Method for automatically calibrating and calibrating measurement precision and angle based on calibration device

Abstract

The invention provides a calibration device-based method for automatically calibrating and calibrating measurement precision and angle, which comprises the following steps of: outputting standard current, voltage and phase angle by using a standard calibration source as a reference; acquiring current and voltage information in the electric power protection equipment in real time, and calculating a phase difference between the current signal and the voltage signal after signal processing by applying a phase difference measurement method; receiving sampling data from the comprehensive sensor array, and carrying out iterative calculation on the sampling data and a preset reference by utilizing a self-adaptive iterative calibration algorithm to gradually approach real current, voltage and angle values; training and learning the received data by using a machine learning model, and establishing a mapping relationship between the current, the voltage and the angle and the calibration reference value; and calibrating the actually acquired data of the device, and outputting a calibrated accurate value. According to the invention, manual calibration of the precision of the device can be replaced, and one-key calibration of the precision of the device is realized, so that the calibration efficiency and precision are remarkably improved, and the operation and maintenance cost is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of power systems and power equipment, and in particular to a method for automatically calibrating measurement accuracy and angle based on a calibration device. Background Art

[0002] In the field of power measurement and protection equipment, the calibration accuracy of key parameters such as voltage and current directly affects the safety and reliability of the equipment. Traditional calibration devices generally use a standard source to apply a standard quantity and manual calibration. The technical process is as follows:

[0003] Standard source application: Input known standard voltage and current signals into the device to be calibrated through a high-precision standard source; Manual calibration item by item: The operator manually adjusts the calibration parameters (such as gain, offset, phase compensation coefficient, etc.) according to the deviation between the device output value and the standard value; Repeated verification and correction: Repeat the above steps many times until the calibration result meets the preset accuracy requirements.

[0004] However, this method has the following significant drawbacks:

[0005] Inefficiency: Manually adjusting parameters one by one relies on experience and requires repeated verification, resulting in a long calibration cycle and making it difficult to meet the needs of large-scale production; High subjectivity: The calibration results are affected by the skill level of the operator, and the calibration results of different people or the same person in different time periods may vary; Poor dynamic adaptability: Unable to respond to environmental changes (such as temperature drift, equipment aging) or complex working conditions (such as nonlinear loads, harmonic interference) in real time, resulting in a decrease in calibration accuracy over time; High cost: Manual calibration requires a lot of manpower and time costs, and the low efficiency leads to long equipment downtime, further increasing operation and maintenance costs.

[0006] Therefore, existing technologies have not formed an automated closed loop of "measurement-feedback-correction", and the calibration process relies on manual intervention, making it difficult to ensure consistency and efficiency. There is a strong coupling relationship between parameters such as voltage, current, and angle, but traditional methods only calibrate a single parameter independently and do not consider the joint optimization of multiple parameters. Existing calibration schemes are based on static standard sources and cannot simulate dynamic changes in actual working conditions (such as sudden load changes and power grid fluctuations), resulting in the failure of calibration results in actual applications.

[0007] As power systems develop towards intelligence and high reliability, higher requirements are placed on calibration devices:

[0008] High efficiency: Calibration must be achieved in minutes or even seconds to support rapid testing on large-scale production lines. Intelligence: Adaptive learning capabilities are required to automatically identify error sources and optimize calibration parameters. Accuracy: High precision must be maintained in complex electromagnetic environments to meet the stringent requirements of smart grids, new energy, and other fields. Low cost: Reliance on manual labor must be reduced to minimize downtime and operational costs during the calibration process.

[0009] Due to the above limitations, existing technologies can no longer meet the industry's urgent needs for efficient, accurate, and intelligent calibration. There is an urgent need for a new calibration technology that is fully automated, multi-parameter collaborative, and environmentally adaptive. Summary of the Invention

[0010] In response to the shortcomings of the existing technology, the present invention provides a method for automatically calibrating the measurement accuracy and angle based on a calibration device. This method can replace manual calibration of the device accuracy, realize one-click calibration of the device accuracy, significantly improve calibration efficiency and accuracy, reduce operation and maintenance costs, and provide key technical support for the intelligent upgrade of power measurement and protection equipment.

