Storage tank metering data processing method, computer equipment and storage medium

Abstract

The invention discloses a storage tank measurement data processing method, computer equipment and a storage medium, and relates to the technical field of measurement modeling, and the method comprises the steps: collecting multi-physics field operation parameter data, carrying out the filtering, interpolation and abnormity elimination, and obtaining a multi-dimensional original data set; establishing a digital twin model based on the data set and calculating an initial parameter set; collecting real-time observation data, and executing residual optimization inversion to obtain a correction parameter set; performing uncertainty analysis and confidence weighted fusion by using the correction parameter set to obtain a fusion measurement parameter set; constructing a multi-objective cost function and updating the weight by adopting an exponential gradient evolution algorithm to obtain feed-forward control intensity; and in combination with the correction parameter set and the feedforward control intensity, carrying out cross-storage-tank robust aggregation and gating judgment to obtain new digital twinborn model parameters. According to the method, self-adaptive fusion and model closed-loop evolution of multi-source metering information are realized through confidence weighted fusion and an exponential gradient evolution weight optimization mechanism and by executing cross-storage-tank robust aggregation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of metering modeling, and in particular to a tank metering data processing method, computer equipment and a storage medium. BACKGROUND

[0002] With the acceleration of digital transformation of petrochemical, chemical and energy industries, tank metering technology gradually develops from traditional static liquid level measurement to multi-source sensing, dynamic modeling and intelligent calibration. In recent years, the use of liquid level, temperature, pressure and structural strain and other multi-physical field parameters to realize accurate estimation of tank volume and mass has become a research hotspot. In order to improve the measurement accuracy, the industry has tried to introduce Internet of Things sensor networks, thermal compensation models and cloud data analysis platforms to realize online monitoring and metering correction of tank state.

[0003] The existing tank metering method mainly uses a single tank and a single sensing signal as the core of the data processing mode, lacks system modeling of the coupling relationship of liquid level, density, temperature and structural strain and other multi-physical fields, and is difficult to reflect the real volume change of the tank under complex thermal and mechanical action. At the same time, the metering results across the tanks have not established global consistency constraints, do not have self-evolution optimization capability, and cannot realize dynamic calibration and collaborative correction among multiple tanks. SUMMARY

[0004] In view of the above existing problems, the present application is proposed.

[0005] Therefore, the present application provides a tank metering data processing method to solve the problems of not establishing a multi-physical field coupling model and being unable to realize dynamic calibration and collaborative correction among multiple tanks in the prior art.

[0006] To solve the above technical problems, the present application provides the following technical solutions: In a first aspect, the present application provides a tank metering data processing method, which comprises collecting multi-physical field operating parameter data, performing noise filtering, time interpolation and abnormality rejection, obtaining a multi-physical field state vector, using a spatial inverse distance weighted algorithm to correct and fuse the multi-physical field state vector in space and time, and obtaining a multi-dimensional original data set; Based on the multi-dimensional original data set, a digital twin model is established, an initial parameter set is calculated, historical calibration data is read, a least squares method is used to fit the parameters of the digital twin model, the error between the output of the digital twin model and the historical calibration data is minimized, and the initial parameter set is obtained; Real-time observation data is collected, the initial parameter set and the real-time observation data are input into the digital twin model, a residual error is calculated, a proximal optimization inversion is performed, and a corrected parameter set is obtained; The uncertainty propagation analysis is performed on the multi-physical field operation parameter data by the correction parameter set, the volume and mass uncertainties are calculated, the uncertainty is weighted and fused by a confidence weighted fusion algorithm, the fused volume and fused mass are calculated, the comprehensive uncertainty evaluation is performed, and the fused metrological parameter set is obtained; According to the fused metrological parameter set, a multi-objective cost function is constructed, weight updating is performed through an exponential gradient evolution algorithm, and a feedforward control strength is obtained. Based on the correction parameter set and the feedforward control strength, cross-tank robust aggregation is performed to obtain a global correction parameter, and the new digital twin model parameter is obtained through consistency and uncertainty gating.

[0007] As a preferred scheme of the storage tank metrological data processing method, the multi-physical field operation parameter data is collected, noise filtering, time interpolation and abnormal rejection are performed, a multi-physical field state vector is obtained, a spatial inverse distance weighting algorithm is used to correct and fuse the multi-physical field state vector in space and time, and a multi-dimensional original data set is obtained, and the specific steps are as follows, A distributed sensor network is established, and the multi-physical field operation parameter data is collected through a master clock synchronization mechanism; Noise filtering, time interpolation and abnormal rejection are performed on the multi-physical field operation parameter data to obtain a trusted data stream, and a multi-physical field state vector is calculated through the trusted data stream; The spatial inverse distance weighting algorithm is used to correct and fuse the multi-physical field state vector in space and time to obtain a multi-dimensional original data set.

