High-precision electric energy metering error dynamic calibration method and system

Abstract

This invention relates to the field of power system energy metering, specifically to a high-precision dynamic calibration method and system for energy metering errors. The method comprises six steps: multi-dimensional synchronous acquisition, error component decoupling calculation, real-time error benchmark generation, incremental learning model updating, dynamic calibration compensation output, and closed-loop verification of calibration results. The system includes a multi-dimensional synchronous acquisition module, an error component decoupling module, a standard metering benchmark module, a real-time error benchmark generation module, an incremental learning calibration module, a dynamic compensation output module, and a closed-loop verification module. This invention can decouple the three-dimensional coupled non-stationary metering errors of load rate, harmonic order, and transient duration under nonlinear impulsive loads in real time. It uses an incremental learning algorithm to update the calibration model online, eliminating the need to store massive amounts of historical data or perform offline retraining, thus achieving continuous dynamic calibration without interrupting metering.

Description

Technical Field

[0001] This invention belongs to the field of power system energy metering, specifically relating to a method and system for dynamic calibration of high-precision energy metering errors. Background Technology

[0002] With the rapid development of new energy power generation technology and the electric vehicle industry, a large number of power electronic devices, such as charging piles, photovoltaic inverters, and wind power converters, have been connected to the distribution network. These devices have nonlinear, impulsive, and intermittent load characteristics, generating a large amount of harmonic currents and transient impacts during operation, leading to a serious deterioration of the power quality of the distribution network. In this complex grid environment, traditional electricity metering devices face severe challenges, with significantly increased metering errors, seriously affecting the fairness of electricity trading and the safe and stable operation of the power grid.

[0003] In existing technologies, electricity metering calibration mainly employs three methods: offline calibration, fixed parameter calibration, and single-dimensional error compensation. Offline calibration involves removing the electricity metering device from the field and sending it to a metrology verification institution for calibration in a standard laboratory environment. While this method offers high calibration accuracy, it has significant drawbacks: First, the calibration process requires interrupting the metering workflow, resulting in missing electricity metering data for that period and affecting the continuity of electricity trading; second, the calibration cycle is long, typically requiring several days or even weeks, failing to reflect changes in field operating conditions in a timely manner; and finally, the laboratory environment differs significantly from the actual field operating environment, often leading to deviations in calibration results when applied in the field.

[0004] Fixed parameter calibration refers to setting a set of fixed calibration coefficients before the electricity metering device leaves the factory, based on standard load conditions. These coefficients are then used directly for error correction during field operation. This method is simple to implement, low in cost, and suitable for conventional electricity metering scenarios with stable load characteristics and no complex harmonics or transient impacts. However, when a large number of nonlinear impulsive loads are connected to the distribution network, parameters such as load rate, harmonic order, and transient duration change rapidly over a wide range. Fixed calibration coefficients cannot adapt to these dynamic changes, leading to a sharp increase in metering errors.

[0005] Single-dimensional error compensation refers to establishing a corresponding error compensation model to correct metering errors for a specific influencing factor, such as load factor or harmonic content. For example, some existing technologies exist for harmonic error compensation, which corrects metering results based on preset harmonic error curves by detecting the harmonic content in the power grid. However, these methods only consider the impact of a single dimension of error, ignoring the coupling effect between errors of different dimensions. In reality, when the load factor, harmonic order, and transient duration change rapidly simultaneously, the metering error exhibits strongly coupled non-stationary characteristics. Error components of each dimension influence and superimpose each other. Single-dimensional compensation methods cannot accurately separate the error contributions of each dimension, thus failing to achieve effective error calibration.

[0006] In recent years, with the development of artificial intelligence technology, some scholars have proposed machine learning-based methods for calibrating electricity metering errors. These methods collect a large amount of historical data, train a complex machine learning model, and use the model to predict and compensate for metering errors. Although these methods have improved metering accuracy under complex operating conditions to some extent, the following problems still exist: First, model training requires massive amounts of historical data, which not only consumes a lot of storage space but also incurs very high costs for data collection and annotation; second, when the on-site operating conditions change significantly, the prediction accuracy of the original model will decrease significantly, requiring data to be collected again and the model to be retrained offline, which will also lead to metering interruptions; finally, traditional machine learning models are based on batch learning, requiring all historical data for each update, resulting in high computational complexity, poor real-time performance, and inability to meet the requirements of dynamic calibration.