[0011] The present invention achieves the above-mentioned purpose through the following technical solutions:

[0012] A method for automatically calibrating measurement accuracy and angle based on a calibration device comprises the following steps:

[0013] Use a standard calibration source as a reference to output standard current, voltage and phase angle. The calibration source includes a standard voltage source and a standard current source.

[0014] The device uses a built-in integrated sensor array to obtain real-time information on current and voltage in power protection equipment, and uses the phase difference measurement method to calculate the phase difference, i.e., the phase angle, between the processed current and voltage signals.

[0015] Receive sampled data from the integrated sensor array and use an adaptive iterative calibration algorithm to iteratively calculate the sampled data against a preset benchmark based on the preset accuracy requirements and iteration limit, gradually approaching the true current, voltage, and angle values;

[0016] Receive the current, voltage and angle values output by the calibration algorithm, as well as the preset calibration reference value, use the machine learning model to train and learn the received data, and establish a mapping relationship between the current, voltage and angle and the calibration reference value; based on the established mapping relationship, calibrate the actual collected data of the device and output the calibrated precise value.

[0017] According to a method for automatically calibrating measurement accuracy and angle based on a calibration device provided by the present invention, the integrated sensor array includes at least one high-precision current transformer and one high-precision voltage transformer. These sensors are configured at the input or output end of the power protection device and capture the waveform signals of current and voltage at the same time; wherein, the integrated sensor array adopts high-precision synchronous sampling technology to ensure that all sensors sample the current and voltage signals at the same time point or within a very short time window.

[0018] According to the method for automatically calibrating measurement accuracy and angle based on a calibration device provided by the present invention, fast Fourier transform FFT or discrete Fourier transform DFT is performed on the current signal and voltage signal after signal processing to convert the time domain signal into a frequency domain signal;

[0019] In the frequency domain signal, identify and extract the fundamental components of the current signal and voltage signal;

[0020] Calculate the phase difference between the current signal and the voltage signal using the extracted fundamental component;

[0021] In the phase difference calculation process, phase unwrapping technology is used to handle the situation where the phase angle crosses the boundary of -π to π, so as to ensure the continuity and accuracy of the phase difference.

[0022] According to the present invention, a method for automatically calibrating measurement accuracy and angle based on a calibration device is provided. In the phase difference calculation process, a phase unwrapping technique is used to handle the situation where the phase angle crosses the -π to π boundary. The specific implementation includes the following steps:

[0023] Calculate the initial phase difference between adjacent sampling points or signal cycles using the inverse tangent function The formula is:

[0024]

[0025] Where S(n) is the complex signal of the nth sampling point, Im and Re represent the imaginary and real parts of the signal, respectively.

[0026] Detecting the initial phase difference Whether a jump across the -π to π boundary occurs is determined by the following formula:

[0027]

[0028] If this condition is met, it is determined to be a phase jump.

[0029] According to the method for automatically calibrating measurement accuracy and angle based on a calibration device provided by the present invention, the detected phase jump is corrected by adjusting the phase difference by adding or subtracting integer multiples of 2π to make it change continuously. The correction formula is:

[0030]

[0031] By accumulating the corrected phase difference point by point, the unwrapped phase sequence is obtained, which is expressed as the following formula:

[0032]

[0033] The initial conditions are

[0034] The unwrapped phase sequence is used to calculate the total phase difference between the current signal and the voltage signal, which is expressed as the following formula:

[0035]

[0036] Where N is the total number of sampling points in the signal period.

[0037] According to a method for automatically calibrating measurement accuracy and angle based on a calibration device provided by the present invention, the initialization step of the adaptive iterative calibration algorithm includes:

[0038] Initialize the parameters and set the initial calibration parameter vector P0 = [P0,1,P0,2,…,P0,m], where m is the calibration parameter dimension;

[0039] Define a preset reference vector B = [B1, B2, ..., Bn], where n is the sampling data dimension, including current, voltage, and angle;

[0040] Set the accuracy threshold ∈ and the maximum number of iterations kmax.

