Industrial process system data online reconciliation method based on dynamic data reconciliation

Abstract

The application relates to an industrial process system data online coordination method based on dynamic data coordination, which comprises the following steps: acquiring random errors of each measurement variable of a system, establishing linear constraints between coordination values corresponding to measurement values of the measurement variables and unmeasured variables based on system mechanism, constructing a dynamic data coordination model based on the linear constraints, remodeling and augmenting a model data matrix and a vector, solving the model to obtain a least square solution, and reasonably setting a sliding window length and constraint equations to obtain high-precision and high-timeliness system online operation data during actual operation. When the method faces dynamic operation of a thermal power unit, the precision of high-pressure heater measurement data can be greatly improved in real time and accurately, and high-quality data are provided for dynamic performance monitoring.

Description

Technical Field

[0001] This invention relates to the field of data coordination methods, and in particular to an online data coordination method for industrial process systems based on dynamic data coordination. Background Technology

[0002] With the rapid penetration of intermittent renewable energy, frequent changes in the operating conditions of thermal power units have become commonplace. Comprehensive online monitoring of the units, enabling real-time detection and analysis of equipment and system operating indicators and status, can more effectively identify anomalies or potential problems, allowing for early fault detection and handling, and improving the reliability of unit operation. This is crucial for ensuring the safe and stable operation of thermal power units. However, since sensors inevitably have measurement errors, the effectiveness of online monitoring methods largely depends on data quality. Poor data quality can lead to inaccurate or delayed production decisions, resulting in production delays and reduced capacity. In the long run, inefficient use and operation of equipment will cause greater damage, shorten equipment lifespan, and ultimately affect the stable operation of the unit.

[0003] Data coordination methods offer a rapid and efficient way to improve the quality of performance monitoring data. This method utilizes the operating data of thermal power units and the physical characteristics of the equipment to perform data coordination modeling on the measurement redundancy of specific equipment and thermodynamic processes, thereby reducing the measurement standard deviation. Depending on the application scope, data coordination is divided into steady-state data coordination and dynamic data coordination. In steady-state data coordination research, steady-state analysis is required to select a suitable steady-state dataset; conversely, dynamic methods are applicable to all datasets, including both steady-state and dynamic datasets. However, due to the complexity of thermal power units and the lack of operating data, research on dynamic data coordination in actual power plants is quite insufficient. Especially for high-pressure heaters, which operate under high pressure and are prone to failure, considering the unique dynamic characteristics of mass accumulation in high-pressure heaters, their liquid levels are constantly changing, and the calculated values ​​obtained from steady-state models often deviate significantly from the measured values. Therefore, steady-state monitoring and steady-state data coordination are no longer applicable to thermal power units operating under dynamic conditions, making research on dynamic monitoring and dynamic data coordination urgently needed. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides an online data coordination method for industrial process systems based on dynamic data coordination. This method solves the problem of low quality of dynamic data in industrial process systems and improves the efficiency, economy, and security of system operation.

[0005] The technical solution adopted in this invention is as follows:

[0006] This invention provides an online data coordination method for industrial process systems based on dynamic data coordination, comprising:

[0007] (1) Select the measured variable and the unmeasured variable u from the target industrial process system, and collect the dataset of the measured variable. Where y represents the sample of measured values ​​of the measured variable, and the subscript t1t N Represents the sampling time; defines the true value of the measured variable. and random error ε, Define the coordination value x corresponding to y. p represents the number of measurement variables;

[0008] Wherein, ε follows a mean of 0 and a covariance of . A multidimensional standard Gaussian distribution, σ1,σ2,σ3…σ p This represents the uncertainty corresponding to each measured value in the sample y.

[0009] (2) Based on the dynamic model of the equipment in the target industrial process system, establish a set of constraint equations for the coordination value x and the unmeasured variable u;

[0010] A dynamic data reconciliation model is constructed using measured sample y, reconciliation value x, and a set of constraint equations:

[0011]

[0012]

[0013] Among them, R MSE Let f(·) be the objective function, N be the number of sampling times (i.e., the number of samples), i be the sampling time, and T and -1 be the transpose and inverse of the matrix, respectively; f(·) is the system of differential equations; g(·) is the system of algebraic equations of inequalities.

[0014] (3) The dataset The data in the sample is first sorted by the order of the measured variables, then by the order of the samples, and then rearranged. Correspondingly, the coordination values ​​x are also rearranged in the same order to form x′, where x′ and y′ have the same dimension. This transforms the above dynamic data coordination model into:

[0015]

[0016] stA N x′=O

[0017] Wherein, the random error matrix ∑ t Let be the covariance matrix of the uncertainty of the sample measurement at time t;

[0018] constraint equation matrix A tLet be the set of constraint equations followed by the measured values ​​at time t, which contains M linear equations and O is a zero matrix.

[0019] (4) In the dynamic data coordination model Perform matrix whitening, turning the diagonal matrix Absorb and construct the whitening residual vector The above dynamic data coordination model is further transformed into:

[0020]

[0021]

[0022] (5) Solve the transformed dynamic data coordination model to obtain its unique least squares analytical solution and obtain the coordination value x′ of the measurement variable;

[0023] (6) Conduct online monitoring of the target industrial process system under the dynamic data coordination model to obtain the adjusted real-time operation dataset.

[0024] The further technical solution is as follows:

[0025] The solution to the transformed dynamic data coordination model yields its unique least-squares analytical solution, obtaining the coordinated value x′ of the measured variable, including:

[0026] The dynamic data reconciliation problem is equivalent to a non-homogeneous linear system of equations Az = b with respect to z, and the solution space is obtained...

[0027]

[0028] in:(·) + For Moore-Penrose generalized inverse, I1 is the AND... A unit matrix of the same dimension, where B is the data coordination and transformation matrix.

[0029] The online monitoring of the target industrial process system under the dynamic data coordination model, to obtain the adjusted real-time operating dataset, includes:

[0030] Determine the number of real-time samplings, i.e., the real-time sliding window length N, obtain measurement samples, and calculate the constraint equation matrix A. N and random error matrix ∑ N ;

[0031] Calculate the Moore-Penrose generalized inverse to obtain the data reconciliation transformation matrix B;

[0032] After rearranging the measurement samples into a vector, the data coordination value x′=By′ is obtained.

[0033] The true value of the measured variable Numerical simulation values ​​are used.

[0034] The target industrial process system is a high-pressure heater system.

[0035] The differential equality constraint f(x,u)=0 represents the mass and energy conservation equations for the high-pressure heater system under dynamic conditions.

[0036] Linearize nonlinear equations by replacing differential equations with difference equations.

[0037] The beneficial effects of this invention are as follows:

[0038] The dynamic data coordination method of the present invention takes into account both the operating mechanism of the industrial process system and the timeliness of the data, and realizes the reconstruction of the online operating data of the industrial process system to obtain high-precision data coordination values.

[0039] This invention has good interpretability, a simple architecture, high computational efficiency, and strong feasibility. It can adapt to changes in the operating conditions of thermal power units. Based on effectively characterizing the thermodynamic characteristics of high-pressure heaters, it meets the high-precision data requirements for online monitoring of high-pressure heaters. It can improve the accuracy of high-pressure heater measurement data in real time, accurately and significantly, and provide high-quality data for its dynamic performance monitoring.

[0040] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. Attached Figure Description

[0041] Figure 1 This is a flowchart of a high-pressure heater data coordination method based on dynamic data coordination, according to an embodiment of the present invention.

[0042] Figure 2 This is a simplified diagram of the thermodynamic process of a high-pressure heater based on mechanistic analysis.

[0043] Figure 3 This is a diagram showing the deviation before and after coordinating the mass conservation equation of the high-pressure heater in an embodiment of the present invention.

[0044] Figure 4 This is a diagram showing the deviation before and after coordinating the energy conservation equation of the high-pressure heater in an embodiment of the present invention.

[0045] Figure 5 This is a scatter plot comparing the measured, coordinated, and simulated values ​​of the feedwater flow rate of the high-pressure heater in an embodiment of the present invention. Detailed Implementation

[0046] The specific embodiments of the present invention are described below with reference to the accompanying drawings.

[0047] See Figure 1This embodiment of a method for online coordination of high-pressure heater data based on dynamic data coordination includes the following steps:

[0048] (1) Data acquisition and preprocessing:

[0049] The dataset Y was obtained by sampling 200 times for 15 measured variables, including extraction steam flow, pressure, feedwater flow, and high-pressure heater water level, in a certain time period using the thermal power unit operation database. Let y represent a measurement sample obtained from a certain sampling, and define the true value of the data corresponding to this sample as y. (Simulation data can be used instead) and random error ε. The random error ε ~ N(0,Σ) means that ε (a multidimensional continuous random variable) follows a multidimensional standard Gaussian distribution with a mean of 0 and a covariance of ∑.

[0050] σ1,σ2,σ3,…,σ 15 Let be the uncertainty corresponding to the measured value of each measured variable in the measured value sample y; define the conformity value x corresponding to each measured value in the measured value sample y.

[0051] Here, ∑ represents the covariance matrix of the random variable ε, one of the two key parameters of the multidimensional Gaussian distribution, characterizing the linear correlation between different sub-variables of the multidimensional random variable and the variance of each sub-variable. The elements on the diagonal of the covariance matrix represent the variance of each sub-variable, while the elements on the off-diagonal represent the covariance between different sub-variables. If the off-diagonal elements of the covariance matrix are 0, it indicates that the variables are independent of each other.

[0052] (2) Establishment of dynamic data coordination model for high-pressure heater:

[0053] Using the statistic R MSE To optimize the function, the data reconciliation problem is transformed into a constrained optimization mathematical model, namely:

[0054]

[0055]

[0056] Where: N is the number of sampling times, i.e., the number of samples, and i is the sampling time, taking the value t1t. N The superscripts T and -1 represent the transpose and inverse of the matrix, respectively; f(·) represents the differential equation and other constraints; g(·) represents the inequality constraints; and u represents the vector of unmeasured values.

[0057] For the high-pressure heater, the inequality constraints can be flexibly adjusted, while the equality constraints are the mass and energy conservation equations under dynamic conditions. Because the sampling interval is sufficiently small, difference equations are used to replace the differential equations in the constraint equations, thus linearizing the nonlinear equations. The constraint equations can be written as a system of linear equations A, containing M linear constraint equations:

[0058] Ax = O

[0059] in: O is a zero matrix.

[0060] Specifically, based on, for example Figure 2 The mechanism analysis shown below yields the following mass conservation equation (energy conservation follows similarly):

[0061]

[0062] In the formula: H lev The working fluid level is represented by T, and the measurement sampling time interval is T. The term represents the first derivative of the effective volume of the working fluid with respect to the horizontal liquid level elevation (due to the liquid level height H). lev For common measuring points, the volume change rate can be determined by H. lev (Calculated based on the equipment's geometric characteristics). Δq is the net flow difference between a certain inlet and outlet of the high-pressure heater, and ∑Δq is the overall net flow difference between the inlet and outlet of the high-pressure heater. and These are the total inlet flow rate and total outlet flow rate of the high-pressure heater. and For t respectively i The liquid density and liquid level height in the steam-water space inside the high-pressure heater are measured at all times. Similarly, For t i+1 The liquid level height at any given time.

[0063] The effective volume of the high-pressure heater can be simplified to a horizontal cylinder, and the relationship between the liquid level and the liquid volume is given by the following formula:

[0064]

[0065] In the formula: D is the diameter of the bottom surface of the horizontal cylinder.

[0066] (3) Dynamic coordination model data matrix and vector reshaping and augmentation:

[0067] The 200 samples in the measurement dataset Y are sorted first by the order of the measurement variables, and then by the order of the samples, and rearranged into y′. Specifically:

[0068] Suppose the measured variable is y = [y1, y2, ..., y]. p ]T Since a total of 200 samples y were sampled, then The i-th sample If we consider a row vector, then:

[0069]

[0070] Correspondingly, the coordination values ​​x are also rearranged in this order to x′, where x′ and y′ have the same dimension. The above constrained optimization mathematical model is then transformed into:

[0071]

[0072] stA N x′=O

[0073] Where: assuming that the random errors and constraint equations between samples have no mutual influence, then:

[0074] random error matrix

[0075] constraint equation matrix A t Let A be the constraint equation followed by the sample at time t, and let A be the constraint relationship between samples at different times. N The elements that are not on the diagonal of the block matrix determine the outcome.