[0007] In summary, existing technologies cannot effectively solve the calibration problem of three-dimensional coupled non-stationary metering errors under nonlinear impact loads. They suffer from defects such as metering interruption, low calibration accuracy, high data storage pressure, and poor real-time performance, which seriously restrict the development of power metering technology in new energy distribution networks. Summary of the Invention

[0008] The purpose of this invention is to provide a high-precision dynamic calibration method for power metering errors, and at the same time, to provide a high-precision dynamic calibration system for power metering errors, so as to solve the problems mentioned in the background art.

[0009] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0010] A high-precision dynamic calibration method for electrical energy metering errors includes the following steps:

[0011] The voltage and current analog signals of the metering circuit are collected in real time, and the analog-to-digital conversion is performed to obtain digital data. Simultaneously, three-dimensional feature data such as load rate, harmonic order, and transient duration are extracted. The acquisition cycle is consistent with the metering cycle of the power metering device.

[0012] Based on the three-dimensional feature data, a three-dimensional coupling error decoupling model is constructed, the model parameters are solved, and the mutually independent load rate error components, harmonic error components and transient error components are obtained.

[0013] Using high-precision standard electricity metering data as a reference, and based on the original metering data and the decoupled three-dimensional error components, a real-time metering error benchmark value is generated for the current moment without interrupting the normal metering process. An incremental learning algorithm is used to update the electricity metering error calibration model parameters online with the current three-dimensional feature data and the real-time metering error benchmark value as incremental samples, without storing the full historical samples.

[0014] Based on the updated calibration model, the metering error is predicted, the original metering results are compensated and corrected in real time, and the calibrated power metering data is output to maintain continuous metering operation.

[0015] The system monitors calibration errors in real time, and triggers rapid incremental updates when the error exceeds a preset threshold to maintain stable calibration accuracy.

[0016] Furthermore, the load factor is calculated by the ratio of the average active power to the rated active power within a metering cycle; the harmonic order is extracted by spectral analysis of the current data using fast Fourier transform; and the transient duration is calculated by detecting the start and end times of abrupt changes in the current signal.

[0017] Furthermore, the constructed three-dimensional coupling error decoupling model is a multivariate nonlinear regression model. The model coefficients are solved in real time using the recursive least squares method, thereby separating the error components of each dimension.

[0018] Furthermore, the step of using high-precision standard electrical energy metering data as a reference benchmark includes: using a standard electrical energy metering module to output standard electrical energy data; taking the difference between the original metering value and the standard metering value at the same time as the original total error; and obtaining the real-time metering error benchmark value after verification based on the decoupling results.

[0019] Furthermore, the stochastic gradient descent incremental learning algorithm is adopted to calculate the prediction error and gradient on each incremental sample, fine-tune the model weight parameters in the opposite direction of the gradient, and discard the incremental sample after the update is completed.

[0020] Furthermore, the real-time compensation and correction of the original metering results includes using a composite compensation method combining additive and multiplicative compensation. The formula for calculating the calibrated energy metering value is as follows: ,in The value after calibration. The original value, For the prediction error value, and This is the preset compensation coefficient.

[0021] Furthermore, the real-time monitoring of calibration error includes setting a preset error threshold based on the accuracy level of the power metering device. When the calibration error exceeds the threshold for three consecutive metering cycles, a rapid incremental update is triggered to increase the step size and frequency of model updates.

[0022] A high-precision dynamic calibration system for electrical energy metering errors includes:

[0023] A multi-dimensional synchronous acquisition module is used to acquire voltage and current signals in real time and extract three-dimensional feature data synchronously.

[0024] The error component decoupling module is connected to the multi-dimensional synchronous acquisition module and is used to build a decoupling model to separate the three-dimensional error components.

[0025] The standard metering reference module is connected in parallel with the multi-dimensional synchronous acquisition module into the metering loop to output standard electrical energy metering data;

[0026] The real-time error benchmark generation module is connected to the error component decoupling module and the standard metrology benchmark module respectively, and is used to generate real-time metrology error benchmark values.

[0027] The incremental learning calibration module is connected to the multi-dimensional synchronous acquisition module and the real-time error benchmark generation module, respectively, and is used to update the calibration model parameters online.

[0028] The dynamic compensation output module, connected to the incremental learning calibration module, is used to compensate and correct the original measurement results and output them.