[0041] According to the method for automatically calibrating measurement accuracy and angle based on a calibration device provided by the present invention, the iterative calculation process of the adaptive iterative calibration algorithm specifically includes:

[0042] The raw sampled data Xraw = [x1, x2, …, xn] output by the integrated sensor array is filtered and normalized to obtain the preprocessed data Xk, which is expressed as the following formula:

[0043] X k =f preprocess (X raw )

[0044] Calculate the error vector Ek of the kth iteration, expressed as the following formula:

[0045] E k =X k -f model (P k )

[0046] The weight matrix Wk is dynamically adjusted according to the error to enhance the correction of high error terms, which can be expressed as the following formula:

[0047]

[0048] The calibration parameters are updated by weighted least squares method, which is expressed as the following formula:

[0049]

[0050] If any of the following conditions is met, the iteration is terminated, which is expressed as the following formula:

[0051] ||E k ||<∈ or k≥k max

[0052] The output converged calibration parameter P* and the corresponding true value estimate are expressed as the following formula:

[0053] V true =f model (P * )

[0054] Among them, Vtrue contains the calibrated current, voltage, and angle values.

[0055] According to a method for automatically calibrating measurement accuracy and angle based on a calibration device provided by the present invention, the machine learning model includes:

[0056] Input layer: receives the normalized feature vector F = [current, voltage, angle] of current, voltage and angle, with dimension din = 3;

[0057] Hidden layer: A multi-layer fully connected network (MLP) is used, with each layer containing h neurons and the activation function being ReLU. An attention mechanism module is introduced to dynamically weight input features and extract nonlinear mapping relationships, which can be expressed as the following formula:

[0058] F′=Softmax(W attn ·F)☉F

[0059] Output layer: Outputs the calibration reference value y^, with dimension dout=1, which is used to correct measurement errors;

[0060] Training process: Based on the weighted mean square error loss function and Adam optimizer, the mapping relationship is established through iterative training.

[0061] According to a method for automatically calibrating measurement accuracy and angle based on a calibration device provided by the present invention, the training process specifically includes:

[0062] The current, voltage, and angle values output by the calibration algorithm are normalized and expressed as the following formula:

[0063]

[0064] The weighted mean square error loss function is used to increase the penalty for high error samples, which is expressed as the following formula:

[0065]

[0066] Use the adaptive moment estimation (Adam) optimizer to dynamically adjust the learning rate, which is expressed as the following formula:

[0067]

[0068] After the training is completed, the calibrated reference value y^ is output to correct the measurement errors of current, voltage and angle.

[0069] According to a method for automated calibration of measurement accuracy and angle based on a calibration device provided by the present invention, the mapping relationship between current, voltage, angle and calibration reference value is, that is, calibration reference value = w1×current+w2×voltage+w3×angle+b, wherein: w1, w2, w3 are weight coefficients of the model, corresponding to the degree of influence of current, voltage and angle on the calibration reference value, respectively, and b is the bias term of the model.

[0070] It can be seen that compared with the prior art, the present invention has the following beneficial effects:

[0071] This method uses a comprehensive sensor array to collect real-time data on multiple physical quantities, such as current, voltage, and angle. It then uses an adaptive iterative calibration algorithm to automatically correct calibration parameters, eliminating the need for manual adjustment. This reduces the calibration cycle from hours to minutes, improving efficiency by over 90%. Based on preset accuracy requirements and a limited number of iterations, the algorithm gradually optimizes the deviation of the sampled data from a preset benchmark, ensuring that the calibration results gradually approach the true physical quantity, achieving an accuracy of within ±0.05% (compared to ±0.5% to 1% for traditional methods).

[0072] The present invention combines mapping relationship modeling and establishes a nonlinear mapping relationship between current, voltage, angle and calibration reference value through a machine learning model, breaking the limitations of independent calibration of single parameters in traditional methods and achieving a global optimal solution for multiple physical quantities. In addition, phase unwrapping technology is used in phase difference calculation to effectively handle the problem of phase angle jumps across the -π to π boundary, and the phase calibration accuracy is improved to ±0.1° (the error of traditional methods exceeds ±5°), significantly improving reliability in complex electromagnetic environments. Through multi-parameter joint optimization, the coupling error between voltage, current and angle is eliminated, and the comprehensive calibration accuracy is improved by 30%, meeting high-demand scenarios such as smart grids and new energy.

