JP6140427B2 - Model predictive control apparatus and method - Google Patents

Description

  The present invention relates to a model predictive control apparatus and method for controlling a controlled object using a model.

  Model predictive control is a control method that predicts the future response of a control variable using a model, and calculates an operation variable that makes the predicted response asymptotic to a reference trajectory or a target value. Model predictive control is characterized by the ability to directly handle constraints (upper and lower limits) on control variables and manipulated variables, and is used in control industries that require control that takes variable constraints into account, such as the process industry. (See Patent Documents 1 and 2).

  Hereinafter, model predictive control will be described with reference to FIG. 7 taking a case where a process of 1 input and 1 output is a control target. FIG. 7 is a configuration diagram illustrating a configuration example of a system that performs model prediction control. This system includes a control target process 701, a data collection unit 702, a response prediction generation unit 703, a target value trajectory generation unit 704, a constraint condition setting unit 705, and an operation variable calculation unit 706. In the following, the time series of control variables output by the control target process 701 is represented by y (k), and the time series of operation variables output to the control target process 701 (input by the control target process 701) is u ( k).

  The data collection unit 702 collects data such as control variables and operation variables of the control target process 701. The data collection unit 702 collects data such as disturbance variables.

  The response prediction generation unit 703 predicts a future response of the control target process 701 based on the data collected by the data collection unit 702 and the data collected in the past. The future response predicted and output by the response prediction generation unit 703 is time-series data such as “yf (k + 1), yf (k + 2),..., Yf (k + Hp)”, for example. The section [k + 1, k + Hp] is a section for which a future response is to be predicted / evaluated, and is called a predicted horizon. k + Hp is the end point of the predicted horizon. The start point of the predicted horizon section is often other than “k + 1”, but here, for simplicity, the start point is unified to “k + 1”.

  The target value trajectory generation unit 704 generates a desired value series (target value trajectory) in the prediction horizon of the control variable. For example, the target value trajectory generation unit 704 generates “yr (k + 1), yr (k + 2),..., Yr (k + Hp)” as the target value trajectory.

  The feature of model predictive control is that variable constraints can be directly considered. The constraint of this variable is set by the constraint condition setting unit 705. Here, the upper limit of the manipulated variable is uH (k), the lower limit of the manipulated variable is uL (k), the upper limit of the control variable is yH (k), the lower limit of the control variable is yL (k), and the upper limit of the change amount of the manipulated variable Is expressed as ΔuH (k), and the lower limit of the change amount of the manipulated variable is expressed as ΔuL (k). In addition, here, an expression is given in consideration of the case where the constraint depends on the time k. However, a constant value may be used regardless of the time. Note that, as these constraints (upper and lower limit values), values determined by the user or values generated based on user operations are set in the constraint condition setting unit 705 in advance. For example, when the user determines a constraint that narrows the upper and lower limits, the constraint is not set in the constraint condition setting unit 705 as it is, but is set so as to gradually approach the new upper and lower limits from the current upper and lower limit values. For example, the value determined by the user may be set after being processed by a filter or the like.

  The operation variable calculator 706 calculates a time series of operation variables to be given to the controlled process 701 in the future in accordance with the following purpose.

• The manipulated variable, the amount of change in the manipulated variable, and the control variable do not deviate outside the given upper and lower limits.
• The response of the control variable is as close as possible to the given target value trajectory.
・ The amount of change in the manipulated variable should be as small as possible.

  In accordance with these objectives, the problem of obtaining the optimum manipulated variable variation “Δu (k), Δu (k + 1),..., Δu (k + Hc)” is formulated as a quadratic programming method. Δu (k) is a backward difference of u (k) and is defined by “Δu (k) = u (k) −u (k−1)”. If the sequence of Δu (k) is obtained, the sequence of u (k) is obtained by adding the obtained Δu (k) to the control variable u (k−1) of the control one cycle before.

  Note that the interval [k, k + Hc] for obtaining the sequence of Δu (k) is called a control horizon. The end point k + Hc of the control horizon needs to be before the end point k + Hp of the predicted horizon.

  The problem formulated in this way is as shown in the following formula (1).

  However, “Δy (k + 1), Δy (k + 2),..., Δy (k + Hp)” is “Δu (k + 1), Δu (k + 2),. This is the amount by which the response of the control target process 701 changes when input. Further, q (i) and r (i) are weighting factors of the evaluation function.

  Finally, the first value (u (k)) in the obtained time series is output to the control target process 701.

  By this calculation, it is possible to automatically perform control such that the control variable follows the target value and the control variable deviated from the target value due to disturbance is returned to the target value. Further, here, for the sake of simplicity, the description has been made on the 1-input 1-output system, but it can be extended to a multi-input multi-output system.

  Here, it is important as one of the purposes of control to suppress the influence of disturbance on the control target. There are several methods for suppressing disturbances in model predictive control. One of these is an operation that predicts the amount of control variable response change due to disturbances from the collected data and suppresses the effects of the predicted changes. There is a method for calculating a variable (see Non-Patent Document 1).

  The method of Non-Patent Document 1 can be easily incorporated into a normal model predictive control algorithm if a model with disturbance as an input and a control variable as an output is prepared. In this method, as shown in FIG. 8, the difference between the response prediction (a) and the target value (b) in consideration of the influence of disturbance is eliminated, and an operation variable that matches the target value is calculated and calculated. Control with manipulated variables. In addition, (c) of FIG. 8 is the response prediction when there is no disturbance.

  The difference between the case where the method of Non-Patent Document 1 is applied to the above-described system example and the above-described example is that “yf (k + 1), yf (k + 2),..., Yf ( k + Hp) ”is the difference in whether the influence of disturbance is taken into consideration. Furthermore, if the accuracy of the model is high, the influence of the disturbance on the response of the control variable can be accurately predicted, and the influence of the disturbance can be removed by the model predictive control.

  However, it is not always best to construct a control system that completely eliminates the influence of the predicted disturbance. Dare to control without completely removing the effects of disturbance. For example, in Patent Document 1, the difference between the output of the process and the predicted value by the model is not processed as an estimated value of the unmeasurable disturbance as it is, but is processed through the disturbance filter, and the response prediction using the processed result is sent to the manipulated variable calculator. As an input.

  For example, if the difference between the process output and the predicted value from the model is the true value of the unmeasurable disturbance, and this value will continue in the future, the disturbance filter described above is not necessary. However, in reality, since it is difficult to obtain accurate disturbance information, the method of Patent Document 1 that smoothes the disturbance estimated value using a filter often exhibits good control performance.

