnormal conditions.

2. The metering error correction method based on power data acquisition according to claim 1, characterized in that: Wherein in step S1, the following sub-steps are further included: S1-1, automatically run the self-check function at startup, check the connection status of each device and sensor, power supply situation, and communication situation; S1-2, use the built-in temperature and humidity sensors to collect environmental data in real time, the power metering transformer extracts electrical parameters, and the metering transformer collects ratio difference and angular difference error information.

3. The metering error correction method based on power data acquisition according to claim 1, characterized in that: Wherein in step S2, the following sub-steps are further included: S2-1, feature data selection, calculate each variable based on the transfer entropy theory, including temperature, humidity, load voltage, secondary voltage, and the error correlation of the power metering transformer, screen the key feature data affecting the error, when the transfer entropy value approaches 0, it indicates that the relationship between variables is weak, and select the variable with a larger entropy value as the feature input, as shown in formula (1) specifically: Formula (1) Among them, is the transfer entropy value, is the length of the data time series, and are the dimensions of the feature data respectively. The larger the transfer entropy value, the stronger the influence of the feature data on the error. is the value of the target variable at the future time n + 1, is the value sequence of the target variable at the past k time points, is the value sequence of the feature variable y at the past l time points, is the joint probability distribution, is the conditional probability distribution; S2-2, data standardization, use the Min-Max standardization formula to perform standardization processing on the feature data to eliminate dimensional differences, as shown in formula (2) specifically: Formula (2) Among them, is the original data, and are the minimum and maximum values of the data, respectively.

4. The metering error correction method based on power data acquisition according to claim 1, characterized in that: Wherein in step S3, the following sub-steps are further included: S3-1, construction of LSTM and SVR models, use the improved LSTM and SVR as sub-models for error compensation, input the selected feature data into the LSTM and SVR models for training, and generate prediction outputs; S3-2, calculate the weight coefficients of each model, and use the standard deviation and the sum of squares of the lowest prediction errors to optimize the weights, as shown in formulas (3)-(4) specifically: Equation (3) Formula (4) Among them, is the weight of the SVR model, is the weight of the LSTM model, and are the standard deviation differences of the SVR model and the LSTM model, respectively; Fuse the prediction results of each sub-model through the average weighting method to form a combined prediction output, as shown in formula (5) specifically: Formula (5) Among them, is the final prediction result, and are the prediction results of the SVR and LSTM models respectively.

5. The metering error correction method based on power data acquisition according to claim 1, characterized in that: Wherein in step S4, the following sub-steps are further included: S4-1. Using the ARIMA model, take the output result of the combined prediction model as the input of the ARIMA model, use it to model the historical observation data by compensating the error in real time, reveal the dynamic relationship between the error parameters, and generate the compensation value; S4-2. Error time series processing. Use the AFC method to detect the stationarity of the error time series, select the ARIMA model, predict and compensate the next error through linear combination prediction, take the output of the combined prediction model as the input of the ARIMA model, model the historical error series, and achieve dynamic compensation, as shown in equations (6)-(7): Formula (6) Equation (7) Among them, is the current error value, is the actual observed value of the electric energy metering transformer, is the final output data after compensation, is the predicted error compensation value to obtain the final compensated data output.

6. The metering error correction method based on power data acquisition according to claim 1, characterized in that: Wherein in step S5, the following sub-steps are further included: S5-1. Output the compensated power metering transformer data, and use the low-pass filtering method to smooth the voltage data to remove high-frequency noise, as shown in equation (8): Formula (8) Among them, is the output after current filtering, is the current input data, is the output after previous filtering, is the smoothing factor, and the range is The smaller it is, the stronger the smoothing effect; S5-2. Error compensation effect evaluation. Use the root mean square error and the mean absolute error to quantify the compensation effect, as shown in equations (9)-(10): Formula (9) Formula (10) Among them, is the root mean square error, is the mean absolute error, is the model prediction value, is the actual value, and n is the total number of data points.

7. The metering error correction method based on power data acquisition according to claim 1, characterized in that: Wherein in step S6, the following sub-steps are further included: S6-1. Monitor the error compensation effect, review the error compensation effect at fixed intervals, regularly draw the time trend chart of the compensation value and the actual value, and observe the consistency of their trends; S6-2. Regularly optimize the feature selection and model parameters. By setting the performance threshold, if the real-time error or trend deviation exceeds this threshold, trigger the model update mechanism to re-optimize the feature selection and model parameters.

8. The metering error correction method based on power data acquisition according to claim 1, characterized in that: Wherein in step S7, the following sub-steps are further included: S7-1. Use the Z-score detection to judge the outliers of the collected data and the predicted errors, as shown in equation (11): Formula (11) Among them, is the value of the current data point, is the average value of the data, is the standard deviation of the data. When | | exceeds the set threshold, this data is regarded as abnormal; S7-2. Through the backup plan module, judge the type of anomaly after detecting the anomaly, trigger different backup plans according to different anomaly conditions, and adjust the system parameters in real time and feedback.

9. A metering error correction system based on power data acquisition, characterized in that, Including: Self-check and data acquisition module, data preprocessing module, model establishment and fusion module, dynamic error compensation module, data smoothing and evaluation module, monitoring and optimization module, fault diagnosis and emergency module; The self-check and data acquisition module checks the connection status of the equipment and sensors, the power supply situation and the communication status at startup, collects the environmental parameters (and electrical parameters) in real time, and records the metering error data; The data preprocessing module, based on the transfer entropy theory, screens out the key feature data closely related to the error, and uses the Min-Max normalization method to normalize the data to eliminate the dimensional difference of different data; The model establishment and fusion module uses the LSTM and SVR models to construct an error prediction model to cope with different time series data patterns, and performs weighted fusion on the prediction results of LSTM and SVR to optimize the weights; The dynamic error compensation module inputs the fusion prediction results into the ARIMA model for dynamic error correction, and uses time series analysis to smooth the error data; The data smoothing and evaluation module uses a low-pass filtering method to smooth the voltage data and remove high-frequency noise, and uses indicators such as RMSE and MAE to evaluate the compensation effect and quantify the error correction accuracy; The monitoring and optimization module regularly checks the compensation effect, draws a trend chart of the compensation value and the actual value, observes the error change, and automatically triggers the update of the model parameters and feature selection when the error deviation exceeds the set threshold; The fault diagnosis and emergency module detects abnormal data through the Z-score, timely identifies potential faults in the system, automatically triggers a backup plan according to the fault type, and adjusts the system parameters in real time.

Images