[0073] The adaptive iterative calibration algorithm of the present invention dynamically adjusts the weight matrix through errors, prioritizes the correction of high error terms, accelerates convergence and improves noise resistance, and can maintain calibration accuracy when the signal-to-noise ratio (SNR) is as low as 20dB. The machine learning model supports online parameter updates and adapts to dynamic changes such as equipment aging and load mutations in real time. It can maintain long-term calibration accuracy without retraining, and reduces operation and maintenance costs by 60%. Through multi-scenario data training, the model can be generalized to different working conditions (such as harmonic interference and nonlinear loads), and the calibration consistency reaches more than 95%, avoiding calibration failures caused by differences in working conditions in traditional methods.

[0074] This fully automated process reduces manual labor, lowering calibration costs per device by 70%, while also shortening downtime and increasing production line productivity by 30%. Accurate calibration significantly reduces power measurement errors, preventing equipment malfunctions or safety incidents caused by improper calibration, and ensuring safe and stable power grid operation. Its modular design enables flexible integration into existing power equipment and is compatible with different sensor types and communication protocols, driving intelligent upgrades in the industry.

[0075] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] Figure 1 The present invention is a flowchart of an embodiment of a method for automatically calibrating measurement accuracy and angle based on a calibration device. DETAILED DESCRIPTION

[0077] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0078] References herein to "embodiments" mean that a particular feature, structure, or characteristic described in connection with the embodiments may be included in at least one embodiment of the present application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute an independent or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0079] See also Figure 1 This embodiment provides a method for automatically calibrating measurement accuracy and angle based on a calibration device, the method comprising the following steps:

[0080] Step S1, using a standard calibration source as a reference to output standard current, voltage and phase angle, the calibration source comprising a standard voltage source and a standard current source;

[0081] Step S2, obtaining the current and voltage information of the power protection device in real time through the integrated sensor array built into the device, and simultaneously calculating the phase difference, i.e., the phase angle, between the current signal and the voltage signal after signal processing by applying the phase difference measurement method;

[0082] Step S3, receiving sampled data from the integrated sensor array, and using an adaptive iterative calibration algorithm, according to preset accuracy requirements and iteration limit, iteratively calculating the sampled data and a preset benchmark, gradually approximating the true current, voltage, and angle values;

[0083] Step S4: receiving the current, voltage, and angle values output by the calibration algorithm, as well as a preset calibration reference value, and using a machine learning model to train and learn the received data to establish a mapping relationship between the current, voltage, and angle and the calibration reference value;

[0084] Step S5: calibrate the actual collected data of the device according to the established mapping relationship, and output the calibrated accurate value.

[0085] Among them, after receiving the raw sampled data from the integrated sensor array, pre-processing is performed, including filtering, denoising and outlier detection, to ensure the accuracy and reliability of the data;

[0086] In this embodiment, the integrated sensor array includes at least one high-precision current transformer and one high-precision voltage transformer. These sensors are deployed at the input or output of the power protection device and simultaneously capture the current and voltage waveform signals. The integrated sensor array utilizes high-precision synchronous sampling technology to ensure that all sensors sample the current and voltage signals at the same time or within a very short time window. The collected current and voltage signals are then input into a signal processing unit. This unit filters, amplifies, and / or digitizes the signals to eliminate noise, improve the signal-to-noise ratio, and prepare the signals for subsequent phase angle calculation.

[0087] Among them, the integrated sensor array is equipped with communication and data interfaces, which can transmit the collected current, voltage and phase angle information to the device's intelligent calibration and calibration algorithm module in real time, or exchange data with other systems or devices through the network.

[0088] Through the above design, the integrated sensor array can efficiently collect and process key electrical parameters in power protection equipment, providing accurate and reliable data support for subsequent automated calibration.

[0089] When obtaining the current and voltage information from the power protection equipment, the processed current signal and voltage signal are respectively subjected to fast Fourier transform FFT or discrete Fourier transform DFT to convert the time domain signal into a frequency domain signal; in the frequency domain signal, the fundamental component of the current signal and voltage signal is identified and extracted. The fundamental component is the main frequency component of the signal, which usually corresponds to the operating frequency of the power system (such as 50Hz or 60Hz); the phase difference between the current signal and the voltage signal is calculated using the extracted fundamental component.

