CN105700358B - A kind of modeling quality control method of the model predictive controller of band drift interference - Google Patents
A kind of modeling quality control method of the model predictive controller of band drift interference Download PDFInfo
- Publication number
- CN105700358B CN105700358B CN201610144819.4A CN201610144819A CN105700358B CN 105700358 B CN105700358 B CN 105700358B CN 201610144819 A CN201610144819 A CN 201610144819A CN 105700358 B CN105700358 B CN 105700358B
- Authority
- CN
- China
- Prior art keywords
- control system
- closed
- loop control
- model
- interference
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
- 238000003908 quality control method Methods 0.000 title abstract 2
- 238000000034 method Methods 0.000 claims abstract description 185
- 230000008569 process Effects 0.000 claims abstract description 157
- 239000011159 matrix material Substances 0.000 claims description 50
- 238000012544 monitoring process Methods 0.000 claims description 30
- 238000004519 manufacturing process Methods 0.000 claims description 12
- 238000005070 sampling Methods 0.000 claims description 11
- 238000004886 process control Methods 0.000 claims description 6
- 230000008859 change Effects 0.000 claims description 5
- 238000000354 decomposition reaction Methods 0.000 claims description 3
- 238000011156 evaluation Methods 0.000 abstract description 5
- 238000012545 processing Methods 0.000 abstract description 3
- 238000012938 design process Methods 0.000 abstract 1
- 238000004088 simulation Methods 0.000 description 7
- 230000000694 effects Effects 0.000 description 4
- 239000007791 liquid phase Substances 0.000 description 4
- 238000010992 reflux Methods 0.000 description 4
- 230000006866 deterioration Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000013404 process transfer Methods 0.000 description 3
- 230000006872 improvement Effects 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 238000013142 basic testing Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000005284 excitation Effects 0.000 description 1
- 238000009776 industrial production Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
- G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
- G05B13/04—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators
- G05B13/042—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators in which a parameter or coefficient is automatically adjusted to optimise the performance
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Feedback Control In General (AREA)
Abstract
The invention discloses a kind of modeling quality control methods of the model predictive controller of band drift interference, include the following steps：Establish the interference model of closed-loop control system；According to the actual conditions of closed-loop control system and given control targe, the dynamic model MPC controller of design process；Using interference model and MPC controller control closed-loop control system, and acquire the process data of closed-loop control system operation gained；According to closed-loop control system structure, rectangular projection is carried out to the output of process and process input data, obtains process estimation interference update；According to closed-loop control system given reference signal and process reality output, the actual tracking error of closed-loop control system is obtained；Interference update and the actual tracking error are estimated according to the process, obtain the model quality index of closed-loop control system；According to closed-loop control system structure, modeling quality is monitored using the model quality index；With feasibility height, processing consumes the characteristics of resource is few, and evaluation result accuracy is high.
Description
Technical Field
The invention belongs to the field of model predictive control, and particularly relates to a modeling quality monitoring method of a model predictive controller with drift interference.
Background
Model Predictive Control (MPC) is an advanced Control method based on a Model widely applied to the field of industrial process Control, and has the advantages of good Control effect, strong robustness and low requirement on Model accuracy.
The practical application of model predictive control to industrial processes is called model predictive Controller (MPC Controller). The MPC controller has the characteristics of simple modeling, good dynamic control effect and strong robustness, and has good control performance at the initial stage of production; however, over time, the MPC controller will gradually degrade and eventually even have to switch to conventional PID control. The main factors that cause the performance degradation of the controller are noise interference, model mismatch, valve sticking, sensor bias, etc. Therefore, the method has important practical significance for ensuring safe, efficient, high-quality and low-consumption operation of the production process.
In recent years, data-driven based approaches have been increasingly applied to control system performance evaluation issues. For example, the performance of the MPC can be effectively evaluated according to historical performance indexes under an MPC framework, but a section of data with good operation of a control system needs to be acquired to calculate an evaluation standard, and the selection of the good operation stage has no standard, so that certain limitation is brought to the application of the method. In the model-based control technology, the quality of a model plays a key role in the design and setting of a controller, and the performance of a control system depends on the precision of a process model, namely, is influenced by the mismatch degree of the model.
On one hand, white noise is mostly adopted as a system interference noise source in the technology about modeling quality monitoring at the present stage, but in the actual industrial process, the interference is usually gradually increased along with time, and the characteristic of non-Gaussian appearance is presented; on the other hand, the technology related to modeling quality monitoring at the present stage cannot diagnose the deterioration source of the performance deterioration of the controller, and the reason for the performance deterioration of the controller cannot be diagnosed is that the model has mismatch or factors such as noise interference, valve viscosity and sensor deviation.