[0008] As a preferred scheme of the storage tank metrological data processing method, the multi-physical field operation parameter data is collected, noise filtering, time interpolation and abnormal rejection are performed, a multi-physical field state vector is obtained, a spatial inverse distance weighting algorithm is used to correct and fuse the multi-physical field state vector in space and time, and a multi-dimensional original data set is obtained, and the specific steps are as follows, Based on the multi-dimensional original data set, a liquid level-volume sub-model, a thermal coupling sub-model and a structure strain sub-model are established to obtain a digital twin model; The historical calibration data is read, the historical calibration data is input into the digital twin model, and the error between the output of the digital twin model and the calibration data is calculated; The least square method is used to fit the parameters of the digital twin model, the error between the output of the digital twin model and the calibration data is minimized, the initial values of the Bezier shape coefficient, the thermal expansion coefficient, the pressure compression coefficient, the tank deformation influence coefficient and the reference density are calculated, and the initial parameter set is obtained.

[0009] As a preferred scheme of the storage tank metering data processing method, wherein: the real-time observation data is collected, the initial parameter set and the real-time observation data are input into the digital twin model, the residual error is calculated, the proximal optimization inversion is performed, and the corrected parameter set is obtained, and the specific steps are as follows, The real-time observation data is collected, and the real-time observation data and the initial parameter set are input into the digital twin model to calculate the predicted volume and the predicted mass. The predicted volume and the predicted mass are compared with the measured volume value and the measured mass value to obtain the residual error of the volume and the mass, and when the residual error exceeds the dynamic threshold range, the proximal optimization inversion based on the residual error driving is performed to update the parameters of the digital twin model. After multiple iterations, when the mean square value of the residual error of the volume and the mass is lower than the preset convergence threshold, the corrected parameter set is obtained.

[0010] As a preferred scheme of the storage tank metering data processing method, wherein: the uncertainty propagation analysis is performed on the multi-physical field operation parameter data by using the corrected parameter set, the uncertainty of the volume and the mass is calculated, the confidence weighted fusion algorithm is used to weight and fuse the uncertainty, the fused volume and the fused mass are calculated, the comprehensive uncertainty evaluation is performed, and the fused metering parameter set is obtained, and the specific steps are as follows, The corrected parameter set is applied to the digital twin model, a sensitivity mapping relationship is established by applying a small amplitude disturbance to the multi-physical field operation parameter data. According to the sensitivity mapping relationship, the measurement variance, the temperature drift and the noise power of the multi-physical field operation parameter data are weighted and accumulated to calculate the uncertainty of the volume and the mass. The confidence weighted fusion algorithm is used to weight and fuse the uncertainty, the fused volume and the fused mass are calculated, the comprehensive uncertainty evaluation is performed, and the fused metering parameter set is obtained.

[0011] As a preferred scheme of the storage tank metering data processing method, wherein: according to the fused metering parameter set, a multi-objective cost function is constructed, a weight update is performed by using an exponential gradient evolution algorithm, and a feedforward control strength is obtained, and the specific steps are as follows, Based on the fused metering parameter set, precision, stability and calculation efficiency indicators are extracted, a multi-objective cost function is constructed, and the weights of the precision, stability and calculation efficiency indicators are updated by using an exponential gradient evolution algorithm to obtain a weight distribution. According to the weight distribution, the preference of the current working condition for accuracy priority, stability priority and efficiency priority is comprehensively evaluated, and the feedforward control strength is adjusted.

[0012] As a preferred scheme of the method for processing tank metering data, the method comprises the following steps of: Based on the correction parameter set, the parameter increment of the current tank is calculated, and the weighted geometric median-Huber robust aggregation is performed on the parameter increments from the plurality of tanks to obtain the global correction parameter.

[0013] As a preferred scheme of the method for processing tank metering data, the method comprises the following steps of: The global correction parameter is determined by the consistency and uncertainty gating mode, and the full update, partial update or freezing is performed, the running mode of the digital twin model is adjusted based on the feedforward control strength, and the new digital twin model parameter is obtained.

[0014] In a third aspect, the present application provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and wherein the computer program is executed by the processor to implement any step of the method for processing tank metering data according to the first aspect of the present application.

[0015] In a fourth aspect, the present application provides a computer readable storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement any step of the method for processing tank metering data according to the first aspect of the present application.

[0016] The present application has the following advantages: by constructing a multi-submodel digital twin system integrating Bezier fitting, thermal coupling and structure compensation, and based on least squares optimization of historical calibration data, the collaborative modeling of tank geometry, thermal properties and elastic response and the acquisition of initial parameter set are realized; by constructing a digital twin model including liquid level-volume, thermal coupling and structure strain and performing parameter self-calibration based on residual driving, dynamic updating and adaptive convergence of model parameters are realized; by introducing confidence weighted fusion and exponential gradient evolution weight optimization mechanism and performing cross-tank robust aggregation, adaptive fusion of multi-source metering information and model closed-loop evolution are realized. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0018] Fig. 1 The flowchart of the method for processing tank metering data.

