Multi-dimensional working condition coupling dynamic calibration method for electric carbon measuring instrument

Abstract

The application is specifically a multi-dimensional working condition coupling dynamic calibration method for electric carbon metering instruments, relates to the technical field of electric carbon metering and detection, and comprises: converting the prediction ability of an error model into a calibration action.In the application, multi-dimensional working condition coupling simulation breaks through the limitation of single-dimensional detection, covers industrial field core interference factors such as voltage fluctuation, harmonic interference, flue gas temperature and pressure change and electromagnetic interference, ensures the consistency improvement of detection data and actual use scene, and avoids the problem of laboratory qualification and on-site misalignment; meanwhile, an error characteristic is adaptively selected based on a model, and parameters are optimized through a weighted least square method, a particle swarm optimization algorithm and the like, thereby providing accurate and reliable error basis for subsequent dynamic calibration, effectively guaranteeing the accuracy of carbon emission accounting, and meeting the strict requirement of carbon trading on metering precision.

Description

Technical Field

[0001] This invention relates to the field of electrocarbon metering and testing technology, and in particular to a multi-dimensional operating condition coupled dynamic calibration method for electrocarbon metering instruments. Background Technology

[0002] Current calibration technologies for electric carbon metering instruments generally suffer from static limitations, making it difficult to adapt to dynamic changes in on-site working conditions. This is one of the core issues restricting measurement accuracy.

[0003] Existing calibration methods mostly adopt offline laboratory mode, generating calibration parameters only under fixed operating conditions such as rated voltage and standard gas concentration, without establishing a linkage mechanism for real-time operating conditions, error prediction, and parameter updates.

[0004] For example, traditional carbon meter calibration requires disassembling the device and taking it to the laboratory to set parameters based on a single rated operating condition. However, after the device is put into the industrial field, changes in operating conditions such as light load, high harmonics, and flue gas temperature and pressure fluctuations will cause new measurement errors. Static calibration parameters cannot compensate for this deviation in real time, resulting in qualified calibration but inaccurate on-site, which seriously affects the accuracy of carbon emission accounting. This lack of dynamic calibration capability has become a key issue that existing technologies urgently need to overcome.

[0005] Meanwhile, existing detection technologies do not adequately cover multi-dimensional operating condition coupling interference, resulting in error models being unable to accurately capture error patterns in real-world scenarios. This is another core technical pain point.

[0006] Existing methods mostly focus on the detection of single-dimensional electrical parameters (voltage, current) or carbon measurement parameters (gas concentration, flow rate), generally ignoring the coupling effect of environmental parameters (such as electromagnetic interference, temperature and humidity) with electrical and carbon measurement parameters.

[0007] Therefore, a multi-dimensional operating condition coupled dynamic calibration method for electric carbon metering instruments is needed to address the problems mentioned above. Summary of the Invention

[0008] The purpose of this invention is to propose a multi-dimensional operating condition coupled dynamic calibration method for electric carbon metering instruments in order to solve the above-mentioned problems.

[0009] To achieve the above objectives, the present invention adopts the following technical solution:

[0010] A multi-dimensional operating condition coupled dynamic calibration method for electric carbon metering instruments includes:

[0011] Build a realistic working condition simulation environment;

[0012] The performance of the equipment is tested under preset working conditions, and the working condition input and measurement error datasets are collected simultaneously to provide original samples for subsequent error modeling.

[0013] To uncover error patterns in the detection data and construct mathematical models;

[0014] Translate the predictive power of the error model into calibration actions;

[0015] The input data consists of real-time operating condition input variables at typical operating points in the actual application of the tested equipment, the error model that has been constructed and met the standards in the early stage, and the original calibration curve of the equipment.

[0016] The process includes substituting real-time operating condition input variables into the error model to obtain real-time predicted error, solving new calibration parameters based on the original calibration curve using a reverse optimization algorithm, encapsulating the parameters according to the instrument communication protocol and injecting them into the instrument after encryption and authentication, and finally reading the parameters to verify successful injection.

[0017] The final output includes real-time prediction error, final calibration parameter set, and calibration execution and verification data.

[0018] Preferably, the specific content of building the realistic working condition simulation environment includes:

[0019] Connect the tested carbon metering instrument, standard calibration device, and environmental simulation equipment to enable real-time synchronous acquisition of measurement data and environmental parameters from the standard device and the tested instrument.

[0020] Based on the actual application scenarios of electric carbon metering instruments, a multi-dimensional combination of operating conditions is set, including:

[0021] Electrical operating conditions: voltage, current, power factor, harmonic content;

[0022] Carbon metering conditions: standard gas concentration, gas flow rate, flue gas temperature, and flue gas pressure;

[0023] Environmental conditions; ambient temperature, relative humidity, and electromagnetic interference intensity;

[0024] Start the data acquisition software, set the sampling frequency, and determine the acquisition objects as: the true value data output by the standard device, the measurement data output by the instrument under test, and the real-time environmental parameters of the environmental simulation equipment;

[0025] The acquisition system is time-synchronized and calibrated, and pre-acquisition tests are performed. At the same time, the initial preprocessing of the raw data is completed.

[0026] Preferably, the step of testing the instrument performance under preset operating conditions and simultaneously collecting the operating condition input and measurement error datasets to provide original samples for subsequent error modeling specifically includes:

[0027] According to the preset standards, the core performance of the tested equipment is tested under different electrical conditions, including accuracy testing, stability testing, and harmonic influence testing.