[0029] The closed-loop calibration module is connected to both the standard metrology reference module and the dynamic compensation output module to monitor the calibration effect and trigger rapid updates.

[0030] Furthermore, the multi-dimensional synchronous acquisition module includes a high-precision voltage transformer, a high-precision current transformer, an analog-to-digital converter, a synchronous phase-locked loop unit, and a feature extraction unit; the synchronous phase-locked loop unit realizes the time synchronization of each signal, and the feature extraction unit extracts three-dimensional feature data.

[0031] Furthermore, the incremental learning calibration module includes a model initialization unit, an incremental sample receiving unit, a gradient calculation unit, a parameter update unit, and a model storage unit; the model storage unit only stores the updated incremental model parameters.

[0032] This application also discloses an electronic device, including:

[0033] At least one processor; and

[0034] A memory communicatively connected to the at least one processor; wherein,

[0035] The memory stores instructions that can be executed by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform the above-described high-precision dynamic calibration method for electrical energy metering errors of the present invention.

[0036] Beneficial effects: By constructing a three-dimensional coupling error decoupling model, this invention can separate the error contribution components of three dimensions—load rate, harmonic order, and transient duration—in real time, effectively solving the problem of strong coupling and non-stationarity of measurement errors under nonlinear impact loads, and significantly improving the measurement accuracy under complex working conditions.

[0037] This invention employs an incremental learning algorithm for model updates, using only incremental samples at the current moment for training and not storing all historical data, which greatly reduces data storage pressure and computational complexity. At the same time, the model update process is completed online, eliminating the need for offline retraining, avoiding calibration interruptions, and enabling continuous operation of the metrology process.

[0038] This invention designs a closed-loop verification mechanism for calibration effect, which can monitor calibration accuracy in real time. When the error exceeds the threshold, it automatically triggers a rapid incremental update to ensure the long-term stability of calibration accuracy and improve the reliability and robustness of the system.

[0039] The system structure of this invention is simple and easy to implement. It can be directly integrated into existing electricity metering devices without the need for large-scale modification of existing metering systems, and has broad application prospects and promotional value.

[0040] This invention employs a composite compensation method that combines additive and multiplicative compensation, which can more accurately correct different types of measurement errors and further improve the accuracy and effectiveness of calibration. Attached Figure Description

[0041] Figure 1 This is an overall flowchart of the high-precision power metering error dynamic calibration method of the present invention;

[0042] Figure 2 This is a flowchart of the three-dimensional coupling error decoupling calculation in an embodiment of the present invention;

[0043] Figure 3 This is a flowchart illustrating the real-time measurement error benchmark generation process in an embodiment of the present invention.

[0044] Figure 4 This is a flowchart of the incremental learning model update and dynamic compensation process in an embodiment of the present invention;

[0045] Figure 5 This is a graph showing the decoupling effect of three-dimensional error components in an embodiment of the present invention. Detailed Implementation

[0046] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0047] This invention provides a high-precision dynamic calibration method for power metering errors, such as... Figure 1 As shown, the steps include:

[0048] The voltage and current analog signals of the metering circuit are collected in real time, and the analog-to-digital conversion is performed to obtain digital data. Simultaneously, three-dimensional feature data such as load rate, harmonic order, and transient duration are extracted. The acquisition cycle is consistent with the metering cycle of the power metering device.

[0049] Based on the three-dimensional feature data, a three-dimensional coupling error decoupling model is constructed, the model parameters are solved, and the mutually independent load rate error components, harmonic error components and transient error components are obtained.

[0050] Using high-precision standard electrical energy metering data as a reference, and based on the original metering data and the decoupled three-dimensional error components, a real-time metering error reference value is generated for the current moment without interrupting the normal metering process.

[0051] An incremental learning algorithm is used to update the parameters of the power metering error calibration model online using the current three-dimensional feature data and the real-time metering error benchmark as incremental samples, without storing the full historical samples.

[0052] Based on the updated calibration model, the metering error is predicted, the original metering results are compensated and corrected in real time, and the calibrated power metering data is output to maintain continuous metering operation.

[0053] The system monitors calibration errors in real time, and triggers rapid incremental updates when the error exceeds a preset threshold to maintain stable calibration accuracy.

[0054] The present invention will be further described in detail below with reference to specific embodiments. It should be understood that the specific embodiments described herein are only for explaining the present invention and are not intended to limit the present invention.