  The same can be said for a configuration in which the influence of disturbance is predicted using a model. When a model with low accuracy is used, if it is attempted to completely eliminate the influence of disturbance, the disturbance suppression performance is rather deteriorated, and the controlled object is adversely affected. In addition, even if the accuracy of the model is high, it may not be preferred that the manipulated variable behaves so as to completely suppress the disturbance. This is because, in the time series of the manipulated variables in a state in which the disturbance is completely suppressed, the amount of change for each control cycle may be very large, and a load may be applied to the controlled object.

  For this reason, even if some influence of disturbance remains, it is often desirable to behave so that the manipulated variable changes its value slowly and gradually. In addition, if the behavior of the manipulated variable is gentle, there is an advantage that even if the accuracy of the model is low, the deterioration of the control performance and the adverse effect on the control target can be reduced.

  As one of the methods to calm the behavior of the manipulated variable, when predicting the amount of change in the control variable response due to disturbance, the response of the model is not used as it is, but the response prediction change is smoothed with a filter etc. (disturbance There is a way to use it by correcting the influence of the to a gentle change. The response prediction obtained by this method suppresses or ignores the fast movement of the original prediction. This method assumes a response that changes more slowly than the effect of the disturbance that actually appears in the control variable. If the manipulated variable is obtained based on the corrected response prediction in the calculation of the model predictive control after such correction, the difference between the response prediction and the target value trajectory is smaller than the original difference, and Since the response prediction is smooth, the behavior of the manipulated variable for disturbance suppression is also relatively gentle.

  In the following, model predictive control used by correcting the change in response prediction will be described with reference to FIG. FIG. 9 is a configuration diagram illustrating a configuration example of a system that performs model prediction control by correcting a change in response prediction. This system includes a control target process 701, a data collection unit 702, a response prediction generation unit 703, a target value trajectory generation unit 704, a constraint condition setting unit 705, an operation variable calculation unit 706, and a response prediction correction unit 707. In this system, a response prediction correction unit 707 is newly provided between the response prediction generation unit 703 and the operation variable calculation unit 706 in the model prediction control system described above.

In this system, the response prediction yf generated by the response prediction generation unit 703 is corrected by the response prediction correction unit 707, and the corrected result yf ^* is passed to the operation variable calculation unit 706, so that the behavior of the operation variable is moderated. I have to. In the response prediction correction unit 707, the prediction of the future response generated by the response prediction generation unit 703 is made smoother and smaller than the original prediction by applying a filter or the like. In this method, as shown in FIG. 10, the corrected response prediction (a ′) corrected as described above is generated from the response prediction (a) considering the influence of disturbance. Then, the difference between the corrected response prediction (a ′) and the target value (b) is eliminated, an operation variable that matches the target value (b) is calculated, and control is performed using the calculated operation variable.

JP 05-265514 A

Jan M. Msciejowski, Translated by Shuichi Adachi and Masaaki Kanno, “Model Predictive Control: Optimal Control under Constraints”, Tokyo Denki University Press, 2005. S. Joe Qin, Thomas A. Badgwell, "A survey of industial model predictive control technology", Control Engineering Practice, vol.11, pp.733-764, 2003.

  However, when the response prediction considering the influence of disturbance is corrected as described above and the response prediction of the control variable is determined to change more slowly, the amount of change in the manipulated variable increases and the control performance deteriorates. In some cases, the purpose of calming the behavior of the manipulated variable cannot be achieved. In particular, when the range of the upper and lower limits of the control variable is narrowed, the above-described problem is likely to occur.

  The present invention was made in order to solve the above problems, and even if the control variable is obtained as a response prediction that changes more slowly by correcting the response prediction in consideration of the influence of disturbance, It is an object of the present invention to make it possible to suppress deterioration in control performance such as an increase in the amount of change in manipulated variable.

The model predictive control apparatus according to the present invention includes an operation variable used for control by a control target, a control variable output by the control target, a data collection unit that collects disturbance information, and a control variable and operation variable collected by the data collection unit. , And a response prediction generation unit that predicts a control variable to be output in the future based on data including disturbance information and outputs it as a future response, and the influence of disturbance on the time series change of the future response output by the response prediction generation unit A response prediction correction unit that generates a corrected future response that is corrected to a gentle change, and a target value trajectory generation unit that generates and outputs a target value trajectory of a control variable including a time series of desirable values that the control target will output in the future, and , The constraint condition storage unit that stores the upper and lower limits of the control variable that the control target will output in the future, and the corrected future response output from the response prediction correction unit approaches the target value trajectory. And an operation variable operator controlled object is output to the controlled object by calculating the manipulated variables to be used in the future, out of tolerance or tolerance to lower to the upper limit of the modified prospective response, response future by the response prediction correction unit The relative relationship between the future response or the corrected future response and the upper and lower limit values is corrected so as to suppress the allowable range from becoming smaller than the original allowable range from 1 to the upper limit or lower limit .

In the model prediction control device, the future response predicted by the response prediction generation unit based on at least one of the future response predicted by the response prediction generation unit and the first corrected future response output by the response prediction correction unit. A relative constraint correction unit that calculates and outputs a second corrected future response that is corrected, and a correction upper limit value and a correction lower limit value that are corrected upper limit values and lower limit values set in the constraint condition storage unit, The constraint correction unit may change the second corrected future response so that the difference between the second corrected future response and the correction upper limit value and the correction lower limit value is larger than the difference between the first corrected future response and the upper limit value and the lower limit value. The response, the correction upper limit value, and the correction lower limit value may be calculated.

In the model predictive control device, based on the future response predicted by the response prediction generation unit, the corrected upper limit value and the corrected lower limit value obtained by correcting the upper limit value and the lower limit value set in the constraint condition storage unit are calculated and output. with a relative constraint modification unit, relative constraint adjustment unit, the difference between the corrected upper limit value and the corrected lower limit that fixes the upper and lower limit values set in the corrected future response and the constraints storage unit, corrected future response The corrected upper limit value and the corrected lower limit value may be calculated so as to be larger than the difference between the upper limit value and the lower limit value.

In the model predictive control device, based on the future response predicted by the response prediction generation unit, the future response predicted by the response prediction generation unit, and the upper limit value and the lower limit value set in the constraint condition storage unit are corrected. A relative constraint correction unit that calculates and outputs a value and a correction lower limit value, and the relative constraint correction unit outputs the future response output from the response prediction generation unit and the upper limit value and the lower limit value set in the constraint condition storage unit the difference between the corrected upper limit value and the corrected lower limit that fixes are modified so that Do greater than the difference between the future response and the upper limit value and the lower limit value, may be calculated correction upper limit value and the corrected lower limit value.