[0028] The input variables of each set of electrical operating conditions are recorded synchronously. True value of standard device and the measured values ​​of the tested instruments Calculate electrical error Input variables In For voltage, For current, For power Harmonic content rate;

[0029] Based on preset standards, the performance of the equipment is tested under different carbon-related operating conditions, including concentration measurement accuracy testing, flow rate influence testing, and flue gas temperature and pressure compensation performance testing.

[0030] Synchronously record the input variables for each set of carbon operating conditions. True value of standard device and the measured values ​​of the tested instruments Calculate carbon measurement error ;

[0031] Input variables In For standard gas concentration, For standard gas flow rate, For flue gas temperature, This refers to the flue gas pressure.

[0032] Preferred options also include:

[0033] Under electrical pre-carbonization related conditions, adjust environmental parameters and detect the impact of environmental interference on the overall performance of the equipment, specifically including the effects of temperature and humidity, and electromagnetic interference.

[0034] Record environment input variables and with Merge into a complete input vector Error vector ;

[0035] Environment input variables In For ambient temperature, For environmental humidity, Electromagnetic interference intensity;

[0036] Input variables for electrical operating conditions Input variables for carbon operating conditions Input variables for the environment;

[0037] For the collected input vector and error vector The data is structured and divided into training and testing sets.

[0038] Preferably, the content of mining the error patterns in the detection data and constructing the mathematical model specifically includes:

[0039] Based on the error characteristics of the carbon metering instrument, select the appropriate model type, including the following cases:

[0040] A1: If the coupling relationship between the error and the input variables is clear, a multivariate multinomial regression model should be used, with the following structure:

[0041]

[0042] in, The coefficients to be determined are: This is the nth input variable;

[0043] A2: If the error coupling relationship is complex, a multilayer perceptron neural network should be selected, with the structure set as an input layer, a hidden layer, and an output layer.

[0044] A3: If the sample size of the detection data is small, support vector machines can be used to map to a high-dimensional space through kernel functions to achieve high-precision error regression under small sample size.

[0045] Preferably, this also includes model training and parameter optimization:

[0046] For the multinomial regression model in A1: Minimize the model prediction error using the least squares method. With respect to actual error Sum of squares, solve for coefficients ;

[0047] For the multilayer perceptron neural network in A2: the backpropagation algorithm is used, with the mean squared error as the loss function, and the network weights and biases are optimized by gradient descent, iterating until the loss function converges;

[0048] For the vector machine model in A3: Error regression is achieved by minimizing the sum of the penalty term for samples whose errors exceed the allowable threshold and the model complexity penalty term by adjusting the penalty parameter and kernel function parameter.

[0049] Preferred options also include:

[0050] Randomly select 30% of the samples from the original dataset as the test set; perform the same preprocessing on the test set as on the training set;

[0051] The correlation coefficient, mean absolute error, and maximum error are calculated respectively, and the model is judged to meet the standard based on the three calculated values. If the model is judged to be unqualified, corresponding processing is carried out.

[0052] The structure and parameters of the compliant model are serialized and assigned a unique identifier, which is then stored in the model database of the detection system.

[0053] Preferably, translating the predictive power of the error model into calibration actions specifically includes:

[0054] Select 1-3 typical operating conditions of the equipment under test in actual application, and collect the input variables of the current operating conditions in real time. ;

[0055] Input variables Input a fixed error model, output real-time prediction error ;

[0056] Retrieve the original calibration curve of the instrument under test for calibration until the difference between the corrected instrument power measurement value and the instrument concentration measurement value and the true value is less than the preset maximum allowable standard, so as to eliminate real-time prediction error. The impact;

[0057] The new calibration parameters, including power calibration and concentration calibration, are solved using a reverse optimization algorithm; the result is a set of calibration coefficients. .

[0058] Preferred options also include:

[0059] According to the communication protocol of the device under test, the calibration coefficients are encapsulated into executable instruction frames;

[0060] The detection system establishes an encrypted communication link with the device under test and performs identity authentication.

[0061] The instruction frame is sent to the calibration parameter storage area of ​​the instrument, and the instrument MCU updates the internal calibration parameters after receiving it;

[0062] Send a read parameter command to read the updated calibration parameters inside the instrument, compare them with the calculated target value, and confirm successful injection.

[0063] Preferred options also include:

[0064] The calibration effect is ensured through retesting, evaluation, and iterative processing, while the error model is optimized to achieve continuous evolution of the detection system.

[0065] A computer-readable storage medium storing a computer program, which, when executed by a processor, implements all steps of a multi-dimensional operating condition coupled dynamic calibration method for an electric carbon meter.

[0066] A multi-dimensional operating condition coupled dynamic calibration device for electric carbon metering instruments, comprising:

[0067] The working environment setup module is used to build a realistic working condition simulation environment;

[0068] The data acquisition module is used to detect the performance of the electric carbon metering instrument under preset working conditions, and simultaneously collect the working condition input and measurement error datasets to provide raw samples for subsequent error modeling.

[0069] The error model building module is used to mine error patterns in detection data and build mathematical models.

[0070] The dynamic calibration execution module is used to translate the predictive capabilities of the error model into calibration actions, dynamically generate calibration parameters based on real-time operating conditions, and automatically inject them into the electric carbon metering instrument.