Disclosure of Invention
In view of the above defects or improvement requirements of the prior art, the invention provides a modeling quality monitoring method for a model predictive controller with drift interference, which aims to effectively monitor the quality of a model by only utilizing conventional input and output data of a closed-loop control system, thereby improving the accuracy of monitoring the modeling quality.
To achieve the above object, according to an aspect of the present invention, there is provided a modeling quality monitoring method for a model predictive controller with drift interference, including the steps of:
(1) establishing an interference model of a closed-loop control system;
(2) acquiring an MPC controller of a process according to the actual condition of a closed-loop control system and a given control target;
(3) when the closed-loop control system operates under the control of the interference model and the MPC controller, acquiring process data obtained by the operation of the closed-loop control system; the process data comprises process outputs and process inputs of a closed loop control system;
(4) performing orthogonal projection on the process output data and the process input data according to a closed-loop control system structure to obtain process estimation interference update;
(5) acquiring an actual tracking error of the closed-loop control system according to the established reference signal of the closed-loop control system and the actual process output;
(6) obtaining a model quality index of a closed-loop control system according to the process estimation interference update and the actual tracking error;
(7) and monitoring the modeling quality by using the model quality index according to the structure of the closed-loop control system.
Preferably, the step (1) includes the following substeps:
(1.1) establishing a first-order moving average model of the industrial process according to the structure of a closed-loop control system, which comprises the following specific steps:
dk＝dk-1+εk-θεk-1
wherein theta is a white noise average coefficient, and theta is more than-1 and less than 1; epsilonk～N(0,σε 2) Representing white noise; sigmaε 2Represents a white noise variance; dkRepresenting process random interference noise;
(1.2) adding drift to the first-order moving average model to obtain a first-order moving average model with drift interference, wherein the first-order moving average model with drift interference is as follows:
dk＝dk-1+εk-θεk-1+δ
where δ is the drift, dkRepresenting process random interference noise.
Preferably, the step (2) includes the following substeps:
(2.1) defining related controlled variable CV, manipulated variable MV and disturbance variable DV according to a given control target; where DV refers to a measurable disturbance that has an effect on CV, but is not manipulable;
(2.2) step changing the manipulated variable MV and the disturbance variable DV to obtain dynamic change data of each controlled variable CV, and acquiring a dynamic model predictive controller of the process by using an identification algorithm (2.2); and configures the controller using the parameter selection rules.
Preferably, the step (3) includes the following substeps:
(3.1) generating a process set value of the closed-loop control system according to the requirement of the control system, wherein the set value is a constant and is recorded as r (k), and k represents the kth sampling moment;
and (3.2) operating the closed-loop control system, and acquiring process input u (k) and process output y (k) of the closed-loop control system.
Preferably, the step (4) includes the following substeps:
(4.1) according to the structure of the closed-loop control system, establishing a high-order autoregressive model output by the process as follows:
yp(k)＝[y(k) y(k-1) … y(k-p)]
wherein, p represents the size of a data window for estimating interference update by the process, and M represents the order of the process output high-order autoregressive model; y (k) represents the process output obtained by the acquisition at the time k, and y (k-1) represents the acquisition at the time (k-1)Collecting the obtained process output, …, y (k-p) represents the process output obtained by collecting the (k-p) time, yp(k) Represents a 1 x (P +1) -dimensional matrix composed of y (k), y (k-1), …, y (k-P); y isM(k-1) represents by yp(k-1)，yp(k-2)，…，yp(k-M) forming an M x (P +1) -dimensional matrix;
(4.2) according to the structure of the closed-loop control system, establishing a high-order autoregressive model of the process set value as follows:
rp(k)＝[r(k) r(k-1) … r(k-p)]
wherein, p represents the size of a data window for process estimation interference update, and N represents the order of a process set value high-order autoregressive model;
r (k) represents the process set value acquired at the time k, r (k-1) represents the process set value acquired at the time (k-1), …, r (k-p) represents the process set value acquired at the time (k-p)p(k) Represents a 1 x (P +1) -dimensional matrix composed of r (k), r (k-1), … and r (k-P); rN(k-1) represents a group represented by rp(k-1)，rp(k-2)，…，rp(k-N) forming an N x (P +1) -dimensional matrix;
(4.3) according to the structure of the closed-loop control system, establishing a mixed high-order autoregressive model of the process output and the process set value as follows:
wherein,a high-order autoregressive model Y which is a matrix of (M + N) × (P +1) dimensions and is output by the process in the first M rowsM(k-1) and the last N rows are formed by a high-order autoregressive model R of the process set pointN(k-1);
(4.4) acquiring a process estimation interference update vector through an orthogonal projection algorithm according to the high-order autoregressive model:
wherein e isp(k)＝[e(k) e(k-1) … e(k-p)]The estimated process interference update vector of 1 x (P +1) dimension is shown, e (k) shows the estimated process interference update obtained at the time of k, and I shows an identity matrix of (P +1) x (P +1) dimension;
since the closed-loop data has a high correlation, it is possible to detect the correlation between the data and the dataAre ill-conditioned, resulting in unreliable interference updates for the resulting process estimate.