In the model predictive control device, based on the future response predicted by the response prediction generation unit, a third corrected future response obtained by correcting the future response predicted by the response prediction generation unit, and an upper limit set in the constraint condition storage unit with a relative constraint modification unit for outputting the value and the lower limit, the relative constraint adjustment unit, the difference between the difference is corrected future response and the upper limit value and the lower limit value of the third modified prospective response and the upper limit value and the lower limit value The third modified future response may be calculated so as to be larger .

  In the model predictive control apparatus, the manipulated variable calculation unit changes the allowable range used when calculating the operational variable to be used in the future by the control target from the allowable range based on the relationship between the future response and the upper limit value and the lower limit value. It may be adjustable. For example, the relaxation width may be adjustable in a section in which an operation variable used in the future by the control target is calculated. Further, the relaxation width may be adjusted based on the characteristics of the controlled object. Moreover, the characteristic of the controlled object is the step response of the controlled object

The model predictive control apparatus method according to the present invention includes an operation variable used for control by a control object, a control variable output by the control object, a data collection step for collecting disturbance information, and a control variable collected in the data collection step. A response prediction generation step for predicting a control variable to be output in the future based on data including operation variables and disturbance information, and outputting as a future response, and a time series change of the future response output in the response prediction generation step. A response prediction correction step that generates a corrected future response in which the influence of disturbance has been corrected to a gentle change, and a target value trajectory that generates and outputs a target value trajectory of a control variable that consists of a time series of desired values that the controlled object will output in the future A generation step, a constraint condition storing step for storing the upper limit value and lower limit value of the control variable to be output in the future by the controlled object, and a response prediction correction And a manipulated variable calculating step modifying future response output in step is controlled object so as to approach the target value trajectory outputs to the controlled object is calculated and the operating variables used in the future, the permissible range of the upper limit of the modified prospective response or Future response or modified future so that the allowable range to the lower limit is suppressed by the response prediction correction step, which is smaller than the original allowable range from the future response to the upper limit or lower limit. Correct the relative relationship between response and upper and lower limits .

In the model predictive control method, the prediction is generated in the response prediction generation step based on at least one of the future response predicted in the response prediction generation step and the first corrected future response output in the response prediction correction step. A relative constraint correction step of calculating and outputting a second corrected future response in which the future response is corrected, and a correction upper limit value and a correction lower limit value in which the upper limit value and the lower limit value stored in the constraint condition storage step are corrected. In the relative constraint correction step, the second modified future response and the second upper limit value and the lower limit value are set such that the difference between the first corrected future response and the upper limit value and the lower limit value is larger than the second corrected future response. The corrected future response, the correction upper limit value, and the correction lower limit value may be calculated.

In the model predictive control method, based on the future response predicted in the response prediction generation step, the corrected upper limit value and the corrected lower limit value obtained by correcting the upper limit value and the lower limit value stored in the constraint condition storage step are calculated and output. with a relative constraint correction step of, in a relative constraint correction step, the difference between the stored upper limit value and the modified modified the lower limit upper limit value and the corrected lower limit value corrected future response and constraint condition storing step, modified prospective response and The corrected upper limit value and the corrected lower limit value may be calculated so as to be larger than the difference between the upper limit value and the lower limit value.

In the model predictive control method described above, a modification in which the future response predicted in the response prediction generation step and the upper limit value and the lower limit value stored in the constraint condition storage step are corrected based on the future response predicted in the response prediction generation step. It has a relative constraint correction step that calculates and outputs an upper limit value and a corrected lower limit value. In the relative constraint correction step, the upper limit value and the lower limit value stored in the future response output in the response prediction generation step and the constraint condition storage step the difference between the corrected upper limit value and the corrected lower limit that fixes are modified so that Do greater than the difference between the future response and the upper limit value and the lower limit value, may be calculated correction upper limit value and the corrected lower limit value.

In the model predictive control method, the third modified future response obtained by correcting the future response predicted in the response prediction generation step based on the future response predicted in the response prediction generation step, and the constraint condition storage step are set. A relative constraint correction step for outputting an upper limit value and a lower limit value. In the relative constraint correction step, a difference between the third corrected future response and the upper limit value and the lower limit value is calculated as a difference between the corrected future response and the upper limit value and the lower limit value. The third modified future response may be calculated so as to be larger than the difference .

  As described above, according to the present invention, even if the control variable is obtained as a response prediction that changes more slowly by correcting the response prediction considering the influence of disturbance, the amount of change in the manipulated variable increases. It is possible to obtain an excellent effect that the deterioration of the control performance can be suppressed.

FIG. 1 is an explanatory diagram illustrating states of future response (a), corrected future response (a ′), target value trajectory (b), and lower limit (d) that change in time series. FIG. 2 is a configuration diagram showing the configuration of the model predictive control device 100 according to Embodiment 1 of the present invention. FIG. 3 is a flowchart for explaining the operation (model prediction control method) of the model prediction control apparatus according to Embodiment 1 of the present invention. FIG. 4 is a configuration diagram showing the configuration of the model predictive control device 200 according to Embodiment 2 of the present invention. FIG. 5 is a configuration diagram showing the configuration of the model predictive control apparatus 300 according to Embodiment 3 of the present invention. FIG. 6 is a configuration diagram showing the configuration of the model prediction control device 400 according to Embodiment 4 of the present invention. FIG. 7 is a configuration diagram illustrating a configuration example of a system that performs model prediction control. FIG. 8 is an explanatory diagram illustrating the state of future response (a) that changes in time series, target value trajectory (b), and response prediction (c) when there is no disturbance. FIG. 9 is a configuration diagram illustrating a configuration example of a system that performs model prediction control. FIG. 10 is an explanatory diagram illustrating the state of future response (a) changing in time series, modified future response (a ′), target value trajectory (b), and response prediction (c) when there is no disturbance.

[principle]
First, the principle of the present invention is described in which the effect of suppressing deterioration in control performance such as an increase in the change amount of the manipulated variable is obtained.

  As described in the background art section, in model predictive control, two control variables are required. One is to bring the response closer to the target value trajectory as much as possible, and the second is to prevent the response from exceeding the constraints of the set upper limit value and lower limit value. In the conventional technique, it can be considered that the control law is corrected by correcting the response prediction so that the action of bringing the response close to the target value (target value trajectory) is suppressed.

  The inventor studied how the relationship between response prediction and the constraints imposed on the response (future response) changes by modifying the response. As a result of this research, by modifying the response so as to suppress the effect of bringing the response closer to the target value, the movement of the response to keep the upper limit and lower limit constraints in the calculation process of the manipulated variable that the controlled object will use in the future It has been clarified that the movement of the manipulated variable becomes excessive as a result of this change.