[0071] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0072] 1. This invention overcomes the limitations of single-dimensional detection by using multi-dimensional coupled simulation of operating conditions. By covering core interference factors in industrial settings such as voltage fluctuations, harmonic interference, flue gas temperature and pressure changes, and electromagnetic interference, it ensures improved consistency between test data and actual usage scenarios, avoiding the problem of laboratory compliance but inaccurate field results. Simultaneously, based on error characteristics, it adaptively selects a model and optimizes parameters using algorithms such as weighted least squares and particle swarm optimization, providing accurate and reliable error data for subsequent dynamic calibration. This effectively ensures the accuracy of carbon emission accounting and meets the stringent requirements for measurement accuracy in carbon trading.

[0073] 2. This invention collects typical on-site working condition data in real time, calls the bound error model to output the predicted error, and then uses a reverse optimization algorithm to solve the calibration parameters. The parameters are then automatically injected into the instrument through encrypted communication, reducing the measurement error of the instrument after calibration. Even under extreme working conditions, the error can still be controlled within the allowable range. Relying on the retest verification and sample iteration mechanism, the error model is optimized every 50-100 samples or every quarter. This not only improves the error prediction accuracy for different batches and aging instruments, but also improves the testing efficiency of subsequent similar instruments and reduces the maintenance cost of the testing system. It achieves a breakthrough from one-time testing to continuous evolution, and can adapt to the complex working condition changes in industrial plants, carbon monitoring stations and other scenarios for a long time, ensuring the long-term stable operation of electric carbon metering instruments. Attached Figure Description

[0074] Further details, features, and advantages of this application are disclosed in the following description of exemplary embodiments in conjunction with the accompanying drawings, in which:

[0075] Figure 1 This is a flowchart of the method of the present invention;

[0076] Figure 2 This is a schematic diagram of the device module of the present invention. Detailed Implementation

[0077] Several embodiments of this application will now be described in more detail with reference to the accompanying drawings to enable those skilled in the art to implement this application. This application may be embodied in many different forms and for various purposes and should not be limited to the embodiments set forth herein. These embodiments are provided to make this application thorough and complete, and to fully convey the scope of this application to those skilled in the art. The embodiments described do not limit this application.

[0078] Unless otherwise defined, all terms used herein (including technical and scientific terms) shall have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. It will be further understood that terms such as those defined in commonly used dictionaries shall be interpreted as having a meaning consistent with their meaning in the relevant field and / or the context of this specification, and shall not be interpreted in an idealized or overly formal sense unless expressly defined herein.

[0079] Example:

[0080] Its specific implementation method is combined with the appendix Figure 1 and attached Figure 2 Please provide a detailed explanation.

[0081] Appendix Figure 1 The flowchart of the multi-dimensional operating condition coupled dynamic calibration method for electric carbon metering instruments provided in the embodiments of the present invention illustrates the complete steps from building a real operating condition simulation environment to optimizing the error model.

[0082] In this embodiment, it includes:

[0083] Build a realistic working condition simulation environment and complete the preliminary preparations for data collection to ensure that the test data can cover the actual use scenarios of the equipment;

[0084] Specifically, it includes:

[0085] Equipment and Link Setup: Connect the tested carbon metering instrument, standard calibration devices (such as standard energy meters, standard gas analyzers, and standard flow generators), and environmental simulation equipment (such as temperature and humidity controllers and electromagnetic interference generators); configure data interaction interfaces (e.g., RS-485, infrared, Wi-Fi) according to the communication protocol of the tested instrument to enable real-time synchronous acquisition of measurement data and environmental parameters between the standard devices and the tested instrument.

[0086] Based on the actual application scenarios of electric carbon metering instruments (such as industrial plants, park microgrids, and carbon monitoring stations), multi-dimensional operating condition combinations are set, including:

[0087] Electrical operating conditions: voltage, current, power factor (0.5 inductive to 1.0 capacitive), harmonic content (0 to 5%, simulating grid harmonic interference);

[0088] Carbon metering conditions: standard gas concentration, gas flow rate, flue gas temperature, and flue gas pressure;

[0089] Environmental conditions; ambient temperature, relative humidity, and electromagnetic interference intensity;

[0090] Start the data acquisition software, set the sampling frequency (which can be adjusted according to the response speed of the instrument), and determine the acquisition objects as: the true value data output by the standard device, the measurement data output by the instrument under test, and the real-time environmental parameters of the environmental simulation equipment;

[0091] Perform time synchronization calibration on the acquisition system (the error must be ≤1ms), and perform pre-acquisition test to verify that there is no packet loss or error code in the data. At the same time, complete the preliminary preprocessing of the raw data (such as filtering and noise reduction, outlier removal, and unit normalization, such as normalizing the voltage to the 0~1 range).

[0092] The performance of the equipment is comprehensively tested under preset working conditions, and the working condition input and measurement error datasets are collected simultaneously to provide original samples for subsequent error modeling.

[0093] Specifically, it includes:

[0094] According to pre-set standards such as GB / T17215.211 (Electricity Meter Standard) and JJG307 (Verification Procedure for AC Energy Meters), the core performance of the tested equipment is tested under different electrical operating conditions, including accuracy testing, stability testing, and harmonic influence testing. Accuracy testing: Under rated voltage, rated current, and power factor 1.0 (inductive) conditions, the basic error of the equipment is tested; under light load (0.1... ), overload (1.5) ), under low power factor (0.5 inductive) conditions, the load characteristic error of the detection instrument;

[0095] Stability testing: The device operates continuously for 24 hours under rated conditions, with measurement data collected once per hour. The maximum deviation value is calculated to assess the long-term operational stability of the device. Harmonic influence testing: The device's additional harmonic error is tested under voltage harmonic content conditions of 3% (3rd harmonic) and 5% (5th harmonic).