[0019] Fig. 2 Flowchart for digital twin modeling and calibration.

[0020] Fig. 3 Flowchart for uncertainty propagation and control weight update.

[0021] Fig. 4 Flowchart for parameter aggregation and model closed loop update. DETAILED DESCRIPTION

[0022] In order to make the above objectives, features and advantages of the present application more apparent, specific embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0023] In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present application. However, it will be apparent to one skilled in the art that the present application can be practiced without the specific details given herein, that the present application can be practiced with other than the described embodiments, and that variations from the particular embodiments described herein can be made and still be within the scope of the present application.

[0024] Secondly, the "one embodiment" or "embodiment" referred to herein means that the specific features, structures or characteristics can be included in at least one implementation of the present application. "In one embodiment" appearing in different places in the specification does not mean the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments.

[0025] Reference Figs. 1-4 For one embodiment of the present application, the embodiment provides a storage tank metering data processing method, comprising the following steps: S1, collect the multi-physical field operating parameter data of the storage tank, perform noise filtering, time interpolation and abnormality rejection, and obtain a multi-dimensional original data set.

[0026] A distributed sensor network is established to collect the multi-physical field information of the storage tank. The sensor network includes a liquid level sensor, a temperature sensor, a pressure sensor and a strain gauge, which are respectively installed at the liquid region, the tank wall and the support structure of the storage tank, for simultaneously collecting liquid level signals, temperature signals, pressure signals and structural strain signals, and obtaining multi-channel signals. In order to ensure the time consistency of the multi-channel signals, a network time synchronization strategy taking the tank top liquid level sensor as the master clock is adopted to uniformly correct the clocks of each node, so that the data sampling time error of different sensors is not more than 1 millisecond, and a multi-dimensional original data vector is obtained.

[0027] To eliminate random noise in the sensor signal, within a preset time window, several sampling values before and after the current sensor sampling point are weighted and averaged according to the time weight to weaken the high-frequency fluctuation. The cubic spline interpolation algorithm is used to time-interpolate the low-frequency signal, so that all signals are resampled in the same time interval to obtain a uniform sampling step data sequence on the time axis. At each time, the liquid level, temperature, pressure and strain signals are converged to form the multi-physical field state vector of the storage tank at the current time. In order to effectively identify abnormal fluctuations caused by temperature sudden change, bubble disturbance or structure impact, and significantly improve the overall credibility of the data, a multivariate anomaly identification method based on Mahalanobis distance is used to calculate the Mahalanobis distance, which is expressed as: ; wherein, represents the Mahalanobis distance square of the corresponding multi-dimensional observation data at time . represents the multi-physical field state vector at time . represents the mean value of the multi-physical field state vector, represents the inverse matrix of the covariance matrix.

[0028] It should be noted that, is obtained by statistically averaging the credible sampling data in the set historical time window, reflecting the typical physical parameter level of the storage tank in the stable running state.

[0029] When the Mahalanobis distance square exceeds the preset confidence identification threshold, it is determined that the multi-physical field state vector at the current time is abnormal, and is marked as an invalid sample. For continuous abnormal sections, a time neighborhood interpolation algorithm is used for repair to maintain time continuity and obtain a credible data stream that eliminates abnormal interference and maintains time consistency.

[0030] It should be noted that the preset confidence identification threshold is constructed from the residual distribution of the high-confidence historical samples, the Mahalanobis distance square value of each sample is calculated, and the critical value of the chi-square distribution under the corresponding degree of freedom (for example, 95% confidence level) is used as the initial identification threshold. Subsequently, in the real-time running process, the change trend of the residual vector is combined to dynamically calculate the Mahalanobis distance change amplitude and its stability index. When abnormal fluctuations are frequent, the threshold is appropriately adjusted downward to improve sensitivity, and when running smoothly, it is adjusted upward to suppress false positives.

[0031] The space-time correction processing is performed on the credible data stream.

[0032] Further, an algorithm based on spatial inverse distance weighting (IDW) is used to weight each node data according to its distance from the center of the tank, realize spatial fusion at the same time, and obtain a multi-dimensional original data set representing the overall running state of the storage tank, which is expressed as: ; in, Represents the original multidimensional dataset. Indicates the first Each sensing node at time Collected multiphysics state vectors. Indicates the first The spatial distance from each sensor node to the geometric center of the storage tank Indicates the first Spatial weight coefficients for each sensor node.

[0033] S2. Establish a digital twin model based on the multidimensional original dataset and calculate the initial parameter set.