Preferably, the method of step (4.4) to solve this problem further comprises the sub-steps of:
(4.4.1) Pair matrixAs QR decomposition, i.e.Obtaining an orthogonal matrix Q1And Q2Diagonal matrix R11Diagonal matrix R22Sum row vector R21；
(4.4.2) according to the orthogonal matrix Q1And Q2Diagonal matrix R11Diagonal matrix R22Sum row vector R21The following results were obtained:
(4.4.3) obtaining a reliable process estimate interference update from the results obtained in (4.4.2):
preferably, the step (5) is as follows:
acquiring an actual tracking error of the closed-loop control system according to a set reference signal (or a control signal) of the closed-loop control system and the actual process output:
wherein,the actual tracking error of the closed-loop control system at the time k is shown, r (k) shows a set reference signal at the time k, and y (k) shows the actual output of the process at the time k.
Preferably, the step (6) includes the following sub-steps:
(6.1) defining key performance indicators KPI according to the output weight matrix Q and the input weight matrix S:
wherein, JNKey performance indicators KPI are represented, N represents the length of sampling data required by the index, y (k) represents the actual output of the process at the time k, r (k) represents a set reference signal at the time k, and delta u*(k) U (k) -u (k-1) represents the difference between the time k process control input and the time (k-1) process control input;
(6.2) estimating interference update e (k) according to the process obtained in the step (4) and the actual tracking error obtained in the step (5)Obtaining a model quality index MQI of a closed-loop control system:
wherein N represents the length of sampling data required for obtaining the index,and e (k) represents the process estimation interference update of the closed-loop control system at the moment k, and Q represents an output weight matrix.
Preferably, the step (7) is as follows:
evaluating the performance of the controller according to the key performance index KPI of the closed-loop control system, wherein the smaller the KPI value is, the better the performance of the controller is, evaluating the modeling quality according to the model quality index MQI of the closed-loop control system, wherein the value range of the MQI is η E (0, 1), the closer the model quality index η is to 1, the better the modeling quality of the closed-loop control system is, the closer the model quality index is to 0, and the worse the modeling quality of the closed-loop control system is.
In general, compared with the prior art, the above technical solution contemplated by the present invention can achieve the following beneficial effects:
(1) according to the modeling quality monitoring method provided by the invention, the conventional closed loop data of the current closed loop control system is adopted to estimate the process interference update, and additional information of an open loop process which is difficult to obtain, such as process time lag or an interaction matrix, is not needed, so that the method is high in feasibility and consumes few resources for processing;
(2) in the modeling quality monitoring method provided by the invention, in order to describe the actual interference situation of the industrial process, the interference noise is described by adopting the IMA (1,1) model with drift, so that the actual situation of the industrial process is better met, and the evaluation result of the modeling quality is more accurate;
(3) the modeling quality monitoring method provided by the invention does not need to adjust the current closed-loop control system in the processing process and add any external excitation signal, so that the influence on the industrial production process is extremely small, the monitoring cost is greatly reduced, and the product quality and the system safety and maintainability in the industrial process are improved;
(4) the modeling quality monitoring method provided by the invention provides a model quality index MQI by using a feedback invariance principle of a disturbance sequence, and can find whether the model in the control system is mismatched or not in time according to the index so as to realize real-time evaluation on the performance of the control system; and because the relation between the process estimation interference update and the process actual tracking error, the model quality index MQI is not influenced by the change of the controller adjusting parameters and the change of the closed-loop control system interference model, compared with the traditional KPI index, the MQI index can more accurately distinguish the factors which can influence the overall control performance of the controller except the model quality, thereby more accurately and effectively representing the modeling quality.
Drawings
FIG. 1 is a schematic overall flow chart of a model quality monitoring method according to an embodiment of the present invention;
FIG. 2 is a schematic structural diagram of a closed loop system in an embodiment;
FIG. 3 is a schematic view of a Wood-Berry rectifying column in the example;
FIG. 4 is a graph of simulation results of estimating an interference noise sequence based on a process obtained by the Wood-Berry rectifying tower in the example;
FIG. 5 is a diagram showing simulation results of a model residual sequence based on the Wood-Berry rectification tower in the example;
FIG. 6 is a graph showing the simulation results of the model output y1 and the set value r1 based on the Wood-Berry rectifying tower in the embodiment;
FIG. 7 is a graph showing the simulation results of the model output y2 and the set value r2 based on the Wood-Berry rectifying tower in the example.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. In addition, the technical features involved in the embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.