[0096] The input variables of each set of electrical operating conditions are recorded synchronously. True value of standard device and the measured values ​​of the tested instruments Calculate electrical error ;

[0097] Input variables In Voltage (using voltage sensors (such as high-precision voltage transformers and voltage probes) to collect the voltage signal of the electrical circuit where the tested device is located, and then using a data acquisition system (including an analog-to-digital converter module) to convert the analog signal into a digital value to obtain the voltage value). The current is collected using a current sensor (such as a current transformer or Hall effect current sensor), and the current value is obtained after processing by a data acquisition system. Power (can be directly measured by a power analyzer (which can simultaneously acquire voltage and current and calculate power); or it can be calculated from the acquired voltage, current, and power factor). The harmonic content rate is calculated by first performing a Fourier transform on the voltage / current signal using a harmonic analyzer (or a data acquisition device with harmonic analysis capabilities) to decompose the amplitudes of the fundamental wave and each harmonic; then, it is calculated using the harmonic content rate formula.

[0098] Measurement results from the standard calibration device:

[0099] A standard device is a reference device that has undergone metrological verification and has a much higher accuracy than the device under test (such as a standard energy meter or a standard power source). Under the same electrical operating conditions as the device under test, the standard device synchronously measures the target parameters (such as power and energy), and its output measured value is the true value, which serves as the benchmark value for judging the error of the device under test.

[0100] The output result of the tested carbon meter itself:

[0101] The measured values ​​of the current electrical parameters can be read directly from the output interface (such as a communication interface or a local display screen) of the instrument under test; or the signals output by the instrument (such as pulse signals or digital messages) can be collected through a data acquisition system and converted into numerical values ​​to obtain the measurement results of the instrument under test.

[0102] Electrical error Measured values ​​of the instrument under inspection Subtract the true value of the standard device The difference is the systematic error of the tested instrument under that operating condition, and is used to quantify its measurement deviation.

[0103] Based on preset standards such as HJ1242 (Infrared Absorption Method for Determination of Carbon Dioxide and Oxygen in Exhaust Gas from Stationary Sources) and JJG635 (Verification Procedure for Gas Volumetric Flow Meters), the performance of the instruments is tested under different carbon-related operating conditions, including concentration measurement accuracy testing, flow effect testing, and flue gas temperature and pressure compensation performance testing.

[0104] Concentration measurement accuracy: The concentration measurement error of the instrument is detected by introducing standard carbon dioxide gas at different concentrations (e.g., 400ppm, 1000ppm, 5000ppm).

[0105] Flow rate effect detection: At a standard concentration (1000 ppm), adjust the gas flow rate (e.g., 0.5 L / min, 2 L / min, 5 L / min) and detect the additional error of flow rate change on concentration measurement;

[0106] Flue gas temperature and pressure compensation performance: Under standard concentration and rated flow, adjust the flue gas temperature (e.g., 50℃, 100℃, 150℃) and pressure (e.g., -5kPa, 0kPa, 5kPa) to detect the impact of temperature and pressure changes on the measurement results.

[0107] Synchronously record the input variables for each set of carbon operating conditions. True value of standard device and the measured values ​​of the tested instruments Calculate carbon measurement error ;

[0108] Input variables In The standard gas concentration (the known accurate concentration of a standard carbonaceous gas used for detection / calibration; provided by standard gas cylinders (factory-calibrated and labeled with precise concentration), or prepared according to a set ratio using a high-precision device such as a dynamic gas mixing instrument). Standard gas flow rate (the rate at which standard gas flows in the detection system; the flow rate value is measured in real time and output by a standard gas flow meter (such as a mass flow meter, volumetric flow meter, etc., which has passed metrological verification). The flue gas temperature (the temperature of the flue gas being detected or the simulated flue gas environment; using a calibrated temperature sensor (such as a thermocouple or RTD) inserted into the flue gas channel / simulated environment to collect and convert the data into a temperature value). Flue gas pressure (the pressure of the flue gas being detected or the simulated flue gas environment; a calibrated pressure sensor (such as a differential pressure transmitter or absolute pressure transmitter) is installed in the flue gas circuit to measure and output the pressure value).

[0109] The benchmark accurate value is obtained by measuring the target parameters (such as gas concentration, total amount corresponding to carbon flow rate, etc.) under the same carbon operating conditions as the instrument under test using a high-precision standard carbon metering device (such as a calibrated standard gas analyzer, standard carbon flow meter, etc.). The standard device is connected to the carbon operating condition detection environment synchronously with the instrument under test and directly outputs its measured true value (because the accuracy of the standard device is much higher than that of the instrument under test, its result is used as the benchmark for error judgment).

[0110] The measurement results of the tested electric carbon metering instrument on the current carbon operating parameters (such as gas concentration, carbon flow rate, etc.); directly read from the data interface of the tested instrument (such as communication interface, local display screen, internal register), or collect the signal (analog quantity / digital message) output by the instrument through the data acquisition system and convert it into a value.

[0111] Under electrical pre-carbonization related operating conditions, environmental parameters were adjusted to detect the impact of environmental interference on the overall performance of the equipment, specifically including the effects of temperature and humidity, and electromagnetic interference. Temperature and humidity effects: the ambient temperature was increased from -10℃ to 40℃ (staying for 1 hour at every 5℃) and the humidity was increased from 30% to 90% (staying for 1 hour at every 10%RH), and the measurement errors at each node were collected. Electromagnetic interference effects: electromagnetic interference of 0~10V / m (frequency 30MHz~1GHz) was applied, and the anti-interference error of the equipment was detected.