[0034] Based on the multidimensional original dataset, a digital twin model is established. The digital twin model includes three sub-models: a liquid level-volume sub-model, a thermo-coupling sub-model, and a structural strain sub-model. To explicitly express the nonlinear relationship between the tank liquid level and volume, a geometric fitting model based on piecewise Bessel functions is adopted to establish the liquid level-volume sub-model, with the expression: ; in, Indicates at time From liquid level height The corresponding storage tank geometric volume, Represents pi (π). This indicates the reference radius at the bottom of the storage tank. Indicates the first Each Bessel shape coefficient Indicates the first Bessel basis functions This represents the normalized variable along the height of the tank. This represents the total number of orders of Bessel basis functions. Represents the differential symbol.

[0035] It should be noted that the Bessel basis function is a smooth basis function that controls the continuous change of the cross-sectional radius of the tank with height. It achieves the gradual change of shape from the bottom of the tank, the tank wall to the top of the tank by changing the weight of the control points of the Bessel curve, and can keep the curve smooth and without sharp corners when splicing segments.

[0036] To characterize the combined effects of temperature and pressure on liquid density and volume, a thermo-mechanical coupling sub-model is established, with the following expression: ; in, Indicates at time The density of the liquid, This represents the reference density of a liquid under standard conditions and is a common density coefficient for liquids. represents the volumetric thermal expansion coefficient of the liquid, which belongs to the liquid regular coefficient, represents the pressure compressibility coefficient of the liquid, which belongs to the liquid regular coefficient, represents the volumetric thermal expansion coefficient of the liquid at time represents the average temperature of the medium in the tank, represents the reference temperature, represents the volumetric thermal expansion coefficient of the liquid at time represents the absolute pressure value in the tank, represents the reference pressure.

[0037] It should be noted that, , and preferentially adopt the national or industry standard value, and generally take , .

[0038] In order to consider the elastic deformation of the tank body under the action of thermal expansion and internal pressure, a structural strain correction term is introduced to compensate the geometric volume of the liquid level, a structural strain sub-model is established, and the expression is: ; Among them, represents the volumetric thermal expansion coefficient of the liquid at time the volumetric correction amount caused by the deformation of the tank body, used to correct the deviation of , represents the tank body deformation influence coefficient, represents the average strain value of the tank structure at time .

[0039] It should be noted that, is the local strain value obtained by spatially weighted average of the local strain value collected by the distributed strain sensor array in real time; is the elastic deformation under the simulation of typical thermal expansion and internal pressure conditions, and the volume change ratio caused by the average strain is obtained by linear regression.

[0040] Read the historical calibration data, including the liquid level-volume calibration curve, temperature-density correction table and tank body strain response data, calculate the residual error between the model prediction value and the measured calibration value for each group of historical observation data, construct the residual error square sum between the prediction value and the measured calibration value as the objective function, based on the objective function, use the conventional least square method to solve the parameters, take the Bezier shape coefficient, the volumetric thermal expansion coefficient, the pressure compressibility coefficient, the tank body deformation influence coefficient and the reference density as the optimization variables, iteratively update to minimize the residual error square sum, and output the parameter combination that minimizes the residual error as the initial parameter set of the digital twin model.

[0041] S3, collecting real-time observation data, inputting the initial parameter set and the real-time observation data into the digital twin model, calculating the residual error, performing proximal optimization inversion, and obtaining the corrected parameter set.

[0042] In each sampling period (e.g., 1 second), the liquid level, temperature, pressure, and strain signals in the storage tank are read, the real-time input vector is obtained, the real-time input vector is input into the digital twin model together with the initial parameter set, and the theoretical prediction value is calculated, including the predicted volume value for reflecting the geometric volume corrected according to the liquid level and the strain, and the predicted mass value for reflecting the medium mass corrected according to the temperature and the pressure. At the same time, the measured values of volume and mass are synchronously collected, and the measured values of volume and mass are independently measured by the liquid level sensor and the mass flow meter, respectively. The measured values of volume and mass are compared with the theoretical prediction values output by the digital twin model, the deviation between the measured values and the theoretical prediction values is calculated, the residual errors of volume and mass are obtained, and when the residual errors of volume and mass exceed the dynamic threshold range, the digital twin model self-calibration process is performed.

[0043] It should be noted that the dynamic threshold is the mean and standard deviation of the volume residual error and the mass residual error in the latest sliding time window (e.g., the last 24 hours or 1000 sampling periods), the historical mean is taken as the residual error center baseline, and several times of the standard deviation are taken as the upper and lower limits of the deviation to form the dynamic threshold interval.

[0044] In order to enable the digital twin model to automatically correct the parameters in operation, a proximal optimization inversion algorithm based on residual error driving is adopted, the parameters are corrected through the sensitivity matrix output by the digital twin model, the adaptive convergence of the digital twin model from the initial parameters to the corrected parameters is realized, and the expression is: ; wherein, represents the corrected parameter set at time , represents the initial parameter set, represents the convergence learning step, represents the Jacobian matrix of the digital twin model, represents the residual error weighted matrix, represents the residual error vector.