The modeling quality monitoring method of the model predictive controller with drift interference provided by the embodiment of the invention has the flow shown in fig. 1, and specifically comprises the following steps:
(1) establishing an interference model of a closed-loop control system according to the actual interference situation of the industrial process;
(2) designing an MPC controller according to the actual condition of a closed-loop control system;
(3) controlling a closed-loop control system according to the interference model of the closed-loop control system and the MPC controller, and acquiring process output and process input data obtained by the operation of the closed-loop control system;
(4) performing orthogonal projection on the obtained process output and process input data according to the structure of a closed-loop control system to obtain process estimation interference update;
(5) acquiring an actual tracking error of the closed-loop control system according to a set reference signal (or a control signal) and the actual process output of the closed-loop control system;
(6) estimating interference update and the actual tracking error according to the process of the closed-loop control system to obtain a model quality index of the closed-loop control system;
(7) monitoring the modeling quality by using the model quality index according to the structure of a closed-loop control system;
evaluating the performance of the controller according to the key performance index KPI of the obtained closed-loop control system, wherein the smaller the KPI value is, the better the performance of the controller is, evaluating the modeling quality according to the model quality index MQI of the obtained closed-loop control system, wherein the value range of the MQI is η E (0, 1), the closer the model quality index η is to 1, the better the modeling quality of the closed-loop control system is, the closer the model quality index is to 0, and the worse the modeling quality of the closed-loop control system is.
Fig. 2 is a structural diagram of a closed-loop control system adopted in the embodiment of the present invention, in which u (t), y (t) respectively represent process input and process output of the closed-loop control system, r (t) represents a process set value of the closed-loop control system, and d (t) represents external interference of the closed-loop control system; gcWhich is the transfer function of the controller of a typical closed loop control system, i.e., an MPC controller in the example, G is the process transfer function.
The modeling quality monitoring method of the model predictive controller with drift interference provided by the invention is further explained by taking the Wood-Berry rectifying tower process as an example.
In the embodiment, the Wood-Berry rectifying tower is a typical multi-input multi-output system with large pure lag; the process is shown in FIG. 3, with the output being the overhead concentration XDAnd the liquid phase concentration X at the bottom of the columnBThe reflux quantity R at the top of the tower and the steam quantity S of a reboiler at the bottom of the tower are used for controlling; the process model is as follows:
wherein, input u1Represents the reflux rate, and has the unit of 1 b/min; input u2Represents the steam flow rate, and the unit is 1 b/min; output y1Represents the overhead concentration in mol%; output y2Represents the concentration of the liquid phase at the bottom of the column in mol%.
In the examples, the process transfer function matrix of the Wood-Berry rectification column is:
where s represents the laplace operator.
In the embodiment, the process sampling time is 1 min/time, and the process transfer function matrix after discretization is as follows:
in the embodiment, the interference model takes the following diagonal matrix:
the modeling quality monitoring method of the model predictive controller with drift interference provided by the invention is used for carrying out the modeling quality monitoring process on the Wood-Berry rectifying tower process of the embodiment, and the method comprises the following specific steps:
(1) establishing an interference model of a closed-loop control system according to the actual interference situation of the industrial process,
the interference model for the Wood-Berry rectification column process is as follows:
it is expressed in the following general form:
a first order moving average model with drifting interference (IMA (1,1)) is represented as follows:
dk＝dk-1+εk-θεk-1+δ
where δ represents the drift, dkRepresenting process random interference noise;
in the formula,ek～N(δ/(1-θ),σε 2). When δ is equal to 0.005 and θ is equal to 0.5, the process is repeatedIn the examples, σ2Are respectively 7.62And 0.142；dkNamely an interference model adopted in the Wood-Berry rectification tower process, wherein k is 1,2, … … and 500.