[0112] Record environment input variables and with Merge into a complete input vector Error vector ; Environment input variables In The ambient temperature is the air temperature of the environment in which the tested carbon metering instrument is located; a calibrated temperature sensor (such as a platinum resistance thermometer, thermocouple, etc.) is placed in the environment near the tested instrument to collect temperature data in real time and convert it into a numerical value. The ambient humidity is the relative humidity of the air in the environment where the tested instrument is located; relative humidity data is collected by placing humidity sensors (such as capacitive humidity sensors, humidity-sensitive resistors, etc.) in the environment and then converting the data to obtain the humidity value. Electromagnetic interference intensity (the strength of electromagnetic interference in the environment (reflecting the magnitude of interference such as electric and magnetic fields); using an electromagnetic interference tester (such as a spectrum analyzer, field strength meter, etc.), the strength of electromagnetic signals is measured around the device under test, and is usually output in field strength units or power units).

[0113] Input variables for electrical operating conditions Input variables for carbon operating conditions Input variables for the environment;

[0114] The input parameters of the three operating conditions—electrical, carbon, and environmental—are concatenated in sequence to form an input vector that covers all disturbances in the scenario. This vector is then used as the input layer of the subsequent multivariate coupled error model to characterize the comprehensive impact of complex operating conditions on measurement errors.

[0115] For the collected input vector and error vector The dataset is structured and divided into a training set (70% of the total data, used for model training) and a test set (30% of the data, used for model validation) to ensure that the dataset covers all preset working conditions and has no missing or abnormal data.

[0116] By mining the error patterns in the detection data, we can construct a mathematical model that can accurately predict errors and provide a basis for dynamic calibration.

[0117] Specifically, it includes:

[0118] Based on the error characteristics of the electric carbon metering instrument (e.g., nonlinearity, multivariable coupling), select the appropriate model type, including the following cases:

[0119] A1: If the coupling relationship between the error and the input variable is clear (e.g., the effect of temperature on accuracy follows a quadratic curve), a multivariate multinomial regression model is selected, with the following structure:

[0120]

[0121] in, The coefficients to be determined are: This is the nth input variable;

[0122] A2: If the error coupling relationship is complex (such as the nonlinear effect of electromagnetic interference and harmonic superposition), a multilayer perceptron (MLP) neural network is selected, with the structure set as input layer (number of neurons = dimension of input variables), hidden layer (2~3 layers, 16~32 neurons per layer), and output layer (number of neurons = number of error types, such as 2: electrical error, carbon measurement error).

[0123] A3: If the sample size of the detection data is small (such as the lack of historical data for new instruments), support vector machines (SVR) can be selected and mapped to a high-dimensional space through kernel functions (such as RBF kernel) to achieve high-precision error regression under small sample size.

[0124] It also includes model training and parameter optimization:

[0125] The model is trained using the corresponding algorithm:

[0126] For the multinomial regression model in A1: Minimize the model prediction error using the least squares method. With respect to actual error Sum of squares: Solve for the coefficients ;in Let i be the weight of the i-th sample. (The larger the error, the higher the weight); the coefficients are solved directly through matrix operations: ,in For the input variable matrix, This is a weighted diagonal matrix. This represents the actual error vector;

[0127] Select the optimal number of items using the Akaike Information Criterion (AIC): ,in For the number of terms, For the likelihood function value, choose the model structure with the smallest AIC (in order to balance complexity and accuracy).

[0128] For the multilayer perceptron neural network in A2: using the backpropagation (BP) algorithm, with mean squared error:

[0129] The loss function is used to optimize the network weights and biases through gradient descent, iterating until the loss function converges (e.g., ...). <0.001);

[0130] The initial learning rate of the loss function is set to 0.001, and it decays to 0.9 every 50 iterations (to avoid oscillations).

[0131] Iteration termination condition: Stop when the MSE of the validation set does not decrease (change < 1e-5) for 10 consecutive iterations, or when the total number of iterations reaches 1000.

[0132] Optimize hyperparameters using grid search and 5-fold cross-validation:

[0133] Number of hidden layer neurons: Search in [16, 32, 64];

[0134] Dropout rate: Search in [0.1, 0.2, 0.3];

[0135] Choose the hyperparameter combination that minimizes the MSE of cross-validation.

[0136] For the vector machine model in A3: Error regression is achieved by minimizing the sum of the penalty term for samples whose error exceeds the allowable threshold and the model complexity penalty term by adjusting the penalty parameter (controlling fitting accuracy and generalization ability) and the kernel function parameter (controlling kernel function complexity). Specifically, this includes:

[0137] The structural risk is minimized by solving a convex quadratic programming problem.

[0138] The objective function is: ,in For the weight vector, As slack variables, This is the penalty coefficient;

[0139] Constraints: ,in For kernel function mapping, For bias;

[0140] Parameter tuning: Optimize key parameters using the Particle Swarm Optimization (PSO) algorithm.

[0141] Penalty coefficient Search within the interval [1, 100] The larger the value, the more the model tends to fit all samples, making it prone to overfitting.

[0142] Kernel function parameters Search within the range [0.01, 10]. The larger the value, the stronger the locality of the kernel function, and the easier it is to overfit.

[0143] Using the test set MAE as the fitness function, we search for the optimal combination of parameters.