  FIG. 1 shows a future response (a), a corrected future response (a ′), a target value trajectory (b), and a lower limit (d). The future response (a) predicted faithfully using a model including the influence of disturbance is a value that the actual control variable value will take if the prediction is accurate and the value of the manipulated variable is not changed. Shows changes. The modified future response (a ′) is intentionally modified to suppress the movement of the manipulated variable. In addition, the target value trajectory (b) is an object that brings the control variable closer to the possible extent. The lower limit (d) is the lower limit of the control variable, and prevents the control variable obtained in the process of calculating the manipulated variable from becoming smaller than this value.

  In FIG. 1, when the difference Δ1 between the future response (a) and the lower limit (d) is compared with the difference Δ2 between the corrected future response (a ′) and the lower limit (d), the difference Δ2 is smaller. This is because the movement of the operation variable is aimed to be suppressed in the process of calculating the operation variable to be used in the future by the controlled object.

  By the way, in the calculation of the manipulated variable to be used in the future by the controlled object, the restriction of the control variable also affects. In FIG. 1, if the value of the manipulated variable is changed in the calculation process, the future response (a) will change. However, this change is not allowed without limitation, and it is important that the change is within the limits of the upper limit value and the lower limit value. For example, in FIG. 1, a result of calculating a future response that falls below the lower limit (d) is not allowed. In other words, the response of the control variable is allowed to vary up to the lower limit (d). If this allowable range is wide, not only the control variable, but also the allowable range for the movement of the manipulated variable becomes wide, and the calculation of the manipulated variable becomes easy.

  However, when the corrected future response (a ′) is used to calculate the manipulated variable used in the future by the controlled object, the allowable range described above changes. In the case shown in FIG. 1, the difference Δ1 indicating the allowable range to the lower limit is reduced to the difference Δ2. On the other hand, the allowable range (not shown) to the upper limit becomes wider than that in the case of the future response (a).

  As a result, the range allowed as the movement of the operation variable changes in the calculation of the operation variable to be used in the future by the controlled object. In particular, the behavior that brings the control variable closer to the lower limit (d) is narrower than the originally acceptable range. For this reason, the change amount of the manipulated variable may become excessive. In particular, if the set upper / lower limit width (difference) is small, even if the corrected future response (a ') is used, if the upper limit and lower limit values are used as they are, the originally acceptable range and response will be corrected. As a result, the relative difference from the changed range becomes large, and the deterioration of the control performance may occur.

  In order to solve the above-described problem, it is only necessary to suppress the allowable range based on the upper limit value / lower limit value of the future response (corrected future response) resulting from the correction from being reduced. For this reason, for example, in the first embodiment, as described below with reference to FIG. 2, when the constraint correction unit 106 is newly provided and the first response prediction correction unit 103 corrects the future response, the allowable range is The upper and lower limits of the smaller one are corrected to ensure an allowable range equivalent to the original allowable range. By doing so, it is possible to suppress a situation in which an inappropriate operation variable is calculated because it is based on an allowable range that is smaller than actual.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

[Embodiment 1]
First, Embodiment 1 of the present invention will be described. FIG. 2 is a configuration diagram showing the configuration of the model predictive control device 100 according to Embodiment 1 of the present invention. The model prediction control apparatus 100 includes a data collection unit 101, a response prediction generation unit 102, a first response prediction correction unit 103, a target value trajectory generation unit 104, a constraint condition setting unit (a constraint condition storage unit) 105, a relative constraint correction. Unit 106 and manipulated variable calculation unit 107. The relative constraint correction unit 106 includes a constraint correction unit 161 that corrects the constraint condition, and a second response prediction correction unit 162 that corrects the response prediction.

  The data collection unit 101 collects operation variables used by the control target process 120 for control, control variables output by the control target process 120, and disturbance information.

  The response prediction generation unit 102 predicts a control variable to be output in the future by the control target process 120 based on data including the control variable, operation variable, and disturbance information collected by the data collection unit 101, and outputs it as a future response. For example, if the time series of the control variable output from the control target process 120 is y (k), the response prediction generation unit 102, for example, “yf (k + 1), yf (k + 2),..., Yf (k + Hp) ”Is output as a future response.

The first response prediction correction unit 103 generates a corrected future response in which the influence of the disturbance in the time series change of the future response output from the response prediction generation unit 102 is corrected to a gentle change. For example, the first response prediction correction unit 103 outputs time series data such as “y ^* f (k + 1), y ^* f (k + 2),..., Y ^* f (k + Hp)” as a corrected future response. To do.

  The target value trajectory generation unit 104 generates and outputs a target value trajectory of a control variable that includes a time series of desirable values that the control target process 120 will output in the future. For example, the target value trajectory generating unit 104 generates, for example, “yr (k + 1), yr (k + 2),..., Yr (k + Hp)” as the target value trajectory.

  The constraint condition setting unit 105 stores an upper limit value and a lower limit value of the control variable y (k) that the control target process 120 will output in the future (calculated as output in the future). For example, if the upper limit of the control variable is expressed as yH (k) and the lower limit of the control variable is expressed as yL (k), the constraint condition setting unit 105 has “y (k + i) ≧ yL (k + i)” and “y (k + i)”. ≦ yH (k + i) ”is set and stored. The constraint condition setting unit 105 also sets and stores an upper limit value and a lower limit value of the manipulated variable, an upper limit value and a lower limit value of the change amount of the manipulated variable.

  The data collection unit 101, the response prediction generation unit 102, the target value trajectory generation unit 104, and the constraint condition setting unit 105 described above are the same as those in the related art. The first response prediction correction unit 103 is the same as the response prediction correction unit of the prior art.

  The relative constraint correction unit 106 generates a response prediction based on at least one of the future response predicted by the response prediction generation unit 102 and the first corrected future response output by the first response prediction correction unit 103. The second modified future response obtained by correcting the future response predicted by the unit 102 and the corrected upper limit value and the corrected lower limit value obtained by correcting the upper limit value and the lower limit value set (stored) in the constraint condition setting unit 105 are calculated. Output. The constraint correction unit 161 calculates and outputs the correction upper limit value and the correction lower limit value, and the second response prediction correction unit 162 calculates and outputs the second correction future response. The relative constraint correction unit 106 performs the second correction so that the relationship between the second corrected future response, the correction upper limit value, and the correction lower limit value corresponds to the relationship between the future response, the upper limit value, and the lower limit value. The future response, the correction upper limit value and the correction lower limit value are calculated.

  Hereinafter, the calculation of the relative constraint correction unit 106, that is, the relationship between the second corrected future response and the correction upper limit value and the correction lower limit value corresponds to the relationship between the future response, the upper limit value, and the lower limit value. A method for calculating the second corrected future response, the correction upper limit value, and the correction lower limit value will be specifically described.

As an example, a description will be given based on FIG. 1 used in the explanation of the principle. In this example, by correcting the future response, the difference between the first corrected future response and the lower limit value is smaller than the difference between the future response before correction and the lower limit value. That is, when the first corrected future response is y ^* f (k), the following relationship is established.