[0045] It should be noted that, is determined by combining offline calibration and online adaptive adjustment, the initial step is calculated according to the average sensitivity of the residual error to the parameters, and the step is dynamically corrected according to the residual error change trend during online operation, the step is gradually reduced to improve stability when the residual error continuously decreases, and the step is automatically increased to improve the response speed when the residual error rises or oscillates. Sensitivity matrix of digital twin model output to parameters, obtained by finite difference method, and its elements are the first-order partial derivatives of digital twin model output (volume, mass) to each parameter in the parameter set. is determined by statistics of the initial weight of each output channel historical residual variance, and the current weight is updated in real time combined with the sliding window, so that the weight dynamically reflects the confidence degree of different output channels, and can dynamically reflect the confidence difference of each output channel.

[0046] To ensure that the parameters of the digital twin model are physically reasonable and do not diverge, perform a proximal constraint mapping operation on the parameters, project the corrected parameter set into a pre-set physical constraint set, and map the thermal expansion coefficient to 0 to between to cover the thermal expansion characteristics of common liquids, map the pressure compression coefficient to 0 to between to cover the pressure response range of industrial liquids, and map the tank deformation influence coefficient to 0 to 2 to reflect the elastic deformation ratio of different tank structures, and keep the reference density in the range of 500 to 1500 kg / m³ to cover the standard density interval of various liquid media. If these ranges are exceeded, the calculation results of the digital twin model will not conform to the physical laws, causing volume prediction deviation or parameter update divergence.

[0047] When the mean square value of the residual after the digital twin model is iterated for a certain number of times is lower than the preset convergence threshold, it indicates that the parameters of the digital twin model have converged, and the calibration process is stable. The calibrated parameter set is saved to the digital twin model.

[0048] It should be noted that the preset convergence threshold is determined adaptively according to the sensor noise statistics and real-time residual fluctuations.

[0049] S4, propagate the uncertainty of the multi-physical field operating parameter data through the corrected parameter set, and obtain the fused measurement parameter set through the confidence weighted fusion algorithm.

[0050] The liquid level, density and temperature data output by each sensor are input into the digital twin model calibrated by the calibrated parameter set. Taking the calibrated parameter set as the reference, a minimum amplitude disturbance is applied to the liquid level, density and temperature respectively. The change difference of the digital twin model output volume and mass before and after the disturbance is calculated, and the sensitivity value is represented by the ratio of the change difference to the disturbance. Through this one-by-one disturbance and comparison method, the sensitivity of each input variable is obtained. These sensitivity values form a sensitivity mapping matrix. According to the sensitivity mapping matrix, the measurement variance, temperature drift and noise power of each input variable are weighted and accumulated, so that the input signals with high sensitivity and large error have more significant contribution to the output uncertainty, thereby quantitatively calculating the uncertainty of volume and mass.

[0051] Since the liquid level method, density method and mass method estimate the state of the storage tank from the geometric, physical and mechanical angles respectively, Each has different precision and noise characteristics, in order to improve the overall measurement accuracy, the fusion algorithm based on confidence weighting is adopted, the weight is automatically allocated according to the uncertainty of each estimation method, and multiple independent results are fused into a global reliable result, the expression is: ; ; Among them, Indicates the volume or mass at time The fusion estimation value of volume or mass obtained by confidence weighted fusion, Indicates the volume or mass estimation value obtained by the Independent measurement channel (for example, liquid level method, density method or mass method) at time , Indicates the confidence weight coefficient of the Independent measurement channel at time , Indicates the uncertainty variance of the Independent measurement channel at time .

[0052] The comprehensive uncertainty evaluation is performed on the fusion estimation value of volume or mass, the uncertainty variances of each measurement channel are calculated, and the comprehensive uncertainty of the fusion output is calculated by weighted accumulation according to the weight square proportion, the confidence interval is calculated according to the comprehensive uncertainty, which is used to measure the reliability of the fusion result, when the confidence interval width is lower than the preset confidence threshold, it is determined that the current fusion result is reliable, and the fusion volume, fusion mass and comprehensive uncertainty are packaged as a fusion measurement parameter set.

[0053] It should be noted that the preset confidence threshold is based on a large number of known real volume or mass calibration samples, the confidence interval width distribution corresponding to the fusion estimation value is calculated, and the quantile value covering more than 95% of normal samples is selected as the initial confidence threshold, and the confidence interval width distribution of recent fusion results is continuously monitored in the running stage, if there is a clear convergence trend (for example, most of the width is significantly reduced) in a certain time window, it is considered that the measurement quality is improved, and the confidence threshold is automatically tightened, otherwise if the fluctuation is intensified, the threshold is appropriately relaxed to avoid misjudgment.