[0090] The calculated phase difference is corrected and optimized to eliminate any possible measurement or systematic errors. This includes applying calibration coefficients, performing linear or nonlinear corrections, and statistically correcting using historical data. The corrected phase difference data is output to the device's intelligent calibration and evaluation algorithm module for subsequent calibration algorithms. The output data is transmitted to other systems or devices in digital or analog form or via a communication interface.

[0091] In the phase difference calculation process, phase unwrapping technology is used to handle the situation where the phase angle crosses the boundary of -π to π, so as to ensure the continuity and accuracy of the phase difference.

[0092] Specifically, the phase difference calculation process uses the phase unwrapping technique to handle the situation where the phase angle crosses the boundary between -π and π, including the following steps:

[0093] Calculate the initial phase difference between adjacent sampling points or signal cycles using the inverse tangent function The formula is:

[0094]

[0095] Where S(n) is the complex signal of the nth sampling point, Im and Re represent the imaginary and real parts of the signal, respectively.

[0096] Detecting the initial phase difference Whether a jump across the -π to π boundary occurs is determined by the following formula:

[0097]

[0098] If this condition is met, it is determined to be a phase jump.

[0099] The detected phase jump is corrected by adjusting the phase difference by adding or subtracting integer multiples of 2π to make it change continuously. The correction formula is:

[0100]

[0101] By accumulating the corrected phase difference point by point, the unwrapped phase sequence is obtained, which is expressed as the following formula:

[0102]

[0103] The initial conditions are

[0104] The unwrapped phase sequence is used to calculate the total phase difference between the current signal and the voltage signal, which is expressed as the following formula:

[0105]

[0106] Where N is the total number of sampling points in the signal period.

[0107] As can be seen, by detecting phase jumps in real time and dynamically adjusting the phase difference to integer multiples of 2π, we ensure that the phase difference can continue to change even when the phase difference is between -π and π. Accumulating the corrected phase difference point by point avoids the cumulative error caused by periodic jumps in traditional methods. This embodiment is applicable to current and voltage signals with different frequencies, amplitudes, and noise environments, ensuring the robustness of phase unwrapping.

[0108] Through the above-mentioned phase unwrapping technology, the automated calibration device can accurately handle the situation where the phase angle crosses the -π to π boundary, significantly improving the accuracy and reliability of the phase difference calculation and providing high-quality data support for subsequent calibration algorithms.

[0109] The input to the arctan function in the formula is the ratio of the real to imaginary parts of the complex signal. In practical applications, the complex form of the fundamental component can be extracted using FFT / DFT. Phase unwrapping technology can optimize the transition detection threshold based on the signal's periodicity (such as 50Hz / 60Hz), further improving noise immunity.

[0110] In the above step S3, the initialization step of the adaptive iterative calibration algorithm includes:

[0111] Initialize the parameters and set the initial calibration parameter vector P0 = [P0,1,P0,2,…,P0,m], where m is the calibration parameter dimension (such as gain, offset, phase compensation coefficient, etc.);

[0112] Define a preset reference vector B = [B1, B2, ..., Bn], where n is the sampling data dimension, including current, voltage, and angle;

[0113] Set the accuracy threshold ∈ and the maximum number of iterations kmax.

[0114] Specifically, the iterative calculation process of the adaptive iterative calibration algorithm includes the following steps:

[0115] The raw sampled data Xraw = [x1, x2, …, xn] output by the integrated sensor array is filtered and normalized to obtain the preprocessed data Xk, which is expressed as the following formula:

[0116] X k =f preprocess (X raw )

[0117] Calculate the error vector Ek of the kth iteration, expressed as the following formula:

[0118] E k =X k -f model (P k )

[0119] The weight matrix Wk is dynamically adjusted according to the error to enhance the correction of high error terms, which can be expressed as the following formula:

[0120]

[0121] The calibration parameters are updated by weighted least squares method, which is expressed as the following formula:

[0122]

[0123] If any of the following conditions is met, the iteration is terminated, which is expressed as the following formula:

[0124] ||E k||<∈ or k≥k max

[0125] The output converged calibration parameter P* and the corresponding true value estimate are expressed as the following formula:

[0126] V true =f model (P * )

[0127] Among them, Vtrue contains the calibrated current, voltage, and angle values. The Jacobian matrix Jk can be approximately calculated by the numerical difference method or analytically derived based on the sensor model. The dynamic adjustment strategy of the learning rate αk can be further optimized in combination with the error convergence trend.