[0055] This embodiment provides a high-precision dynamic calibration method for electricity metering errors, applied to a distribution network area with centralized charging pile access. This area has 20 7kW AC charging piles and 10 120kW DC charging piles, and is also connected to a 500kW distributed photovoltaic system. During peak electricity consumption periods, the charging piles and photovoltaic inverters operate simultaneously, with the load rate fluctuating rapidly between 10% and 90%. Multiple harmonics, including the 3rd, 5th, 7th, 11th, and 13th harmonics, exist in the power grid, with transient impacts lasting between 10ms and 100ms. The metering error exhibits strongly coupled and non-stationary characteristics.

[0056] The high-precision dynamic calibration method for power metering errors in this embodiment specifically includes the following steps:

[0057] Step 1: Multi-dimensional synchronous acquisition; A high-precision acquisition unit with a sampling frequency of 12.8kHz is used to acquire the analog voltage and current signals of phases A, B, and C of the metering circuit in real time. A 16-bit analog-to-digital converter converts the analog signals into digital voltage and current data with a conversion accuracy of 0.01%. A GPS-based phase-locked loop (PLL) technology is employed to achieve time synchronization of the three-phase voltage signals, three-phase current signals, and subsequently extracted three-dimensional feature data, with the time synchronization error controlled within 0.5 microseconds.

[0058] Based on digitized voltage and current data, the average active power within each metering cycle (20ms) is calculated. The average active power is then divided by the rated active power of the metering device (1000kW) to obtain the load factor L at the current moment. For example, when the average active power is 500kW, the load factor L = 0.5.

[0059] Fast Fourier Transform (FFT) is used to perform spectral analysis on the current data for each metering cycle to obtain the spectral distribution of the current signal. A harmonic amplitude threshold of 1% of the rated current is set, and the order and amplitude of all harmonics with amplitudes greater than this threshold are extracted to form a harmonic order vector H. For example, when the amplitude of the 3rd harmonic is detected to be 5A, the amplitude of the 5th harmonic to be 3A, and the amplitude of the 7th harmonic to be 2A, the harmonic order vector H = [3, 5, 7], and the corresponding amplitude vector is [5, 3, 2].

[0060] Wavelet transform is used to detect abrupt changes in the current signal. The transient start point is determined when the rate of change of the current signal exceeds a preset threshold (10% of the rated current / ms); the transient end point is determined when the rate of change of the current signal returns to below the preset threshold. The time interval between the transient start and end points is calculated to obtain the transient duration T. For example, if the transient starts at t=100ms and ends at t=150ms, the transient duration T=50ms.

[0061] The load factor L, harmonic order vector H, and transient duration T extracted from each metering cycle are packaged into a three-dimensional feature vector X=[L,H,T] and transmitted to the subsequent error component decoupling module.

[0062] Step 2, decoupling calculation of error components; such as Figure 2 As shown, a three-dimensional coupling error decoupling model is constructed, and a multivariate nonlinear regression model is used to describe the relationship between total measurement error and three-dimensional features. The specific form of the model is as follows: ; Where E is the total metering error and L is the load factor. Let i be the order of the i-th harmonic. Let be the amplitude of the i-th harmonic, and T be the transient duration. These are the model coefficients.

[0063] The model coefficients are solved in real time using the recursive least squares (RLS) method. The basic idea of ​​recursive least squares is to use new observation data at each time step to correct the model coefficients from the previous time step, thereby obtaining the optimal coefficient estimate for the current time step. Its recursive formula is: ; ; ; Where k is the current time, The feature vector at the current time. This represents the total error observation value at the current moment. This represents the model coefficient vector at the current moment. Let be the error covariance matrix. Here is the gain matrix. The forgetting factor has a value of 0.99.

[0064] By solving the above model, the error components of each dimension are obtained: Load factor error component: ; Harmonic error components: ; Transient error components: ; Constant term error: ;

[0065] The total measurement error is the sum of the individual components: ;

[0066] The decoupling calculation process is completed within each metrology cycle, with a calculation time of less than 1ms, and does not affect subsequent calibration procedures.

[0067] Step 3: Real-time error benchmark generation; such as Figure 3 As shown, a high-precision standard energy meter with an accuracy class of 0.01 is connected in parallel in the metering circuit as a standard metering reference module. The standard energy meter and the energy metering device being calibrated use the same voltage transformer and current transformer to synchronously acquire voltage and current signals and calculate the standard energy value. .