[0054] S5, according to the fusion measurement parameter set, a multi-objective cost function is constructed, the weight is updated by exponential gradient evolution algorithm, and the feedforward control strength is obtained.

[0055] In each sampling period, according to the change of fusion volume and fusion quality, precision index, stability index and calculation efficiency index are extracted, the precision index is obtained by comparing the proportion of comprehensive uncertainty and allowable upper limit, the stability index is obtained by calculating the short-term variance through the fluctuation amplitude of fusion volume and quality in the sliding window, and the calculation efficiency index is obtained by measuring the proportion of calculation time and target time limit in this period, reflecting the efficiency of calculation resources.

[0056] It should be noted that the allowable upper limit is the initial relative uncertainty limit value set by the industry measurement standard, and is corrected according to the geometric structure of the storage tank and the resolution parameters of the sensor. According to the statistical mean of the comprehensive uncertainty in the recent fusion measurement process, the allowable upper limit is dynamically updated by exponential smoothing. The target time limit is set to 60% to 80% of the sampling period, and is dynamically updated by exponential smoothing according to the statistical results of the calculation time in the recent period. If the target time limit is set to less than 50% of the sampling period, the fusion task may be interrupted or the result may be delayed due to calculation timeout when the data volume fluctuates or the operation load is high. If the target time limit is set too high, more than 90% of the sampling period, there will be no buffer margin when a sudden delay occurs, affecting the stability and continuity of the fusion output, and even may conflict with the next sampling period.

[0057] The precision index, stability index and calculation efficiency index are used as three sub-targets of multi-objective optimization, and a comprehensive cost function is constructed by linear weighting to balance the contradictory relationship between different targets. In order to make the weight automatically adjust with the working condition, exponential gradient evolution algorithm is used for updating to maintain non-negativity and normalization, while ensuring exponential sensitive response to index change, the expression is: ; Among them, represents the weight coefficient of the i-th target at time t, represents the updated weight of the i-th target at time t, represents the normalized value of the i-th target at time t, represents the weight learning step.

[0058] It should be noted that, is determined by off-line simulation calibration to determine the initial value, and in the running process, the ratio of the normalized value and the weight change rate is automatically adjusted.

[0059] ​​​​​​According to the weight distribution updated by the exponential gradient evolution algorithm in the last period, the current preference for accuracy priority, stability priority or efficiency priority is comprehensively evaluated. If the accuracy weight and stability weight are higher under the current working condition, it indicates that the digital twin model needs to maintain strong measurement reliability and data smoothness, so the feedforward control strength is automatically increased, the data sampling frequency is increased, the filter window is widened, and the accuracy level is improved, thereby strengthening the suppression of small fluctuations and noise disturbances. If the calculation efficiency weight rises, it is judged that the current working condition requires higher response speed, at this time the feedforward control strength is reduced, the data sampling frequency or the calculation window is appropriately reduced, so as to reduce the operation load and improve the real-time performance. Through this adaptive adjustment method based on weight proportion, the feedforward control strength can automatically realize the optimal compromise between stability and efficiency under different operating conditions, ensuring the accuracy of the tank measurement results and maintaining the real-time response capability.

[0060] S6, based on the correction parameter set and the feedforward control strength, cross-tank robust aggregation is performed to obtain global correction parameters, which are fused through consistency and uncertainty gating to obtain new digital twin model parameters.

[0061] The difference between the calibration parameter set and the parameters in the last period of stable operation is taken as the parameter increment, the reliability weight is calculated according to the comprehensive uncertainty level of each tank, the smaller the uncertainty, the higher the weight, the reliability weight is obtained, the parameter increment and the reliability weight from multiple tanks are weighted and geometrically median-Huber robustly aggregated to resist abnormal sites and working condition mutations, and the expression is: ; Among them, represents the global correction parameter vector, represents the variable that minimizes the objective function , that is, to find the optimal global correction vector, represents the aggregation sum of data from tank sites, represents the reliability weight coefficient of the th tank, represents the Huber loss function, represents the local parameter increment vector of the th tank.

[0062] Compare the change direction of each component of the global correction parameter vector with the change direction at the last parameter calibration. If the change directions are consistent, it indicates that the global correction is consistent with the convergence trend of the digital twin model, and the current correction is considered reasonable in physical sense, and is directly updated. If the change directions are opposite, it is considered inconsistent, and the correction amplitude of the component is reduced or temporarily suspended to prevent introducing reverse bias.