  The relative constraint correcting unit 106 corrects the future response and the lower limit value so that the difference between the second corrected future response and the correction lower limit value is at least larger than the difference between the first corrected future response and the lower limit value. . Here, the problem is the difference between the second corrected future response and the correction lower limit value, that is, the relative relationship. If this difference is larger than the difference between the first corrected future response and the lower limit value, either the future response or the lower limit value may be corrected, or both may be corrected. The reason why the portion for performing this calculation is the relative constraint correction unit is that, as described above, the purpose is to correct the relative relationship between the future response and the upper and lower limit values.

The calculation of the relative constraint correction unit 106 needs to make at least the difference between the second corrected future response and the correction lower limit value larger than the difference between the first corrected future response and the lower limit value. When the modified future response of 2 is represented by y ^# f (k) and the modified lower limit value is represented by y ^# L (k), it is necessary to satisfy the following formula (3) at a minimum.

  In general, the difference between the second corrected future response and the correction lower limit value is preferably made to correspond to the difference between the future response before correction and the lower limit value, so that the following equation (4) is satisfied. It is a desirable embodiment to be corrected to.

  Of course, as long as Expression (3) is satisfied, the left side of Expression (4) may be made larger or smaller than the right side. However, if the left side of Formula (4) is made small, the effect of this invention will fall. Also, increasing the left side of equation (4) sets a constraint that is more lenient than the original lower limit for the control variable, so there is a possibility of performance degradation in terms of maintaining the control variable constraint. is there.

  So far, the example of correcting the future response and the lower limit value has been described with reference to FIG. 1, but in the opposite case, that is, the difference between the upper limit value and the first corrected future response is the upper limit value and the future before the correction. The case where the difference from the response becomes smaller can be dealt with by correcting the future response and the upper limit value in the same manner. Further, this correction may be performed only for the upper limit or the lower limit where the difference from the first corrected future response is smaller than the difference from the future response before the correction. If the difference from the first modified future response is larger than the difference from the future response before the modification, the modification may be performed in the same manner, and the performance may be improved from the viewpoint of maintaining the constraints on the control variable. There is no need to make corrections.

  The manipulated variable calculation unit 107 performs the first response prediction correction under the condition that the change amount of the control variable is in a range determined by the second corrected future response output from the relative constraint correction unit 106, the correction upper limit value, and the correction lower limit value. The control target process 120 calculates an operation variable to be used in the future so that the corrected future response output from the unit 103 approaches the target value trajectory, and outputs it to the control target process 120. For example, the operation variable calculator 107 calculates a time series of operation variables to be given to the control target process 120 in the future in accordance with the following purpose.

• The manipulated variable, the amount of change in the manipulated variable, and the control variable do not deviate outside the given upper and lower limits.
• The response of the control variable is as close as possible to the given target value trajectory.
・ The amount of change in the manipulated variable should be as small as possible.

  If the time series of operation variables output to the control target process 120 (input by the control target process 120) is u (k), the operation variable calculation unit 107 performs the optimal operation variable in accordance with the above-described purpose. The problem of obtaining the change amount “Δu (k), Δu (k + 1),..., Δu (k + Hc)” is formulated as a quadratic programming method as shown in the following equation (5).

  Δu (k) is a backward difference of u (k) and is defined by “Δu (k) = u (k) −u (k−1)”. If the sequence of Δu (k) is obtained, the sequence of u (k) is obtained by adding the obtained Δu (k) to the control variable u (k−1) of the control one cycle before.

  By the way, when the last equation of equation (5) is modified, the following equation (6) is obtained. As shown in the equation (6), the difference between the corrected lower limit value and the second corrected future response, and the difference between the corrected upper limit value and the second corrected future response becomes the upper and lower limit values with respect to the change amount of the control variable. That is, the operation in the relative constraint correction unit is equivalent to correcting the upper and lower limit values for the change amount of the control variable.

  Next, the operation (model prediction control method) of the model prediction control apparatus according to Embodiment 1 of the present invention will be described using the flowchart of FIG.

  First, in step S <b> 101, the upper limit value and lower limit value of the control variable that the control target process 120 outputs in the future (calculated as output in the future) is set in the constraint condition setting unit 105. Thereby, the constraint condition setting unit 105 stores the upper limit value and the lower limit value of the control variable that the controlled process 120 will output in the future (constraint condition setting step). Next, in step S102, the target value trajectory generation unit 104 generates and outputs a target value trajectory of a control variable including a time series of desirable values that the control target process 120 will output in the future (target value trajectory generation step).

  Next, in step S103, the data collection unit 101 collects operation variables used by the control target process 120 for control, control variables output by the control target process 120, and disturbance information (data collection step). Next, in step S104, the response prediction generation unit 102 predicts the control variable that the control target process 120 will output in the future based on the data including the control variable, the operation variable, and the disturbance information collected by the data collection unit 101. Output as a response (response prediction generation step).

  Next, in step S105, a first corrected future response in which the influence of disturbance in the time series change of the future response output from the response prediction generation unit 102 is corrected to a gentle change is generated (response prediction correction step). Next, in step S106, the relative constraint correction unit 106 responds based on at least one of the future response output from the response prediction generation unit 102 and the first corrected future response output from the first response prediction correction unit 103. A second corrected future response that is corrected from the future response output from the prediction generation unit 102, and a corrected upper limit value and a corrected lower limit value that are corrected from the upper limit value and the lower limit value set in the constraint condition setting unit 105 are calculated. Output (relative constraint correction step). At this time, the relative constraint correction unit 106 corrects the relationship so that the relationship between the second corrected future response, the correction upper limit value, and the correction lower limit value corresponds to the relationship between the future response, the upper limit value, and the lower limit value. Calculate the upper limit and the corrected lower limit.

  Next, in step S107, the operation variable calculation unit 107 calculates an operation variable to be used in the future by the control target process 120 so that the corrected future response output from the first response prediction correction unit 103 approaches the target value trajectory ( Manipulation variable calculation step). The above calculation is performed under the condition that the control variable falls within the range of the second corrected future response, the correction upper limit value, and the correction lower limit value output from the relative constraint correction unit 106. Thereafter, in step S108, the operation variable calculator 107 outputs the calculated operation variable to the control target process 120.

  Thus, according to the first embodiment, based on the future response output by the response prediction generation unit 102 and the correction amount of the first corrected future response output by the first response prediction correction unit 103, the second response The relationship between the corrected future response, the correction upper limit value, and the correction lower limit value corresponds to the relationship between the future response before correction and the upper limit value and the lower limit value. As a result, even if the response prediction considering the influence of disturbance is corrected and the response prediction of the control variable is determined to change more slowly, deterioration of the control performance such as an increase in the change amount of the manipulated variable can be suppressed. become.