[0068] In each metering cycle, the raw electrical energy value output by the calibrated device is measured. Compared with standard electrical energy value Compare the results and calculate the original total error:

[0069] The calculated original total error The total error E obtained from decoupling in step 2 is compared to verify the accuracy of the decoupling result. If the deviation between the two is less than 0.05%, the decoupling result is considered valid. This serves as the real-time measurement error benchmark value for the current moment; if the deviation is greater than 0.05%, the decoupling calculation is re-performed until the deviation meets the requirements.

[0070] The entire error benchmark generation process is carried out in parallel in the background without affecting the normal metrology process of the calibrated device. The metrology data is continuously output without interruption or loss.

[0071] Step 4: Incremental learning model update; such as Figure 4 As shown, the initialization model for electricity metering error calibration is a three-layer feedforward neural network. The input layer has three neurons, corresponding to the load rate, total harmonic content, and transient duration, respectively; the hidden layer has ten neurons, using the ReLU activation function; and the output layer has one neuron, corresponding to the predicted metering error value. The model's weight parameters are randomly initialized, with initial values ​​uniformly distributed between [-0.1, 0.1].

[0072] The stochastic gradient descent (SGD) incremental learning algorithm is used to update the model online. At each econometric interval, the current 3D feature vector is... (in The total harmonic content (equal to the square root of the sum of the squares of the amplitudes of all harmonics) and the corresponding real-time measurement error benchmark value. It is input into the model as an incremental training sample.

[0073] Calculate the model's predicted value on this sample. Then calculate the mean squared error loss function: ;

[0074] Calculate the gradient of the loss function with respect to the model weight parameters W: ;

[0075] Update the model weight parameters using gradient descent: ; in, The learning rate is initially set to 0.01 and gradually decreases as training progresses.

[0076] After the update is complete, the incremental training sample is immediately discarded without any storage. The model only retains the updated weight parameters, requiring less than 100KB of storage space, which greatly reduces storage pressure.

[0077] The model update process is completed within each metrology cycle, with an update time of less than 2ms, ensuring the real-time nature of the calibration.

[0078] Step 5: Dynamic calibration compensation output; Based on the updated calibration model, input the current three-dimensional feature vector to predict the measurement error value under the current operating conditions. .

[0079] A composite compensation method combining additive and multiplicative compensation is used to correct the original metering results. The formula for calculating the calibrated electricity metering value is as follows: ； in, This is the rated electrical energy value of the metering device.

[0080] The calculated calibrated electricity metering value The final metering output is used for electricity trading settlement and grid operation monitoring. The entire compensation process is implemented at the hardware level of the metering chip, with a latency of less than 1 microsecond. The metering process runs continuously without any interruption or switching.

[0081] Step 6: Closed-loop verification of calibration effect; real-time monitoring of calibrated power metering values. Compared with standard electrical energy value Calculate the calibration error based on the deviation between them: ;

[0082] The preset error threshold is 0.2% (corresponding to the accuracy requirement of a 0.5-level energy metering device). When the calibration error exceeds this threshold for three consecutive metering cycles... When all values ​​exceed ±0.2%, a fast incremental update process is triggered.

[0083] In the rapid incremental update process, the learning rate is... Increase the learning rate to 0.05, raising the model update frequency from once per measurement cycle to twice per measurement cycle, thereby increasing the model's adaptability to changes in operating conditions. Continue rapid updates until the calibration error for 10 consecutive measurement cycles recovers to within ±0.2%, then restore the learning rate and update frequency to normal values.

[0084] Through the aforementioned closed-loop verification mechanism, a decrease in calibration accuracy can be detected in a timely manner and adjustments can be made automatically, ensuring that the system can maintain stable high-precision measurement under various complex operating conditions.

[0085] like Figure 5 The figure shows the decoupling effect curves of the three-dimensional error components. The horizontal axis represents 30 metering cycles, and the vertical axis represents the percentage of error components. The four curves independently represent the load rate, harmonics, transient error, and total error. Data shows that under steady-state conditions, the transient error components... Constantly 0, load factor error Harmonic error Independent and smooth changes, without cross-coupling; periods 7-11 and 21-24 are transient impact intervals. Independently activated and linearly decaying, , The original pattern of change remains undisturbed; the total error E is strictly equal to + + This method perfectly achieves linear decoupling and independent quantization of error components. Each error component has a clear physical boundary and is free from coupling interference, enabling precise separation of error sources under different operating conditions and providing a unique solution for accuracy compensation in the metering system.