[0144] Also includes:

[0145] After the model is trained, its generalization ability needs to be verified using an independent test set. Once it meets the standard, it is solidified as the unique error prediction model for the tested instrument.

[0146] Randomly select 30% of the samples from the original dataset as the test set (which must cover all working conditions and avoid duplication with the training set samples); perform the same preprocessing on the test set as on the training set (such as normalization and outlier handling) to ensure consistent data distribution.

[0147] The correlation coefficient, mean absolute error, and maximum error are calculated respectively, and the model is judged to meet the standard based on the three calculated values. If the model is judged to be unqualified, corresponding processing is carried out.

[0148] Correlation coefficient ,in, The mean of the prediction error. This is the average of the actual errors; it needs to meet the following requirements. A value greater than 0.95 indicates that the prediction error is highly linearly correlated with the actual error, and the model captures the main patterns of the error.

[0149] Mean Absolute Error ; needs to meet ≤10% of the permissible error of the tested instrument; for example, if the permissible error of the instrument is ±2%, then ≤0.2%, ensuring that the average prediction bias of the model is within an acceptable range.

[0150] Maximum error ; needs to meet ≤20% of the allowable error of the tested instrument (to avoid excessive prediction deviation under extreme working conditions).

[0151] like If the value is less than 0.95, it is considered as not meeting the standard, indicating that the model has not captured the key error patterns. It is necessary to supplement the collection of working condition samples with large error fluctuations (such as high electromagnetic interference and high harmonic scenarios), or replace it with a more complex model (such as upgrading the polynomial regression to MLP).

[0152] like and If the value exceeds the preset threshold, it is considered unqualified. Overfitting is suppressed by regularization enhancement (such as increasing the C value of SVR and the Dropout rate of MLP), or outliers in the test set are manually removed (which need to be verified as data acquisition errors).

[0153] The structure (such as the number of polynomial terms and the structure of neural network layers) and parameters (coefficients, weights, and biases) of the compliant model are serialized; and a unique identifier (bound to the SN code of the tested instrument) is assigned and stored in the model database of the testing system for subsequent dynamic calibration.

[0154] The predictive power of the error model is transformed into calibration actions, and calibration parameters are dynamically generated based on real-time operating conditions and automatically injected into the equipment.

[0155] Specifically, it includes:

[0156] Select 1-3 typical operating conditions of the equipment under test in actual application, and collect the input variables of the current operating conditions in real time. ;

[0157] Input variables Input a fixed error model, output real-time prediction error (e.g., electrical errors) Carbon concentration error This value reflects the systematic error of the instrument under the current operating conditions.

[0158] Retrieve the original calibration curve of the instrument under test (stored in the testing system or the instrument's internal register). Electrical metrology and carbon metrology typically follow a linear calibration model.

[0159] Electrical metering: ,in, This is the measured power value of the appliance. This is the true value of the power. For power calibration slope, The intercept;

[0160] Carbon metering: ,in, This refers to the concentration measurement value of the instrument. For the true concentration, For concentration calibration slope, The intercept;

[0161] Perform calibration until the corrected appliance power measurement value is obtained. Instrument concentration measurement value The difference between the real value and the true value is less than the preset maximum allowable standard to eliminate real-time prediction errors. The impact;

[0162] The new calibration parameters are solved using a reverse optimization algorithm, including power calibration and concentration calibration.

[0163] Power calibration:

[0164] Concentration calibration:

[0165] After calculation, a set of calibration coefficients is generated: .

[0166] According to the communication protocol of the device under test, the calibration coefficients are encapsulated into executable instruction frames;

[0167] The detection system establishes an encrypted communication link with the tested device and performs identity authentication (such as device SN code matching and key verification) to prevent misoperation or malicious tampering.

[0168] The instruction frame is sent to the instrument's calibration parameter storage area. After receiving the instruction frame, the instrument's MCU updates its internal calibration parameters, that is, replaces the original ones. ;

[0169] Send a read parameter command to read the updated calibration parameters inside the instrument, compare them with the calculated target value, and confirm successful injection.

[0170] For traditional electric carbon metering instruments without an automatic communication interface, the detection system generates a "Manual Calibration Guide Manual," which includes:

[0171] Target calibration parameters: for example ;

[0172] Operating steps: Open the instrument calibration panel and enter the concentration calibration mode; rotate the slope adjustment potentiometer and observe the display until the displayed value is 1025.03 ppm when 1000 ppm standard gas is introduced; save the parameters and exit the calibration mode;

[0173] Verification method: Introduce 500ppm of standard gas, and the required measurement value is 512.53ppm (calculation process: 500×1.025+0.03), with an error of ≤±1ppm.

[0174] The calibration effect is ensured through retesting, evaluation, and iterative processing, while the error model is optimized to achieve continuous evolution of the detection system. Specifically, this includes:

[0175] At the real-time operating point in step four and other key operating points preset in step two (such as light load, high humidity, and electromagnetic interference conditions), performance testing is performed again.

[0176] Repeat the detection process in step two and collect the calibrated measurement values. Calculate the new error ;

[0177] The focus is on verifying operating conditions with large errors before calibration (such as low power factor and high flue gas temperature) and evaluating the compensation effect of calibration on key errors.

[0178] Calibration effect judgment: If all operating points All meet If the calibration is within the allowable error range of the tested instrument, the calibration is deemed qualified. If individual operating points do not meet the standard, return to step four, fine-tune the calibration coefficient, and re-inject and retest. If multiple fine-tunings still fail to meet the standard, a hardware abnormality alarm will be triggered, indicating that the instrument sensor (such as current transformer, infrared gas detection module) or circuit needs to be repaired.