  The model predictive control device is, for example, a computer device including a CPU, a main storage device, an external storage device, a network connection device, and the like, and the CPU operates by a program developed in the main storage device. Each function is realized. Each function may be distributed among a plurality of computer devices.

[Embodiment 2]
Next, Embodiment 2 of the present invention will be described with reference to FIG. FIG. 4 is a configuration diagram showing the configuration of the model predictive control device 200 according to Embodiment 2 of the present invention. The model prediction control apparatus 200 includes a data collection unit 101, a response prediction generation unit 102, a first response prediction correction unit 103, a target value trajectory generation unit 104, a constraint condition setting unit 105, and an operation variable calculation unit 107. The model prediction control apparatus 200 also includes a relative constraint correction unit 206, and the relative constraint correction unit 206 includes a constraint correction unit 261 that corrects the constraint condition.

  In the second embodiment, the relative constraint correcting unit 206 is different from the above-described first embodiment. In the relative constraint correcting unit 206, the constraint correcting unit 261 corrects only the upper and lower limit constraints, and outputs the first modified future response as it is as the second modified future response in the first embodiment described above. As described above, in the second embodiment, the first modified future response and the second modified future response are the same, and hereinafter simply referred to as a modified future response. In the second embodiment, since the relative constraint correction unit 206 and the constraint correction unit 261 are different from the above-described first embodiment, these will be described in detail, and description of other configurations will be omitted.

  The relative constraint correction unit 206 (constraint correction unit 261) is configured so that the relationship between the corrected future response, the correction upper limit value, and the correction lower limit value corresponds to the future response before correction, the upper limit value, and the lower limit value. The correction lower limit value is calculated. Similarly to the first embodiment, the difference between the corrected future response and the correction upper limit value and the correction lower limit value needs to be at least larger than the difference between the corrected future response and the upper limit value and the lower limit value.

For example, the constraint correction unit 261 changes “y ^* f (k + i) + Δy (k + i) ≧ yL (k + i)” to “y ^* f (k + i) + Δy (k + i) ≧ yL (k + i) −α (i) max {0 , Yf (k + i) −y ^* f (k + 1)} ”. As a result, the correction lower limit value is yL (k + i) −α (i) max {0, yf (k + i) −y ^* f (k + i)}.

Further, the constraint correcting unit 261 sets “y ^* f (k + i) + Δy (k + i) ≦ yH (k + i)” to “y ^* f (k + i) + Δy (k + i) ≦ yH (k + i) + β (i) max {0 , Y ^* f (k + i) −yf (k + 1)} ”. As a result, the correction lower limit value is yH (k + i) + β (i) max {0, y ^* f (k + i) −yf (k + i)}. Note that max {a, b} is an operation for taking the maximum value (the larger value) of a and b, and α (i) and β (i) are α (i)> 0, β (i). It shall satisfy> 0. It is not necessary to modify the constraints on the manipulated variable.

When correcting as described above, both α (i) and β (i) are positive, and the value in max {} is always 0 or more, so either the correction upper limit value or the correction lower limit value and the corrected future response The difference is always wider than when no correction is made. Further, if yf (k + i)> y ^* f (k + i), the lower limit is widened, and if yf (k + i) <y ^* f (k + i), the upper limit is widened. Will spread the side that became. Although it is possible to similarly widen the restrictions on the side opposite to the side where the allowable range is narrowed, in other words, the side where the allowable range is widened, this operation may not be performed.

  Next, α (i) and β (i) will be described. If the values of α (i) and β (i) are set to 0, there is no relaxation effect, so it is necessary to make them positive. In order to secure a difference (original allowable range) between the uncorrected future response and the upper limit value and the lower limit value, both α (i) and β (i) may be set to 1. This is a standard setting value. When α (i) and β (i) are set to values between 0 and 1, a relaxation effect is obtained according to the set value, but the effect is smaller than when 1 is set.

  It is also possible to set α (i) and β (i) to 1 or more. This sets a permissible range wider than the original permissible range, and therefore temporarily allows the control variable to deviate from the upper limit / lower limit constraints. Such a setting can be expected to suppress the movement of the manipulated variable in the process of calculating the manipulated variable to be used in the future by the controlled object, but it must also be noted because it may degrade the transient response of the controlled variable. It is.

  α (i) and β (i) may be constants that do not depend on i, or may vary depending on i. For example, the value may be increased when i = 1, and the value may gradually approach 0 as i increases. In this way, it is possible to set such that the allowable range is widened recently and the allowable range is gradually narrowed in the future. Α (i) and β (i) may be fixed values or may be adjustable parameters.

  Further, α (i) and β (i) may be changed according to the characteristics of the controlled object. For example, when a large overshoot or undershoot is included in the step response from the observable disturbance to the control variable, it may be more easily affected by the narrowing of the allowable range. If it is predicted from the characteristics of the controlled object that this is the case, the effect of the present invention can be more reliably increased by making α (i) and β (i) larger than the standard value (1). It is thought that it will come to be obtained.

  Next, the operation of the manipulated variable calculation unit 107 will be described in detail. The operation of the operation variable calculation unit 107 is the same as that of the prior art except that the correction upper limit value and the correction lower limit value corrected by the constraint correction unit 261 are used. For example, the problem formulated in the following formula (7) Is used.

  Solving the above-described problem provides “Δu (k + 1), Δu (k + 2),..., Δu (k + Hc)” which is the time series of the optimum manipulated variable change amount. “U (k−1) + Δu (k)” obtained by adding Δu (k) to the operation variable value u (k−1) before one control cycle is output to the control target process 120 as the operation variable value of the current control cycle. Is done. This operation is the same as general model predictive control.

  In summary, the present invention provides a data collection unit 101, a response prediction generation unit 102, a first response prediction correction unit 103, a target value trajectory generation unit 104, a constraint condition setting unit 105, an operation The model predictive control by the variable calculation unit 107 is characterized in that the corrected future response approaches the target value trajectory within an allowable range corresponding to the relationship between the future response and the upper limit value and the lower limit value.

[Embodiment 3]
Next, Embodiment 3 of the present invention will be described with reference to FIG. FIG. 5 is a configuration diagram showing the configuration of the model predictive control apparatus 300 according to Embodiment 3 of the present invention. The model prediction control apparatus 300 includes a data collection unit 101, a response prediction generation unit 102, a first response prediction correction unit 103, a target value trajectory generation unit 104, a constraint condition setting unit 105, and an operation variable calculation unit 107. The model prediction control apparatus 300 includes a relative constraint correction unit 306, and the relative constraint correction unit 306 includes a constraint correction unit 361 that corrects the constraint condition.