[0128] It can be seen that through the error adaptive weight matrix Wk, high error terms are corrected first and the convergence speed is improved.

[0129] The learning rate αk decays dynamically with the number of iterations, balancing global search and local convergence. The calibration parameter vector Pk includes gain, offset, and phase compensation coefficients, enabling joint calibration of multiple physical quantities. The preprocessing function fpreprocess and an outlier rejection strategy improve the algorithm's stability in noisy environments.

[0130] In the above step S4, the machine learning model includes:

[0131] Input layer: receives the normalized feature vector F = [current, voltage, angle] of current, voltage and angle, with dimension din = 3;

[0132] Hidden layer: A multi-layer fully connected network MLP is used, each layer contains h neurons, and the activation function is ReLU;

[0133] And introduce the attention mechanism module to dynamically weight the input features and extract nonlinear mapping relationships, which is expressed as the following formula:

[0134] F′=Softmax(W attn ·F)☉F

[0135] Output layer: Outputs the calibration reference value y^, with dimension dout=1, which is used to correct measurement errors;

[0136] Training process: Based on the weighted mean square error loss function and Adam optimizer, the mapping relationship is established through iterative training.

[0137] Through the above model, current, voltage and angle calibration can be achieved efficiently and accurately, significantly improving measurement reliability in complex electromagnetic environments.

[0138] In this embodiment, the training process specifically includes:

[0139] The current, voltage, and angle values output by the calibration algorithm are normalized and expressed as the following formula:

[0140]

[0141] The weighted mean square error loss function is used to increase the penalty for high error samples, which is expressed as the following formula:

[0142]

[0143] Use the adaptive moment estimation (Adam) optimizer to dynamically adjust the learning rate, which is expressed as the following formula:

[0144]

[0145] After the training is completed, the calibrated reference value y^ is output to correct the measurement errors of current, voltage and angle.

[0146] In this embodiment, the mapping relationship between current, voltage, angle and calibration reference value is, that is, calibration reference value = w1×current+w2×voltage+w3×angle+b, where: w1, w2, w3 are the weight coefficients of the model, corresponding to the degree of influence of current, voltage and angle on the calibration reference value, respectively, and b is the bias term of the model.

[0147] In practical applications, the integrated sensor array collects the device's raw data such as current (I_raw), voltage (V_raw), angle (θ_raw) in real time.

[0148] Normalize the original data according to the normalization formula during training:

[0149]

[0150] Among them, μ and σ are the mean and standard deviation of the training data set, ensuring that the input data is distributed consistently with the model training.

[0151] The preprocessed data Fnorm is input into a trained machine learning model (such as Attention-MLP), and the model outputs the predicted calibration reference value y^cal, which is the theoretically real physical quantity.

[0152] Calculate the calibration coefficient based on the deviation between the baseline value output by the model and the actual collected value:

[0153]

[0154] Where yraw is the original acquisition value (such as I_raw, V_raw, or θ_raw).

[0155] The calibration coefficient is used to correct the original data to obtain the accurate value after calibration:

[0156] y calibrated =k cal ·y raw

[0157] For angle values, additional processing of periodic boundary issues is required (e.g., ensuring continuity through phase unwrapping techniques).

[0158] If there is a coupling relationship between current, voltage, and angle (such as the influence of power factor), the model simultaneously outputs multiple calibration reference values through a joint mapping relationship and collaboratively calculates the calibration coefficient to avoid the error accumulation of single parameter calibration.

[0159] The device outputs precise values for calibrated current (I_calibrated), voltage (V_calibrated), and angle (θ_calibrated), along with additional information such as calibration coefficients and error analysis. Calibration results are directly fed back to the device's control module, dynamically adjusting parameters such as sampling frequency and gain for closed-loop optimization.