[0063] By synthesizing the uncertainty and the aggregation uncertainty, the uncertainty gating judgment is performed on each parameter component. When the root mean square of the synthesized uncertainty is lower than the allowed upper limit, it is considered that the accuracy is higher, and when the aggregation uncertainty does not exceed 1.5 times of the long-term average value, it is considered that the aggregation result is stable, and the update weight of the current component is set to 1 to perform full update. If one of the synthesized uncertainty and the aggregation uncertainty does not meet the standard, the update weight is reduced to 0.5 for partial update, and if both the synthesized uncertainty and the aggregation uncertainty exceed the limit, the current component is frozen, and the next period is waited for re-aggregation.

[0064] It should be noted that the aggregation uncertainty is obtained by statistically analyzing the variance of the aggregation residuals of the plurality of tank parameters.

[0065] Based on the feedforward control strength, adaptive adjustment is performed on the digital twin model.

[0066] Further, when the feedforward control strength is high, it means that the accuracy weight and the stability weight are large, and the digital twin model pays more attention to accuracy maintenance, and then enters the recalibration mode. By shortening the data sampling interval, increasing the data filtering order and reducing the parameter update step, the digital twin model re-estimates the key parameters under high confidence, thereby eliminating residual errors and improving the stability of the digital twin model. When the feedforward control strength is at a medium level, the normal operation is maintained, and only light recalibration is performed within a preset time window. Without interrupting the operation, the real-time monitoring data within the short time window is used to perform small-step local correction on the key parameters with large deviation, so as to quickly eliminate small drifts and maintain the accuracy of the digital twin model. When the feedforward control strength is low, it means that the calculation efficiency weight in the last period is high, and then the fast running mode is automatically switched to, the sampling frequency is reduced, the filtering constraint is relaxed, and the existing parameters are maintained, so as to improve the calculation response speed. Through the adaptive control mechanism based on the feedforward control strength, the balance between accuracy and efficiency can be dynamically adjusted according to the actual working condition in each running period, so as to realize the continuous self-calibration and stable operation of the digital twin model. In the next sampling period, the global correction parameter vector is set as the new initial parameter set, the parameters are continuously iteratively updated between periods, and a closed-loop optimization mechanism that can adaptively evolve is formed.

[0067] The embodiment also provides a computer device suitable for the tank metering data processing method, which includes a memory and a processor. The memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions to realize the tank metering data processing method proposed in the above embodiment.

[0068] The computer device can be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected by a system bus. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for running the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is configured to perform wired or wireless communication with an external terminal. The wireless communication can be achieved by WIFI, an operator network, NFC (Near Field Communication) or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the computer device, or an external keyboard, touchpad or mouse, etc.

[0069] The embodiment also provides a storage medium having a computer program stored thereon, the program being executed by a processor to implement the processing method for realizing tank metering data as proposed in the above embodiment; and the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic memory, a flash memory, a magnetic disk or an optical disk.

[0070] To sum up, the present application achieves the following effects: a multi-submodel digital twin system is constructed by fusing Bezier fitting, thermal coupling and structure compensation, and based on least square optimization of historical calibration data, the collaborative modeling of tank geometry, thermal properties and elastic response and the acquisition of initial value parameter set are realized; a digital twin model is constructed by fusing liquid level-volume, thermal coupling and structure strain, and based on residual driven parameter self-calibration, the dynamic updating and adaptive convergence of model parameters are realized; a confidence weighted fusion and exponential gradient evolution weight optimization mechanism is introduced, and cross-tank robust aggregation is performed, so that adaptive fusion of multi-source metering information and model closed-loop evolution are realized.

[0071] It should be noted that the above examples are only used to illustrate the technical solutions of the present application but not limit the present application. Although the present application is described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or equivalently replaced, without departing from the spirit and scope of the technical solutions of the present application, which should be covered in the scope of the claims of the present application.

Claims

1. A method of processing tank gauging data, characterized by: comprise, Collecting multi-physical field operation parameter data, performing noise filtering, time interpolation and abnormality rejection, obtaining a multi-physical field state vector, performing space inverse distance weighted algorithm on the multi-physical field state vector for space-time correction and fusion, and obtaining a multi-dimensional original data set; Based on the multi-dimensional original data set, a digital twin model is established, an initial parameter set is calculated, historical calibration data is read, a least square method is used to fit the parameters of the digital twin model, the error between the output of the digital twin model and the historical calibration data is minimized, and the initial parameter set is obtained; Collecting real-time observation data, inputting the initial parameter set and the real-time observation data into the digital twin model, calculating the residual error, performing proximal optimization inversion, and obtaining a corrected parameter set; Through the corrected parameter set, uncertainty propagation analysis is performed on the multi-physical field operation parameter data, the volume and mass uncertainties are calculated, the uncertainties are weighted and fused by using a confidence weighted fusion algorithm, the fused volume and fused mass are calculated, comprehensive uncertainty evaluation is performed, and a fused metrological parameter set is obtained; According to the fused metrological parameter set, a multi-objective cost function is constructed, the weight is updated by using an exponential gradient evolution algorithm, and a feedforward control strength is obtained; Based on the corrected parameter set and the feedforward control strength, cross-tank robust aggregation is performed, global correction parameters are obtained, and the new digital twin model parameters are obtained through consistency and uncertainty gating.