  In the third embodiment, the operation in the relative constraint correcting unit 306 (constraint correcting unit 361) is different from the relative constraint correcting unit 206 in the second embodiment. In the third embodiment, the constraint correction unit 361 causes the relationship between the future response (second corrected future response), the correction upper limit value, and the correction lower limit value to correspond to the future response before correction, the upper limit value, and the lower limit value. Then, the correction upper limit value and the correction lower limit value are calculated. The difference from the second embodiment described above is that the future response before correction is output as it is as the second corrected future response. Other configurations are the same as those of the above-described embodiment.

For example, the constraint correction unit 361 corrects the lower limit value yL (k + i) to yL (k + i) −α (i) max {0, yf (k + i) −y ^* f (k + i)}. Further, the constraint correction unit 361 corrects the upper limit value yH (k + i) to yH (k + i) + β (i) max {0, y ^* f (k + i) −yf (k + 1)}. Note that the allowable ranges of α (i) and β (i) are different from those of the second embodiment. α (i) and β (i) must satisfy α (i)> − 1 and β (i)> − 1. The case where α (i) and β (i) are −1 corresponds to the prior art (the state where the relative constraint correcting unit 306 is not provided).

  In the third embodiment, the second modified future response and the constraint modified by the constraint modification unit 361 are transferred to the manipulated variable computing unit 107, and the manipulated variable is calculated by the formulation shown by the following equation (8). Is done.

  As a special example of the third embodiment, α (i) and β (i) may be 0. In this case, the relationship between the second corrected future response and the corrected upper limit value and the corrected lower limit value automatically matches the relationship between the future response and the upper limit value and the lower limit value. In addition, since the correction upper limit value and the correction lower limit value do not change depending on the value of the first corrected future response, there is no need to transmit to the relative constraint correction unit 306. This method has an advantage that the operation of the relative constraint correction unit 306 is simplified. However, since α (i) and β (i) are fixed, the control behavior cannot be finely adjusted.

[Embodiment 4]
Next, Embodiment 4 of the present invention will be described with reference to FIG. FIG. 6 is a configuration diagram showing the configuration of the model prediction control device 400 according to Embodiment 4 of the present invention. The model prediction control apparatus 400 includes a data collection unit 101, a response prediction generation unit 102, a first response prediction correction unit 103, a target value trajectory generation unit 104, a constraint condition setting unit 105, and an operation variable calculation unit 107. The model prediction control apparatus 400 includes a relative constraint correction unit 406, and the relative constraint correction unit 406 includes a second response prediction correction unit 462 that corrects the response prediction.

  In the fourth embodiment, the relative constraint correction unit 406 outputs the upper limit value and the lower limit value as they are as the correction upper limit value and the correction lower limit value. Hereinafter, the calculation of the relative constraint correction unit 406 will be described.

When the upper and lower limit correction formulas shown in the second embodiment are shifted from the right side to the left side, in calculation, “y ^* f (k + i) + α (i) max {0, yf (k + i) −y ^* f” respectively. (K + 1)} + Δy (k + i) ≧ yL (k + i) ”and“ y ^* f (k + i) −β (i) max {0, y ^* f (k + i) −yf (k + i)} + Δy (k + i) ≦ yH ( k + i) ". This can be interpreted as correcting only the future response (first corrected future response) rather than correcting the upper limit value and the lower limit value. Therefore, for example, in the operation variable calculation unit 107, the operation is performed using the corrected future response in which the relationship between the corrected future response and the upper limit value and the lower limit value is corrected to the state corresponding to the relationship between the future response and the upper limit value and the lower limit value. Variables may be calculated.

That is, as a second modified future response, y ^* f (k + i) + α (i) {yf (k + i) −y ^* f (k + 1)} is set for i satisfying yf (k + i)> y ^* f (k + i). , Yf (k + i) <y ^* f (k + 1) is output as y ^* f (k + i) −β (i) {y ^* f (k + i) −yf (k + i)}. For i where yf (k + i) and y ^* f (k + i) are the same, y ^* f (k + i) is output as it is.

  In the case of the fourth embodiment, the interval between the second corrected future response and the upper and lower limit values becomes narrower on the side opposite to the side where the interval between the corrected future response and the upper and lower limit values is widened. This is because the upper and lower limit values are not corrected. In some cases, this may be a cause of deterioration in control performance, so care must be taken.

  As a result, according to the present invention, even if the response prediction considering the influence of the disturbance is corrected and the response prediction of the control variable is determined so as to change more slowly, the control performance such that the change amount of the manipulated variable increases. It becomes possible to suppress the deterioration of.

  The present invention is not limited to the embodiment described above, and many modifications and combinations can be implemented by those having ordinary knowledge in the art within the technical idea of the present invention. It is obvious. For example, in the above-described embodiment, the description has been given by taking the one-input one-output system as an example. However, the present invention is not limited to this, and if the processing (calculation) of the constraint correction unit is performed for each control variable, the present invention will be described. It can be applied to a multi-input multi-output system.

  DESCRIPTION OF SYMBOLS 101 ... Data collection part, 102 ... Response prediction production | generation part, 103 ... 1st response prediction correction part, 104 ... Target value trajectory generation part, 105 ... Restriction condition setting part, 106 ... Relative restriction correction part, 107 ... Manipulation variable calculation , 120 ... Control target process, 161 ... Restriction correction unit, 162 ... Second response prediction correction unit.

Claims (14)