(2) Designing a Model Predictive Controller (MPC) according to the actual situation of a closed-loop control system, wherein in the embodiment, an MPC controller is designed by utilizing an existing MPC Toolbox in an MATLAB Toolbox; the MPC controller parameter takes the prediction time domain P as 100, the control time domain M as 10, and the index required sampling data length as N as 500. To implement the minimum variance criterion, the weight matrices are set to Q ═ diag {1,100}, respectively, and S ═ 0; the process set value is:
(3) controlling a closed-loop control system according to the interference model of the closed-loop control system and the MPC controller, and acquiring process output and process input data obtained by the operation of the closed-loop control system;
when the closed-loop control system normally operates, the sampling time is set to be 1 min/time, and process output data obtained by the operation of the Wood-Berry rectifying tower of the closed-loop control system is collected, wherein the process output data comprises the concentration X of distillate at the tower topDAnd the liquid phase concentration X at the bottom of the columnBAre respectively denoted by y1And y2(ii) a Process input data, including the top reflux R and the bottoms reboiler steam S, are recorded as u1And u2(ii) a The number of samples N collected was set to 500.
(4) According to the closed-loop control system structure, orthogonal projection is carried out on the obtained process output and process input data, and process estimation interference update is obtained:
according to the process output and process input data obtained by the operation of the closed-loop control system, a high-order autoregressive model of the process output is established as follows:
yp(k)＝[y(k) y(k-1) … y(k-p)]
the high-order autoregressive model of the set value is established as follows:
rp(k)＝[r(k) r(k-1) … r(k-p)]
wherein, the process set value r (k) is a constant, and in the embodiment, the process set value r (k) is set as:p represents the size of a data window for process estimation interference updating, M represents the order of a process output high-order autoregressive model, and N represents the order of a process set value high-order autoregressive model;
establishing a mixed high-order autoregressive model of process output and a process set value (or process input), which is specifically as follows:
for matrixAs QR decomposition, i.e.Obtaining an orthogonal matrix Q1And Q2Diagonal matrix R11Diagonal matrix R22Sum row vector R21；
Obtaining a process estimation interference update vector through an orthogonal projection algorithm:
wherein e isp(k)＝[e(k) e(k-1) … e(k-p)]The estimated process interference update vector of 1 x (P +1) dimension is shown, e (k) shows the estimated process interference update obtained at the time of k, and I shows an identity matrix of (P +1) x (P +1) dimension;
according to an orthogonal matrix Q1And Q2Diagonal matrix R11Diagonal matrix R22Sum row vector R21The following results were obtained:
obtaining a final reliable process estimate interference update:
in the embodiment, k is 1,2, … …, and 500, the number N of samples to be collected is 500, the prediction time domain P is 100, and the control time domain M is 10.
(5) Acquiring an actual tracking error of the closed-loop control system according to a set reference signal (or a control signal) of the closed-loop control system and the actual process output:
according to the structure of the closed loop system, the tracking error of the system is obtained:
wherein,the actual tracking error of the closed-loop control system at time k is shown, r (k) shows the established reference signal at time k, y (k) shows the actual output of the process at time k, and k is 1,2, … …, 500.
(6) Obtaining a model quality index of the closed-loop control system according to the process estimation interference update of the closed-loop control system and the actual tracking error
Defining key performance indicators KPI according to the output weight matrix Q and the input weight matrix S:
wherein, JNThe key performance index KPI is represented, N represents the length of sampling data required by the index, and is set to be 500; y (k) represents the actual output of the process at time k: concentration X of the overhead distillateDAnd the liquid phase concentration X at the bottom of the columnBAre respectively denoted by y1And y2(ii) a The process setpoint r (k) is a constant set to: Δu*(k) u (k) -u (k-1) represents the difference between the process control input at time k and the process control input at time k-1, the output weight matrix Q is diag {1,100}, and the input weight matrix S is 0.
Updating e (k) and actual tracking error based on process estimated interferenceObtaining a model quality index MQI of a closed-loop control system:
wherein, N is a sampling data length required for obtaining the index, and is 500 in the embodiment;and e (k) represents the actual tracking error of the closed-loop control system at the moment k, and Q represents an output weight matrix, wherein Q is diag {1,100 }.
(7) Monitoring the modeling quality by using the model quality index according to the structure of the closed-loop control system; in the examples, the validity of the model quality indicator MQI is verified by the following three different scenarios.
The first situation is as follows: process model matching, but interference model mismatch:
let K1 be 0.5, K2 be 0.3, and when the simulation process model and the interference model are both matched, obtain the corresponding KPI index value of 450.3286 and the MQI index value of 0.9993 by calculating formula (1) and formula (2); by setting different values of K1 and K2, KPI indicators and MQI indicators under the condition that the interference model has different degree of mismatch are obtained, as shown in table 1:
table 1: parameter list under process model matching and interference model mismatching conditions
As can be seen from table 1, the first row results are obtained when both CV variables are white noise interference plus drift, and because a more matched interference model is used, the MQI and KPI indicators in the second, third, and fourth row results are all improved over those in the first row.