2. The method of claim 1, wherein: The specific steps of collecting multi-physical field operation parameter data, performing noise filtering, time interpolation and abnormality rejection, obtaining a multi-physical field state vector, and performing space inverse distance weighted algorithm on the multi-physical field state vector for space-time correction and fusion to obtain a multi-dimensional original data set are as follows, A distributed sensor network is established, and multi-physical field operation parameter data is collected through a master clock synchronization mechanism; Noise filtering, time interpolation and abnormality rejection are performed on the multi-physical field operation parameter data to obtain a trusted data stream, and a multi-physical field state vector is calculated through the trusted data stream; The space inverse distance weighted algorithm is used to correct and fuse the multi-physical field state vector in space and time to obtain a multi-dimensional original data set.

3. The method of claim 2, wherein: The specific steps of establishing a digital twin model based on the multi-dimensional original data set, calculating an initial parameter set, reading historical calibration data, fitting the parameters of the digital twin model by using a least square method, minimizing the error between the output of the digital twin model and the historical calibration data, and obtaining the initial parameter set are as follows, Based on the multi-dimensional original data set, a liquid level-volume sub-model, a thermal coupling sub-model and a structure strain sub-model are established to obtain a digital twin model; The historical calibration data is read and input into the digital twin model, and the error between the output of the digital twin model and the calibration data is calculated; The least square method is used to fit the parameters of the digital twin model, the initial values of the Bezier shape coefficient, the thermal expansion coefficient, the pressure compression coefficient, the tank deformation influence coefficient and the reference density are calculated by minimizing the error between the output of the digital twin model and the calibration data, and the initial parameter set is obtained.

4. The method of claim 3, wherein: The specific steps of collecting real-time observation data, inputting the initial parameter set and the real-time observation data into the digital twin model, calculating the residual error, performing proximal optimization inversion, and obtaining a corrected parameter set are as follows, Collect real-time observation data, input the real-time observation data and the initial parameter set into the digital twin model, calculate the predicted volume and the predicted mass; Compare the predicted volume and the predicted mass with the measured volume value and the measured mass value, obtain the residual error of the volume and the mass, when the residual error exceeds the dynamic threshold range, perform near-end optimization inversion based on the residual error, update the parameters of the digital twin model; After multiple iterations, when the mean square value of the residual error of the volume and the mass is lower than the preset convergence threshold, obtain the corrected parameter set.

5. The method of claim 4, wherein: The method comprises the following steps of: Apply the corrected parameter set to the digital twin model, and establish a sensitivity mapping relationship by applying a small amplitude disturbance to the multi-physical field operation parameter data; According to the sensitivity mapping relationship, the measurement variance, temperature drift and noise power of the multi-physical field operation parameter data are weighted and accumulated, and the uncertainty of the volume and the mass is calculated; The confidence weighted fusion algorithm is used to weight and fuse the uncertainty, calculate the fused volume and the fused mass, perform comprehensive uncertainty evaluation, and obtain the fused metrological parameter set.

6. The method of claim 5, wherein: According to the fused metrological parameter set, a multi-objective cost function is constructed, the weight is updated through an exponential gradient evolution algorithm, and the feedforward control strength is obtained. Based on the fused metrological parameter set, the precision, stability and calculation efficiency indicators are extracted, a multi-objective cost function is constructed, and the weights of the precision, stability and calculation efficiency indicators are updated using an exponential gradient evolution algorithm to obtain the weight distribution. According to the weight distribution, the preference of the current working condition for accuracy priority, stability priority and efficiency priority is comprehensively evaluated, and the feedforward control strength is adjusted.

7. The method of claim 6, wherein: Based on the corrected parameter set and the feedforward control strength, cross-tank robust aggregation is performed to obtain global correction parameters. Based on the corrected parameter set, the parameter increment of the current tank is calculated, and the parameter increments from multiple tanks are weighted and geometrically median-Huber robustly aggregated to obtain global correction parameters.

8. The method of claim 7, wherein: The new digital twin model parameters are obtained by consistency and uncertainty gating, and the specific steps are as follows: The global correction parameters are determined by consistency and uncertainty gating, and full update, partial update or freezing is performed, the running mode of the digital twin model is adjusted based on the feedforward control strength, and the new digital twin model parameters are obtained. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is characterized in that: The processor executes the computer program to realize the steps of the storage tank metrological data processing method of any one of claims 1-7.

10. A computer readable storage medium having stored thereon a computer program, characterized in that: The computer program is executed by the processor to realize the steps of the storage tank metrological data processing method of any one of claims 1-7.

Images