[0179] Overall performance rating: Based on all detection data, from accuracy ( The overall performance rating of the instrument is based on four dimensions: mean, stability (maximum deviation over 24 hours), environmental adaptability (error changes under temperature, humidity / electromagnetic interference), and calibration responsiveness (error reduction after calibration). (e.g., Grade A: meets all operating conditions and...) ≤50% permissible error; Grade B: Core operating conditions meet standards, secondary operating conditions meet standards. ≤ 80% permissible error.

[0180] The operating condition data from this test Error before calibration Calibration parameters Calibration error Performance ratings are compiled, merged, and archived into the appliance lifecycle database.

[0181] The results of this test As new samples, they are added to the training database of the error model; every 50 to 100 test samples of the same type of instrument are accumulated, or the model is retrained once every quarter to optimize the model parameters (such as updating polynomial coefficients and neural network weights), improve the model's error prediction accuracy for instruments of different batches and different aging levels, and realize the continuous learning capability of the detection system.

[0182] The present invention also provides, for example Figure 2 The device shown is a multi-dimensional operating condition coupled dynamic calibration device for electric carbon metering instruments, comprising:

[0183] The working environment setup module is used to build a realistic working condition simulation environment;

[0184] The data acquisition module is used to detect the performance of the electric carbon metering instrument under preset working conditions, and simultaneously collect the working condition input and measurement error datasets to provide raw samples for subsequent error modeling.

[0185] The error model building module is used to mine error patterns in detection data and build mathematical models.

[0186] The dynamic calibration execution module is used to translate the predictive capabilities of the error model into calibration actions, dynamically generate calibration parameters based on real-time operating conditions, and automatically inject them into the electric carbon metering instrument.

[0187] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the multi-dimensional operating condition coupling dynamic calibration method for electric carbon metering instruments as described above.

[0188] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0189] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

[0190] It should be noted that, in this document, the use of relational terms such as "first" and "second" is merely for distinguishing one entity or operation from another, and does not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0191] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0192] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0193] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0194] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0195] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0196] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0197] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A multi-dimensional working condition coupling dynamic calibration method for an electric carbon metering device, characterized in that, The method comprises the following steps: Building a real working condition simulation environment; Detecting the performance of the instrument under the preset working condition, and synchronously collecting the working condition input and measurement error data set to provide original samples for subsequent error modeling; Mining the error rules in the detection data and constructing a mathematical model; Specifically comprising: According to the error characteristics of the electric carbon measuring instrument, select the appropriate model type, including the following cases: A1: If the coupling relationship between error and input variable is clear, select a multivariate polynomial regression model, the structure is: wherein, is the coefficient to be found, is the nth input variable; V is the voltage, I is the current, T is the ambient temperature, C is the standard gas concentration; A2: If the error coupling relationship is complex, select a multilayer perception neural network, the structure is set as input layer, hidden layer and output layer; A3: If the sample size of the detection data is small, select a support vector machine, which is mapped to a high-dimensional space through a kernel function to realize high-precision error regression under small samples; It also includes model training and parameter optimization: For the polynomial regression model in A1: use least squares to minimize the sum of squared prediction errors with the actual errors squared, to solve for the coefficients ; For the multilayer perception neural network in A2: use the back propagation algorithm, take the mean square error as the loss function, and optimize the network weight and bias through gradient descent until the loss function converges; For the vector machine model in A3: adjust the penalty parameter and the kernel function parameter to minimize the sum of the sample penalty term and the model complexity penalty term when the error exceeds the allowed threshold, and realize error regression; The predictive ability of the error model is converted into calibration action; Among them, the input data is the real-time working condition input variable of the instrument to be detected in the typical working condition point of the actual application, the error model constructed in the early stage and meeting the standard, and the original calibration curve of the instrument; The processing process includes substituting the real-time working condition input variable into the error model to obtain the real-time predicted error, solving the new calibration parameters based on the original calibration curve through the reverse optimization algorithm, packaging the parameters according to the instrument communication protocol and injecting the parameters into the instrument after encryption authentication, and finally reading the parameters to verify the success of the injection; The final output is the real-time predicted error, the final calibration parameter set, and the calibration execution and verification data.

2. The multi-dimension working condition coupled dynamic calibration method of electric carbon metrological instruments according to claim 1, characterized in that, The content of building a real working condition simulation environment specifically includes: Connect the electric carbon measuring instrument to be detected, the standard calibration device and the environment simulation equipment, so that the measurement data of the standard device and the instrument to be detected and the environmental parameters are collected in real time; According to the actual application scene of the electric carbon measuring instrument, set up multi-dimensional working condition combination, including: Electrical working condition: voltage, current, power factor, harmonic content rate; Carbon measurement working condition: standard gas concentration, gas flow, flue gas temperature, flue gas pressure; Environmental working condition: environmental temperature, relative humidity, electromagnetic interference intensity; Start the data collection software, set the sampling frequency, and determine the collection object as: the true value data output by the standard device, the measurement data output by the instrument to be detected, and the real-time environmental parameters of the environment simulation equipment; Time synchronization calibration is performed on the collection system, pre-collection test is executed, and preliminary preprocessing of the original data is completed.