An operation variable used for control by the control object, a control variable output by the control object, and a data collection unit for collecting disturbance information;
A response prediction generation unit that predicts a control variable that the control target will output in the future based on data including control variables, operation variables, and disturbance information collected by the data collection unit, and outputs a future response;
A response prediction correction unit that generates a corrected future response in which the influence of disturbance in the time series change of the future response output by the response prediction generation unit is corrected to a gentle change;
A target value trajectory generator for generating and outputting a target value trajectory of a control variable consisting of a time series of desirable values to be output in the future by the control object;
A constraint storage unit that stores an upper limit value and a lower limit value of a control variable to be output in the future by the control target;
An operation variable calculation unit that calculates an operation variable used in the future by the control target so that the corrected future response output from the response prediction correction unit approaches the target value trajectory, and outputs the operation variable to the control target.
Of the allowable range to the upper limit or the lower limit of the corrected future response, the allowable range is smaller for the response prediction correction unit that is smaller than the original allowable range from the future response to the upper limit or the lower limit. The model predictive control device , wherein the relative relationship between the future response or the corrected future response and the upper and lower limit values is corrected so as to suppress the above . The model predictive control apparatus according to claim 1,
Based on at least one of the future response predicted by the response prediction generation unit and the first corrected future response output by the response prediction correction unit, the future response predicted by the response prediction generation unit is corrected. And a relative constraint correction unit that calculates and outputs a corrected future response of 2 and a correction upper limit value and a correction lower limit value obtained by correcting the upper limit value and the lower limit value set in the constraint condition storage unit,
The relative constraint correction unit is configured so that the difference between the second corrected future response and the correction upper limit value and the correction lower limit value is larger than the difference between the first corrected future response and the upper limit value and the lower limit value. A model predictive control apparatus that calculates a corrected future response, a corrected upper limit value, and a corrected lower limit value. The model predictive control apparatus according to claim 1,
Relative constraint correction unit that calculates and outputs a correction upper limit value and a correction lower limit value obtained by correcting the upper limit value and the lower limit value set in the constraint condition storage unit based on a future response predicted by the response prediction generation unit With
The relative constraint adjustment unit, the difference between the upper limit value and the lower limit corrected upper limit value value Fixed and Fixed lower limit is set the corrected future response and the constraint condition storage unit, wherein the modified prospective response and the upper limit value and A model predictive control device that calculates a correction upper limit value and a correction lower limit value so as to be larger than a difference from the lower limit value. The model predictive control apparatus according to claim 1,
Based on the future response predicted by the response prediction generation unit, the future response predicted by the response prediction generation unit, and the corrected upper limit value and the corrected lower limit value corrected by the upper limit value and the lower limit value set in the constraint condition storage unit A relative constraint correction unit that calculates and outputs a value,
The relative constraint adjustment unit, the difference between the upper limit value and the lower limit value modification upper limit value and the corrected lower limit that fixes set in the constraint condition storing unit and outputted future response from said response prediction generation section, wherein modified so that Do greater than the difference between the future response and the upper limit value and the lower limit value, the model predictive control system and calculates the corrected upper limit value and the corrected lower limit value. The model predictive control apparatus according to claim 1,
Based on the future response predicted by the response prediction generation unit, a third corrected future response in which the future response predicted by the response prediction generation unit is corrected, and an upper limit value and a lower limit value set in the constraint condition storage unit And a relative constraint correction unit that outputs
The relative constraint adjustment unit, the difference between the third modification future response and the upper limit value and the lower limit value, the so modified greater than the difference between the future response and the upper limit value and the lower limit value, the third modified prospective A model predictive control device characterized by calculating a response. In the model predictive control apparatus according to any one of claims 1 to 5,
The relaxation range in which the manipulated variable calculator changes the tolerance range used when calculating the manipulated variable to be used in the future by the controlled object from the tolerance range based on the relationship between the future response and the upper limit value and the lower limit value. A model predictive control device characterized by being adjustable. The model predictive control apparatus according to claim 6, wherein
The model predictive control apparatus, wherein the relaxation range is adjustable in a section in which an operation variable to be used in the future by the control target is calculated. The model predictive control apparatus according to claim 6 or 7,
The model predictive control apparatus, wherein the relaxation range is adjusted based on characteristics of the control target. The model predictive control apparatus according to claim 8, wherein
The model predictive control apparatus, wherein the characteristic of the controlled object is a step response of the controlled object. A data collection step for collecting operation variables used by the controlled object, control variables output by the controlled object, and disturbance information;
A response prediction generation step of predicting a control variable to be output in the future by the control object based on data including the control variable, operation variable, and disturbance information collected in the data collection step, and outputting as a future response;
A response prediction correction step for generating a corrected future response in which the influence of disturbance in the time series change of the future response output in the response prediction generation step is corrected to a gentle change;
A target value trajectory generating step for generating and outputting a target value trajectory of a control variable consisting of a time series of desirable values to be output in the future by the control object;
A constraint condition storing step of storing an upper limit value and a lower limit value of a control variable to be output in the future by the control object;
An operation variable calculation step for calculating an operation variable used in the future by the control object so that the corrected future response output in the response prediction correction step approaches the target value trajectory, and outputting the operation variable to the control object.
Of the allowable range to the upper limit or the lower limit of the corrected future response, the allowable range is smaller with respect to the one that becomes smaller than the original allowable range from the future response to the upper limit or the lower limit by the response prediction correction step. The model predictive control method is characterized by correcting a relative relationship between the future response or the corrected future response and the upper and lower limit values so as to suppress the above . The model predictive control method according to claim 10, wherein
Based on at least one of the future response predicted in the response prediction generation step and the first corrected future response output in the response prediction correction step, the future response predicted in the response prediction generation step is determined. A relative constraint correction step of calculating and outputting the corrected second corrected future response and the corrected upper limit value and the corrected lower limit value obtained by correcting the upper limit value and the lower limit value stored in the constraint condition storing step;
In the relative constraint correction step, the second modified future response and the difference between the modified upper limit value and the modified lower limit value are larger than the difference between the first modified future response and the upper limit value and the lower limit value. A model predictive control method comprising: calculating a corrected future response, a corrected upper limit value, and a corrected lower limit value. The model predictive control method according to claim 10, wherein
Relative constraint correction step of calculating and outputting a corrected upper limit value and a corrected lower limit value obtained by correcting the upper limit value and the lower limit value stored in the constraint condition storage step based on the future response predicted in the response prediction generation step With
In the relative constraint correction step, the difference between the upper limit value and the modified modified the lower limit upper limit value and the corrected lower limit value stored in the constraint condition storing step and the corrected future response, the modified prospective response and the upper limit value and lower limit A model predictive control method, wherein a correction upper limit value and a correction lower limit value are calculated so as to be larger than a difference from the value. The model predictive control method according to claim 10, wherein
Based on the future response predicted in the response prediction generation step, the future response predicted in the response prediction generation step, the corrected upper limit value and the correction corrected by the upper limit value and the lower limit value stored in the constraint condition storage step A relative constraint correction step for calculating and outputting the lower limit value is provided.
The Relative constraints correction step, the difference between the response prediction generation upper limit value and modifying the corrected lower limit upper limit value and the corrected lower limit value stored outputted future response and in the constraint condition storing step in step, the modified model predictive control method characterized by calculating the so that Do greater than the difference between the future response and the upper limit value and the lower limit value, the correction upper limit value and corrected lower limit value. The model predictive control method according to claim 10, wherein
Based on the future response predicted in the response prediction generation step, a third corrected future response in which the future response predicted in the response prediction generation step is corrected, and the upper limit value and the lower limit set in the constraint condition storage step A relative constraint correction step for outputting a value and
The Relative constraints correction step, the difference between the third modification future response and the upper limit value and the lower limit value, the so modified greater than the difference between the future response and the upper limit value and the lower limit value, the third modified prospective A model predictive control method characterized by calculating a response.

Images