[0086] This application also provides an embodiment of an electronic device. The electronic device is manifested in the form of a general-purpose computing device. The components of the electronic device may include, but are not limited to: one or more processors or processing units, memory, and buses connecting different components (including memory and processing units).

[0087] A bus refers to one or more of several bus architectures, including memory buses or memory controllers, peripheral buses, graphics acceleration ports, processors, or local buses using any of the various bus architectures. Examples of these architectures include, but are not limited to, Industry Standard Architecture (ISA) buses, Micro Channel Architecture (MCA) buses, Enhanced ISA buses, Video Electronics Standards Association (VESA) local buses, and Peripheral Component Interconnect (PCI) buses.

[0088] Electronic devices typically include a variety of computer-readable media. These media can be any available media that can be accessed by the electronic device, including volatile and non-volatile media, and removable and non-removable media.

[0089] The memory may include computer-readable media in the form of volatile memory, such as random access memory (RAM) and / or cache memory. Electronic devices may further include other removable / non-removable, volatile / non-volatile computer device storage media. By way of example only, the storage system may be used to read and write non-removable, non-volatile magnetic media.

[0090] The electronic device can also communicate with one or more external devices (e.g., keyboard, pointing device, camera, etc.), may include a display, and may communicate with one or more devices that enable a user to interact with the electronic device, and / or with any device that enables the electronic device to communicate with one or more other computing devices (e.g., network card, modem, etc.). This communication can be performed via an input / output (I / O) interface. Furthermore, the electronic device can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN)) and / or public networks, such as the Internet) via a network adapter. The network adapter communicates with other modules of the electronic device via a bus. The processor executes various functional applications and data processing by running programs stored in memory, such as implementing the high-precision dynamic calibration method for electrical energy metering errors provided in the above embodiments of the present invention.

[0091] This application also discloses a high-precision dynamic calibration system for electrical energy metering errors, including:

[0092] A multi-dimensional synchronous acquisition module is used to acquire voltage and current signals in real time and extract three-dimensional feature data synchronously.

[0093] The error component decoupling module is connected to the multi-dimensional synchronous acquisition module and is used to build a decoupling model to separate the three-dimensional error components.

[0094] The standard metering reference module is connected in parallel with the multi-dimensional synchronous acquisition module into the metering loop to output standard electrical energy metering data;

[0095] The real-time error benchmark generation module is connected to the error component decoupling module and the standard metrology benchmark module respectively, and is used to generate real-time metrology error benchmark values.

[0096] The incremental learning calibration module is connected to the multi-dimensional synchronous acquisition module and the real-time error benchmark generation module, respectively, and is used to update the calibration model parameters online.

[0097] The dynamic compensation output module, connected to the incremental learning calibration module, is used to compensate and correct the original measurement results and output them.

[0098] The closed-loop calibration module is connected to both the standard metrology reference module and the dynamic compensation output module to monitor the calibration effect and trigger rapid updates.

[0099] Furthermore, the multi-dimensional synchronous acquisition module includes a high-precision voltage transformer, a high-precision current transformer, an analog-to-digital converter, a synchronous phase-locked loop unit, and a feature extraction unit; the synchronous phase-locked loop unit realizes the time synchronization of each signal, and the feature extraction unit extracts three-dimensional feature data.

[0100] Furthermore, the incremental learning calibration module includes a model initialization unit, an incremental sample receiving unit, a gradient calculation unit, a parameter update unit, and a model storage unit; the model storage unit only stores the updated incremental model parameters.

[0101] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features.