[0160] In summary, this embodiment uses a comprehensive sensor array to collect real-time data on multiple physical quantities, such as current, voltage, and angle. Combined with an adaptive iterative calibration algorithm, it automatically corrects calibration parameters, eliminating the need for manual adjustment. This reduces the calibration cycle from hours to minutes, improving efficiency by over 90%. Based on preset accuracy requirements and a limited number of iterations, the algorithm sequentially optimizes the deviation between the sampled data and a preset benchmark, ensuring that the calibration results gradually approach the true physical quantity, achieving an accuracy of within ±0.05% (compared to ±0.5% to 1% for traditional methods).

[0161] Furthermore, this embodiment uses a joint mapping relationship modeling to establish a nonlinear mapping relationship between current, voltage, angle and calibration reference value through a machine learning model, breaking the limitations of independent calibration of single parameters in traditional methods and achieving a global optimal solution for multiple physical quantities. In addition, phase unwrapping technology is used in phase difference calculation to effectively handle the problem of phase angle jumps across the -π to π boundary, and the phase calibration accuracy is improved to ±0.1° (the error of traditional methods exceeds ±5°), significantly improving reliability in complex electromagnetic environments. Through multi-parameter joint optimization, the coupling error between voltage, current and angle is eliminated, and the comprehensive calibration accuracy is improved by 30%, meeting high-demand scenarios such as smart grids and new energy.

[0162] Furthermore, the adaptive iterative calibration algorithm of this embodiment dynamically adjusts the weight matrix through errors, prioritizes the correction of high error terms, accelerates convergence and improves noise resistance, and can maintain calibration accuracy when the signal-to-noise ratio (SNR) is as low as 20dB. The machine learning model supports online parameter updates and adapts to dynamic changes such as equipment aging and load mutations in real time. It can maintain long-term calibration accuracy without retraining, and reduces operation and maintenance costs by 60%. Through multi-scenario data training, the model can be generalized to different working conditions (such as harmonic interference and nonlinear loads), and the calibration consistency reaches more than 95%, avoiding calibration failures caused by differences in working conditions in traditional methods.

[0163] Furthermore, this embodiment's fully automated process reduces manual labor, lowering calibration costs per device by 70%, while also shortening downtime and increasing production line productivity by 30%. Accurate calibration significantly reduces power measurement errors, avoiding equipment malfunctions or safety incidents caused by improper calibration, and ensuring safe and stable power grid operation. The modular design enables flexible integration into existing power equipment and is compatible with different sensor types and communication protocols, driving intelligent upgrades in the industry.

[0164] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0165] The above embodiments are only preferred embodiments of the present invention and cannot be used to limit the scope of protection of the present invention. Any non-substantial changes and replacements made by technicians in this field on the basis of the present invention fall within the scope of protection required by the present invention.

Claims

1. A method for calibrating measurement accuracy and angle based on an automated calibration device, characterized in that: The following steps are involved: Use a standard calibration source as a reference to output standard current, voltage and phase angle. The calibration source includes a standard voltage source and a standard current source. The device uses a built-in integrated sensor array to obtain real-time information on current and voltage in power protection equipment, and uses the phase difference measurement method to calculate the phase difference, i.e., the phase angle, between the processed current and voltage signals. Receive sampled data from the integrated sensor array and use an adaptive iterative calibration algorithm to iteratively calculate the sampled data against a preset benchmark based on the preset accuracy requirements and iteration limit, gradually approaching the true current, voltage, and angle values; Receive the current, voltage, and angle values output by the calibration algorithm, as well as the preset calibration reference value, and use the machine learning model to train and learn the received data to establish a mapping relationship between the current, voltage, angle, and calibration reference value; According to the established mapping relationship, the actual collected data of the device is calibrated and the calibrated accurate value is output.

2. The method according to claim 1, wherein: The integrated sensor array includes at least one high-precision current transformer and one high-precision voltage transformer. These sensors are configured at the input or output end of the power protection device to simultaneously capture the waveform signals of current and voltage. The integrated sensor array adopts high-precision synchronous sampling technology to ensure that all sensors sample the current and voltage signals at the same time point or within a very short time window.