The simulation results are shown in FIGS. 4-7; as can be gathered from fig. 4, the estimated interference noise sequence is close to but not completely white noise, and has a small drift distributed as a whole. Fig. 5 is a model residual sequence, fig. 6 and fig. 7 are model output and set values, which can be analyzed from fig. 6 and fig. 7, the control effect of the model is relatively ideal, the set value can be tracked as a whole, and the process tends to be stable.
Case two: the process model and the disturbance model both match the situation, but the controller parameters are different:
to verify that the MQI index is not affected by the controller parameters, and that the KPI index has this deficiency, three sets of comparative tests can be performed with both the process model and the disturbance model matched, as follows:
(i) basic test, i.e. case one where K1 ═ 0.5, K2 ═ 0.3;
(ii) the input weight matrix is set to S ═ diag {0.2,0.6 };
(III) reflux Rate i.e. input u1Is defined as u1≤151b/min。
The simulation test results are as follows:
table 2: controller parameter list under condition that process model and interference model are matched
The results obtained from table 2 can be analyzed, and when the process model and the interference model are both matched and the controller parameters are changed, the KPI index is correspondingly changed, but the MQI index is kept unchanged; therefore, the MQI index can be verified to accurately distinguish factors which can influence the overall control performance of the controller and are beyond the quality of the model.
Case three: mismatch of both process and interference models:
when there is a mismatch in the process model, it is assumed that the steady state gain of the process model in MPC contains a 20-40% bias, as follows:
by setting different values of K1 and K2, the interference model has mismatch conditions of different degrees, and the results shown in Table 3 are obtained.
Table 3: parameter list of process model and interference model mismatch
The results obtained from table 3 can be analyzed, and when the mismatch condition of the process model is fixed, the interference models of different mismatch conditions are obtained by setting different kalman gains, and the parameters of the controller are not changed, the MQI index is correspondingly increased along with the decreasing of the KPI index, so that the MQI index can effectively represent the modeling quality.
It will be understood by those skilled in the art that the foregoing is only a preferred embodiment of the present invention, and is not intended to limit the invention, and that any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the scope of the present invention.
Claims (9)
1. A modeling quality monitoring method of a model predictive controller with drift interference is characterized by comprising the following steps:
(1) establishing an interference model of a closed-loop control system;
(2) acquiring an MPC controller of a process according to parameters of a closed-loop control system and a given control target;
(3) when the closed-loop control system operates under the control of the interference model and the MPC controller, acquiring process data obtained by the operation of the closed-loop control system; the process data comprises process outputs and process inputs of a closed loop control system;
(4) performing orthogonal projection on the process output data and the process input data according to a closed-loop control system structure to obtain process estimation interference update;
(5) acquiring an actual tracking error of the closed-loop control system according to the established reference signal of the closed-loop control system and the actual process output;
(6) obtaining a model quality index of a closed-loop control system according to the process estimation interference update and the actual tracking error;
(7) and monitoring the modeling quality by using the model quality index according to the structure of the closed-loop control system.
2. The modeling quality monitoring method according to claim 1, wherein the step (1) includes the substeps of:
(1.1) establishing a first-order moving average model of the industrial process according to the structure of a closed-loop control system, which comprises the following specific steps:
dk＝dk-1+εk-θεk-1
wherein theta is a white noise average coefficient, and theta is more than-1 and less than 1; epsilonk～N(0,σε 2) Representing white noise; sigmaε 2Represents a white noise variance; dkRepresenting process random interference noise;
(1.2) adding drift to the first-order moving average model to obtain a first-order moving average model with drift interference, wherein the first-order moving average model with drift interference is as follows:
dk＝dk-1+εk-θεk-1+δ
where δ is the drift, dkRepresenting process random interference noise.
3. A modeling quality monitoring method according to claim 1 or 2, characterized in that said step (2) comprises the sub-steps of:
(2.1) defining a controlled variable CV, a manipulated variable MV and a disturbance variable DV according to a given control target;
(2.2) obtaining dynamic change data of each controlled variable CV by performing step change on the manipulated variable MV and the disturbance variable DV, and obtaining an MPC controller of the process by using an identification algorithm; and configures the controller using the parameter selection rules.
4. A modeling quality monitoring method according to claim 1 or 2, characterized in that said step (3) comprises the sub-steps of:
(3.1) generating a process set value r (k) of a closed-loop control system according to the requirement of the control system; the set value is a constant, and k represents the kth sampling moment;
and (3.2) operating the closed-loop control system, and acquiring process input u (k) and process output y (k) of the closed-loop control system.