[0076] (4) Matrix whitening and dynamic data reconciliation model transformation:

[0077] Item Perform matrix whitening, turning the diagonal matrix Absorb and construct the whitened residual vector z. The above constrained optimization mathematical model is further transformed into:

[0078]

[0079]

[0080] (5) Solve for the analytical solution of the dynamic data coordination model:

[0081] This problem is equivalent to finding the solution space of the non-homogeneous linear system of equations Az = b with respect to z. The analytical solution is From this we can obtain

[0082] in:(·) + For Moore-Penrose generalized inverse, I1 is the AND... An identity matrix of the same dimension, where B is the data coordination and transformation matrix.

[0083] (6) High-efficiency online monitoring of high-pressure heaters under a dynamic data coordination model:

[0084] Determine the real-time sliding window length T, that is, set the parameter N = T in the data coordination model;

[0085] Calculate the constraint equation matrix A N and random error matrix ∑ N ;

[0086] Calculate matrix operations and the generalized inverse to obtain the data reconciliation matrix B;

[0087] After rearranging the measurement samples into a vector, the data coordination value x′=By′ is obtained.

[0088] In actual operation, the sliding window length and constraint equations are finely adjusted to obtain high-precision and high-time-efficiency system operation data.

[0089] Specifically, when coordinating online data, the specific frequency and accuracy of the data coordination should be determined based on the actual level of big data informatization of the thermal power unit. If the required data coordination frequency and accuracy are high, and there is good hardware support, then N = T = 1, 3, or 5 (where the sampling interval of the samples is 1 second / sample). Conversely, if the system requirements are low and high data accuracy is not required for real-time monitoring, then N = T = 10, 30, or 60.

[0090] Specifically, the constraint equation system matrix A N The specific requirements need to be determined based on the actual high-pressure heater requirements of the thermal power unit. If the required data accuracy is low and each sample is independent, only the basic open system mass conservation equation and energy conservation equation are needed. If the required data accuracy is high, A can be adjusted based on the correlation between the samples. N (If the sample at this moment is related to the sample at a previous moment), and A can also be determined based on the conservation equations of the actual high-pressure heater (such as pressure conservation, water level conservation under the action of the control system, etc.). N Adjustments were made to obtain a high-precision sample set that conforms to the actual situation on site.

[0091] To verify the effectiveness of the method in this embodiment, we define the mass conservation deviation ΔQ and energy conservation deviation ΔE of the measured value to measure the degree of deviation of the measured value from the conservation formula. Based on the operating mechanism of the high-pressure heater (derived above), the formula for calculating ΔQ is as follows (ΔE is calculated similarly):

[0092]

[0093] Figure 3 and Figure 4The figures show the ΔQ and ΔE curves before and after data reconciliation (Z-score normalization has been performed for easier observation). As can be seen from the figures, the measured values ​​fluctuate around the line y = 0, indicating that when random errors are included in the measurements, the actual data deviates somewhat from the theoretical conservation equation.

[0094] After data coordination processing, the coordinated values ​​can obey the conservation equation, and their deviations ΔQ and ΔE both tend to 0.

[0095] Figure 5 A scatter plot comparing the measured, coordinated, and simulated values ​​of feedwater flow rate for high-pressure heaters is presented. Feedwater flow rate is a crucial measurement point for thermal power units, but the measurement signal often fluctuates, resulting in very low measurement accuracy, sometimes reaching 10%–15% of the measured value. After dynamic data coordination calculation, 200 samples of the measured, coordinated, and simulated feedwater flow rates are displayed as shown in the figure. It is evident that random noise significantly impacts the measured values, causing a large deviation between the measured and simulated values. The dynamic data coordination method proposed in this paper can largely suppress measurement noise and approximate the true value. Furthermore, it boasts high computational efficiency, providing a real-time and efficient dataset for online monitoring of thermal power units.

[0096] In summary, the method of the present invention can significantly improve the accuracy of high-pressure heater measurement data in real time when facing the dynamic operation of thermal power units, providing high-quality data for their dynamic performance monitoring.

[0097] It will be understood by those skilled in the art that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for online data coordination in industrial process systems based on dynamic data coordination, characterized in that, include: (1) Select the measured variable and the unmeasured variable u from the target industrial process system, and collect the dataset of the measured variable. Where y represents the sample of measured values ​​of the measured variable, with subscripts t1 to t2. N Represents the sampling time; defines the true value of the measured variable. and random error ε, Define the coordination value x corresponding to y. p represents the number of measurement variables; Wherein, ε follows a pattern with a mean of 0 and a covariance of . A multidimensional standard Gaussian distribution, σ1,σ2,σ3…σ p This represents the uncertainty corresponding to each measured value in the sample y. (2) Based on the dynamic model of the equipment in the target industrial process system, establish a set of constraint equations for the coordination value x and the unmeasured variable u; A dynamic data reconciliation model is constructed using measured sample y, reconciliation value x, and a set of constraint equations: Among them, R MSE Let f(·) be the objective function, N be the number of sampling times (i.e., the number of samples), i be the sampling time, and T and -1 be the transpose and inverse of the matrix, respectively; f(·) is the system of differential equations; g(·) is the system of algebraic equations of inequalities. (3) The dataset The data in the sample is first sorted by the order of the measured variables, then by the order of the samples, and then rearranged. Correspondingly, the coordination values ​​x are also rearranged in the same order to form x′, where x′ and y′ have the same dimension. This transforms the above dynamic data coordination model into: stANx′=O Wherein, the random error matrix ∑ t Let be the covariance matrix of the uncertainty of the sample measurement at time t; constraint equation matrix A t Let be the set of constraint equations followed by the measured values ​​at time t, which contains M linear equations and O is a zero matrix. (4) In the dynamic data coordination model Perform matrix whitening, turning the diagonal matrix Absorb and construct the whitening residual vector The above dynamic data coordination model is further transformed into: (5) Solve the transformed dynamic data coordination model to obtain its unique least squares analytical solution and obtain the coordination value x′ of the measurement variable; (6) Conduct online monitoring of the target industrial process system under the dynamic data coordination model to obtain the adjusted real-time operation dataset.

2. The online data coordination method for industrial process systems based on dynamic data coordination according to claim 1, characterized in that, The solution to the transformed dynamic data coordination model yields its unique least-squares analytical solution, obtaining the coordinated value x′ of the measured variable, including: The dynamic data reconciliation problem is equivalent to a non-homogeneous linear system of equations Az = b with respect to z, and the solution space is obtained... The minimum least squares analytical solution is Therefore, we can conclude that: in:(·) + For Moore-Penrose generalized inverse, I1 is the AND... A unit matrix of the same dimension, where B is the data coordination and transformation matrix.

3. The online data coordination method for industrial process systems based on dynamic data coordination according to claim 2, characterized in that, The online monitoring of the target industrial process system under the dynamic data coordination model to obtain the adjusted real-time operating dataset includes: Determine the number of real-time samplings, i.e., the real-time sliding window length N, obtain measurement samples, and calculate the constraint equation matrix A. N and random error matrix ∑ N ; Calculate the Moore-Penrose generalized inverse to obtain the data reconciliation transformation matrix B; After rearranging the measurement samples into a vector, the data coordination value x′=By′ is obtained.

4. The online data coordination method for industrial process systems based on dynamic data coordination according to claim 1, characterized in that, The true value of the measured variable Numerical simulation values ​​are used.

5. The online data coordination method for industrial process systems based on dynamic data coordination according to claim 1, characterized in that, The target industrial process system is a high-pressure heater system.

6. The online data coordination method for industrial process systems based on dynamic data coordination according to claim 5, characterized in that, The differential equality constraint f(x,u)=0 is the mass and energy conservation equation for the high-pressure heater system under dynamic conditions.

7. The online data coordination method for industrial process systems based on dynamic data coordination according to claim 6, characterized in that, Linearize nonlinear equations by replacing differential equations with difference equations.

Images