3. The multi-dimension working condition coupled dynamic calibration method of electric carbon metrological instruments according to claim 1, characterized in that, The content of detecting the performance of the instrument under the preset working condition and synchronously collecting the working condition input and measurement error data set to provide original samples for subsequent error modeling specifically includes: According to the preset standard, detect the core performance of the instrument to be detected under different electrical working conditions, including precision detection, stability detection and harmonic influence detection; Synchronously recording input variables of each set of electrical operating conditions , standard device true values , and measured values of the inspected appliance , calculating electrical errors ; Input variable in is the voltage, is the current, is the power is the harmonic content; According to the preset standard, detect the performance of the instrument under different carbon related working conditions, including concentration measurement precision detection, flow influence detection, flue gas temperature and pressure compensation performance detection; Synchronously recording input variables for each set of carbon conditions , standard device true values , and measured values of the inspected device , calculating carbon metering errors ； Input variables In For standard gas concentration, For standard gas flow rate, For flue gas temperature, This refers to the flue gas pressure.

4. The multi-dimension working condition coupled dynamic calibration method of the electric carbon weighing instrument according to claim 3, characterized in that, Also including: In the pre-carbon related working condition of the electrical working condition, the environmental parameters are adjusted, and the influence of environmental interference on the overall performance of the appliance is detected, specifically including the influence of temperature and humidity and electromagnetic interference; record environmental input variables , and combine them into a complete input vector , , error vector ;​ environmental input variables in the environment is the ambient temperature, is the ambient humidity, is the electromagnetic interference strength; for electrical operating conditions, for carbon operating conditions, for environmental inputs; The input vector collected and the error vector are structured and divided into training and test sets.

5. The multi-dimension working condition coupled dynamic calibration method of electric carbon metrological instruments according to claim 1, characterized in that, Further comprising: Randomly extract 30% of the original data set as a test set; The test set is preprocessed in the same way as the training set; And calculate the correlation coefficient, mean absolute error, and maximum error, respectively, and determine whether the model meets the standard based on the three calculated values. If it is determined that the model does not meet the standard, appropriate processing is performed; The structure and parameters of the qualified model are serialized; And give a unique identifier and store it in the model database of the detection system.

6. The multi-dimensional working condition coupling dynamic calibration method of the electric carbon weighing instrument according to claim 5, characterized in that, The prediction ability of the error model is converted into the content of the calibration action, which specifically includes: Select 1~3 typical working points of the detected device in practical application, real-time collect input variables of current working point ; input variables input a solidified error model, output real-time predicted error ; The original calibration curve of the detector is called to calibrate until the difference between the corrected instrument power measurement value, the instrument concentration measurement value and the true value is less than the preset maximum allowable standard, so as to eliminate the influence of real-time prediction error ; Solving new calibration parameters by reverse optimization algorithm, including power calibration, concentration calibration; generating calibration coefficient set after solving: .

7. The multi-dimension working condition coupled dynamic calibration method of the electric carbon metrological instrument according to claim 6, characterized in that, Further comprising: According to the communication protocol of the detected appliance, the calibration coefficient is packaged into an executable instruction frame; The detection system establishes an encrypted communication link with the detected appliance and performs identity authentication; The instruction frame is sent to the calibration parameter storage area of the appliance, and the appliance MCU updates the internal calibration parameters after receiving it; Send a parameter reading instruction to read the updated calibration parameters in the appliance and compare them with the calculated target value to confirm the success of the injection.

8. The multi-dimension working condition coupled dynamic calibration method of electric carbon metrological instruments according to claim 1, characterized in that, Further comprising: Through repeated testing, evaluation, and iterative processing to ensure the calibration effect and optimize the error model to achieve the continuous evolution of the detection system.

9. A multi-dimensional working condition coupling dynamic calibration device applied to an electric carbon metering instrument, characterized in that, It includes: A working condition environment building module for building a real working condition simulation environment; A data acquisition module for detecting the performance of the electric carbon measurement appliance under the preset working condition and synchronously acquiring the working condition input and measurement error data set to provide original samples for subsequent error modeling; An error model construction module for mining error rules in detection data and constructing a mathematical model, specifically including: According to the error characteristics of the electric carbon measurement appliance, select the appropriate model type, including the following cases: A1: If the coupling relationship between the error and the input variable is clear, use a multivariate polynomial regression model with the following structure: ； wherein, is the coefficient to be found, is the nth input variable; V is the voltage, I is the current, T is the ambient temperature, C is the standard gas concentration; A2: If the error coupling relationship is complex, use a multilayer perceptron neural network with an input layer, a hidden layer, and an output layer; A3: If the sample size of the detection data is small, use a support vector machine to map to a high-dimensional space through a kernel function to achieve high-precision error regression under small samples; Further comprising model training and parameter optimization: For the polynomial regression model in A1: use least squares to minimize the sum of squared prediction errors with the actual errors squared, to solve for the coefficients ; For the multilayer perceptron neural network in A2: use the back propagation algorithm with mean square error as the loss function to optimize the network weights and biases through gradient descent until the loss function converges; For the vector machine model in A3: adjust the penalty parameter and kernel function parameter to minimize the sum of the penalty term of the samples whose error exceeds the allowed threshold and the model complexity penalty term to achieve error regression; A dynamic calibration execution module for converting the prediction ability of the error model into calibration actions, dynamically generating calibration parameters based on real-time working conditions, and automatically injecting the electric carbon measurement appliance.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program, which is executed by the processor to implement all steps of the multi-dimensional working condition coupling dynamic calibration method of the electric carbon measurement appliance according to any one of claims 1-8.

Images