5. A modeling quality monitoring method according to claim 1 or 2, characterized in that said step (4) comprises the sub-steps of:
(4.1) according to the structure of the closed-loop control system, establishing a high-order autoregressive model output by the process as follows:
yp(k)＝[y(k) y(k-1) … y(k-p)]
wherein, p represents the size of a data window for estimating interference update by the process, and M represents the order of the process output high-order autoregressive model; y (k) represents the process output acquired at the time k, y (k-1) represents the process output acquired at the time k-1, …, y (k-p) represents the process output acquired at the time (k-p)p(k) Represents a 1 x (p +1) -dimensional matrix composed of y (k), y (k-1), …, y (k-p); y isM(k-1) represents by yp(k-1)，yp(k-2)，…，yp(k-M) forming an M x (p +1) -dimensional matrix;
(4.2) according to the structure of the closed-loop control system, establishing a high-order autoregressive model of the process set value as follows:
rp(k)＝[r(k) r(k-1) … r(k-p)]
wherein N represents the order of the high-order autoregressive model of the process set value;
r (k) represents the process set value acquired at the time k, r (k-1) represents the process set value acquired at the time (k-1), …, r (k-p) represents the process set value acquired at the time (k-p)p(k) Represents a 1 x (p +1) -dimensional matrix composed of r (k), r (k-1), …, r (k-p); rN(k-1) represents a group represented by rp(k-1)，rp(k-2)，…，rp(k-N) forming an N x (p +1) -dimensional matrix;
(4.3) according to the structure of the closed-loop control system, establishing a mixed high-order autoregressive model of the process output and the process set value as follows:
wherein,dimensional matrix, first M rows of high order autoregressive model Y output by processM(k-1), wherein the last N rows consist of a high-order autoregressive model RN (k-1) of the process set value;
(4.4) acquiring a process estimation interference update vector through an orthogonal projection algorithm according to the high-order autoregressive model:
wherein e isp(k)＝[e(k) e(k-1) … e(k-p)]The estimated interference update vector of the process of dimension 1 × (p +1) is shown, e (k) shows the estimated interference update of the process obtained at the time k, and I shows the identity matrix of dimension (p +1) × (p + 1).
6. A method of monitoring modelling quality according to claim 5, wherein said step (4.4) comprises the sub-steps of:
(4.4.1) Pair matrixAs QR decomposition, i.e.Obtaining an orthogonal matrix Q1And Q2Diagonal matrix R11Diagonal matrix R22Sum row vector R21；
(4.4.2) according to the orthogonal matrix Q1And Q2Diagonal matrix R11Diagonal matrix R22Sum row vector R21The following results were obtained:
(4.4.3) obtaining a reliable process estimate interference update from the results obtained in (4.4.2):
7. a modeling quality monitoring method according to claim 1 or 2, characterized in that said step (5) is specifically as follows:
acquiring an actual tracking error of the closed-loop control system according to the established reference signal and the actual process output of the closed-loop control system:
wherein,the actual tracking error of the closed-loop control system at the time k is shown, r (k) shows a set reference signal at the time k, and y (k) shows the actual output of the process at the time k.
8. The modeling quality monitoring method according to claim 7, wherein the step (6) includes the substeps of:
(6.1) defining key performance indicators KPI according to the output weight matrix Q and the input weight matrix S:
wherein, JNKey performance indicators KPI are represented, N represents the length of sampling data required by the index, y (k) represents the actual output of the process at the time k, r (k) represents a set reference signal at the time k, and delta u*(k) U (k) -u (k-1) represents the difference between the time k process control input and the time (k-1) process control input;
(6.2) estimating interference update e (k) and the actual tracking error according to the process obtained in the step (4)Obtaining a model quality index MQI of a closed-loop control system:
wherein N represents the length of sampling data required for obtaining the index,and e (k) represents the process estimation interference update of the closed-loop control system at the moment k, and Q represents an output weight matrix.