Claims

1. A high-precision dynamic calibration method for electrical energy metering errors, characterized in that, Includes the following steps: The voltage and current analog signals of the metering circuit are collected in real time, and the analog-to-digital conversion is performed to obtain digital data. Simultaneously, three-dimensional feature data such as load rate, harmonic order, and transient duration are extracted. The acquisition cycle is consistent with the metering cycle of the power metering device. Based on the three-dimensional feature data, a three-dimensional coupling error decoupling model is constructed, the model parameters are solved, and the mutually independent load rate error components, harmonic error components and transient error components are obtained. Using high-precision standard electrical energy metering data as a reference, and based on the original metering data and the decoupled three-dimensional error components, a real-time metering error reference value is generated for the current moment without interrupting the normal metering process. An incremental learning algorithm is used to update the parameters of the power metering error calibration model online using the current three-dimensional feature data and the real-time metering error benchmark as incremental samples, without storing the full historical samples. Based on the updated calibration model, the metering error is predicted, the original metering results are compensated and corrected in real time, and the calibrated power metering data is output to maintain continuous metering operation. The system monitors calibration errors in real time, and triggers rapid incremental updates when the error exceeds a preset threshold to maintain stable calibration accuracy.

2. The high-precision power metering error dynamic calibration method according to claim 1, characterized in that, The load factor is calculated as the ratio of the average active power to the rated active power over a metering cycle; the harmonic order is extracted by spectral analysis of the current data using fast Fourier transform; and the transient duration is calculated by detecting the start and end times of abrupt changes in the current signal.

3. The high-precision power metering error dynamic calibration method according to claim 1, characterized in that, The constructed three-dimensional coupling error decoupling model is a multivariate nonlinear regression model. The model coefficients are solved in real time using the recursive least squares method, and then the error components of each dimension are separated.

4. The high-precision power metering error dynamic calibration method according to claim 1, characterized in that, The method of using high-precision standard electrical energy metering data as a reference includes: using a standard electrical energy metering module to output standard electrical energy data; taking the difference between the original metering value and the standard metering value at the same time as the original total error; and obtaining the real-time metering error reference value after verification based on the decoupling results.

5. The high-precision power metering error dynamic calibration method according to claim 1, characterized in that, The stochastic gradient descent incremental learning algorithm is used to calculate the prediction error and gradient on each incremental sample, fine-tune the model weight parameters in the opposite direction of the gradient, and discard the incremental sample after the update is completed.

6. The high-precision power metering error dynamic calibration method according to claim 1, characterized in that, The real-time compensation and correction of the original metering results includes using a composite compensation method combining additive and multiplicative compensation. The formula for calculating the calibrated energy metering value is as follows: ,in The value after calibration. The original value, For the prediction error value, and This is the preset compensation coefficient.

7. The high-precision power metering error dynamic calibration method according to claim 1, characterized in that, The real-time monitoring of calibration error includes setting a preset error threshold based on the accuracy level of the power metering device. When the calibration error exceeds the threshold for three consecutive metering cycles, a rapid incremental update is triggered, increasing the step size and frequency of model updates.

8. A system applying the high-precision dynamic calibration method for electrical energy metering errors as described in claim 1, characterized in that, include: A multi-dimensional synchronous acquisition module is used to acquire voltage and current signals in real time and extract three-dimensional feature data synchronously. The error component decoupling module is connected to the multi-dimensional synchronous acquisition module and is used to build a decoupling model to separate the three-dimensional error components. The standard metering reference module is connected in parallel with the multi-dimensional synchronous acquisition module into the metering loop to output standard electrical energy metering data; The real-time error benchmark generation module is connected to the error component decoupling module and the standard metrology benchmark module respectively, and is used to generate real-time metrology error benchmark values. The incremental learning calibration module is connected to the multi-dimensional synchronous acquisition module and the real-time error benchmark generation module, respectively, and is used to update the calibration model parameters online. The dynamic compensation output module, connected to the incremental learning calibration module, is used to compensate and correct the original measurement results and output them. The closed-loop calibration module is connected to both the standard metrology reference module and the dynamic compensation output module to monitor the calibration effect and trigger rapid updates.

9. The high-precision power metering error dynamic calibration system according to claim 8, characterized in that, The multi-dimensional synchronous acquisition module includes a high-precision voltage transformer, a high-precision current transformer, an analog-to-digital converter, a synchronous phase-locked loop unit, and a feature extraction unit; the synchronous phase-locked loop unit realizes the time synchronization of each signal, and the feature extraction unit extracts three-dimensional feature data.

10. The high-precision power metering error dynamic calibration system according to claim 8, characterized in that, The incremental learning calibration module includes a model initialization unit, an incremental sample receiving unit, a gradient calculation unit, a parameter update unit, and a model storage unit; the model storage unit only stores the updated incremental model parameters.

Images