3. The method according to claim 1, wherein: Perform fast Fourier transform (FFT) or discrete Fourier transform (DFT) on the processed current signal and voltage signal to convert the time domain signal into a frequency domain signal; In the frequency domain signal, identify and extract the fundamental components of the current signal and voltage signal; Calculate the phase difference between the current signal and the voltage signal using the extracted fundamental component; In the phase difference calculation process, phase unwrapping technology is used to handle the situation where the phase angle crosses the boundary of -π to π, so as to ensure the continuity and accuracy of the phase difference.

4. The method according to claim 3, wherein: The phase difference calculation process uses the phase unwrapping technology to handle the situation where the phase angle crosses the boundary between -π and π. The specific implementation includes the following steps: Calculate the initial phase difference between adjacent sampling points or signal cycles using the inverse tangent function The formula is: Where S(n) is the complex signal of the nth sampling point, Im and Re represent the imaginary and real parts of the signal, respectively. Detecting the initial phase difference Whether a jump across the -π to π boundary occurs is determined by the following formula: If this condition is met, it is determined to be a phase jump.

5. The method according to claim 4, characterized in that: The detected phase jump is corrected by adjusting the phase difference by adding or subtracting integer multiples of 2π to make it change continuously. The correction formula is: By accumulating the corrected phase difference point by point, the unwrapped phase sequence is obtained, which is expressed as the following formula: The initial conditions are The unwrapped phase sequence is used to calculate the total phase difference between the current signal and the voltage signal, which is expressed as the following formula: Where N is the total number of sampling points in the signal period.

6. The method according to claim 1, characterized in that The initialization step of the adaptive iterative calibration algorithm includes: Initialize the parameters and set the initial calibration parameter vector P0 = [P0,1,P0,2,…,P0,m], where m is the calibration parameter dimension; Define a preset reference vector B = [B1, B2, ..., Bn], where n is the sampling data dimension, including current, voltage, and angle; Set the accuracy threshold ∈ and the maximum number of iterations kmax.

7. The method according to claim 6, characterized in that The iterative calculation process of the adaptive iterative calibration algorithm specifically includes: The raw sampled data Xraw = [x1, x2, …, xn] output by the integrated sensor array is filtered and normalized to obtain the preprocessed data Xk, which is expressed as the following formula: X k =f preprocess (X raw ) Calculate the error vector Ek of the kth iteration, expressed as the following formula: E k =X k -f model (P k ) The weight matrix Wk is dynamically adjusted according to the error to enhance the correction of high error terms, which can be expressed as the following formula: The calibration parameters are updated by weighted least squares method, which is expressed as the following formula: If any of the following conditions is met, the iteration is terminated, which is expressed as the following formula: ||E k ||<∈ or k≥k max The output converged calibration parameter P* and the corresponding true value estimate are expressed as the following formula: V true =f model (P * ) Among them, Vtrue contains the calibrated current, voltage, and angle values.

8. The method according to claim 1, characterized in that The machine learning model includes: Input layer: receives the normalized feature vector F = [current, voltage, angle] of current, voltage and angle, with dimension din = 3; Hidden layer: A multi-layer fully connected network (MLP) is used, with each layer containing h neurons and the activation function being ReLU. An attention mechanism module is introduced to dynamically weight input features and extract nonlinear mapping relationships, which can be expressed as the following formula: F′=Softmax(W attn ·F)⊙F Output layer: Outputs the calibration reference value y^, with dimension dout=1, which is used to correct measurement errors; Training process: Based on the weighted mean square error loss function and Adam optimizer, the mapping relationship is established through iterative training.

9. The method according to claim 8, characterized in that The training process specifically includes: The current, voltage, and angle values output by the calibration algorithm are normalized and expressed as the following formula: The weighted mean square error loss function is used to increase the penalty for high error samples, which is expressed as the following formula: Use the adaptive moment estimation (Adam) optimizer to dynamically adjust the learning rate, which is expressed as the following formula: After the training is completed, the calibrated reference value y^ is output to correct the measurement errors of current, voltage and angle.

10. The method according to any one of claims 1 to 9, characterized in that: The mapping relationship between current, voltage, angle and calibration reference value is: calibration reference value = w1 × current + w2 × voltage + w3 × angle + b, where w1, w2, and w3 are the weight coefficients of the model, corresponding to the degree of influence of current, voltage, and angle on the calibration reference value, respectively, and b is the bias term of the model.

Images