9. The modeling quality monitoring method according to claim 8, wherein the step (7) is specifically as follows:
evaluating the performance of the controller according to key performance indicators KPI of the closed-loop control system: the smaller the KPI value is, the better the performance of the controller is represented;
and evaluating the modeling quality according to the model quality index MQI of the closed-loop control system, wherein the value range of the MQI is η E (0, 1), the closer the model quality index is to 1, the better the modeling quality of the closed-loop control system is indicated, and the closer the model quality index is to 0, the worse the modeling quality of the closed-loop control system is indicated.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610144819.4A CN105700358B (en) | 2016-03-14 | 2016-03-14 | A kind of modeling quality control method of the model predictive controller of band drift interference |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610144819.4A CN105700358B (en) | 2016-03-14 | 2016-03-14 | A kind of modeling quality control method of the model predictive controller of band drift interference |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105700358A CN105700358A (en) | 2016-06-22 |
| CN105700358B true CN105700358B (en) | 2018-06-12 |
Family
ID=56221642
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610144819.4A Expired - Fee Related CN105700358B (en) | 2016-03-14 | 2016-03-14 | A kind of modeling quality control method of the model predictive controller of band drift interference |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105700358B (en) |
Families Citing this family (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106843111B (en) * | 2017-03-10 | 2019-04-05 | 中国石油大学(北京) | The accurate source tracing method of hydrocarbon production system alarm signal root primordium and device |
| CN107272640A (en) * | 2017-06-12 | 2017-10-20 | 华中科技大学 | A kind of modeling quality control method and system based on model predictive controller |
| CN108536127B (en) * | 2018-04-20 | 2019-08-13 | 华中科技大学 | A kind of model mismatch diagnostic method of the multivariable control system of data-driven |
| CN108646553A (en) * | 2018-04-20 | 2018-10-12 | 华中科技大学 | A method of statistics on-line monitoring closed-loop control system model quality |
| CN112424705A (en) * | 2018-07-13 | 2021-02-26 | 西门子股份公司 | Process control prediction model adjusting method and device and process controller |
| CN111176155B (en) * | 2019-12-20 | 2021-07-02 | 华中科技大学 | Process model mismatch detection method of closed-loop model predictive control system |
| CN111300984B (en) * | 2020-03-11 | 2021-07-16 | 华南理工大学 | Parameter self-tuning method for roll printing system and roll printing system |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP3966115B2 (en) * | 2002-08-05 | 2007-08-29 | 株式会社デンソー | Design support apparatus, design support method, and design support program |
| US7421374B2 (en) * | 2005-11-17 | 2008-09-02 | Honeywell International Inc. | Apparatus and method for analyzing model quality in a process control environment |
| CN101349893B (en) * | 2007-07-18 | 2011-03-16 | 太极光控制软件（北京）有限公司 | Forecast control device of adaptive model |
| CN104932488B (en) * | 2015-06-30 | 2017-06-16 | 南京工业大学 | Model predictive control performance evaluation and diagnosis method |
-
2016
- 2016-03-14 CN CN201610144819.4A patent/CN105700358B/en not_active Expired - Fee Related
Also Published As
| Publication number | Publication date |
|---|---|
| CN105700358A (en) | 2016-06-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105700358B (en) | A kind of modeling quality control method of the model predictive controller of band drift interference | |
| Annaswamy et al. | A historical perspective of adaptive control and learning | |
| CN104932488B (en) | Model predictive control performance evaluation and diagnosis method | |
| CN107272640A (en) | A kind of modeling quality control method and system based on model predictive controller | |
| Zhang et al. | Low-computation adaptive fuzzy tracking control of unknown nonlinear systems with unmatched disturbances | |
| CN105807611B (en) | A kind of model of closed-loop control system and the unmatched detection method of object | |
| CN108536127B (en) | A kind of model mismatch diagnostic method of the multivariable control system of data-driven | |
| CN113297798B (en) | Robot external contact force estimation method based on artificial neural network | |
| CN106383443B (en) | Anti-interference control method and system | |
| di Capaci et al. | A valve stiction tolerant formulation of MPC for industrial processes | |
| Chen et al. | Design fixed-time practical distributed average tracking algorithms for nonlinear signals with bounded-and Lipschitz-type derivatives | |
| CN107168066B (en) | Greenhouse environment self-adaptive control method | |
| Pengfei et al. | Soft sensor model derived from wiener model structure: modeling and identification | |
| Duţescu et al. | Model predictive control of a nonlinear laboratory twin rotor aero-dynamical system | |
| CN112925207A (en) | Greenhouse environment temperature self-adaption method based on parameter identification | |
| Simkoff et al. | Plant-model mismatch evaluation for unconstrained MPC with state estimation | |
| Tiwari et al. | Learning digital emulators for closed architecture machine tool controllers | |
| Ji et al. | Parameters tuning of model free adaptive control based on minimum entropy | |
| Waleed et al. | Integration of a robust kalman filter with model-free control | |
| Jian-hong et al. | Virtual reference feedback tuning control design for closed loop system | |
| Guojun et al. | A real-time updated model predictive control strategy for batch processes based on state estimation | |
| CN110703606A (en) | Novel self-coupling PID cooperative control theory method | |
| Fan et al. | A method for parameter identification of combined integrating systems | |
| Salhi et al. | Identification of nonlinear dynamic systems described by Hammerstein state-space models with discontinuous nonlinearities | |
| Zhang et al. | Robust adaptive neural control of SISO nonlinear systems with unknown dead-zone and completely unknown control gain |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20180612 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |