Control device, control method and computer readable medium for controlling plant based on prediction model and constraint condition

Patent metadata

Field	Value
Publication number	US-11112760-B2
Application number	US-201916387120-A
Country	US
Kind code	B2
Filing date	Apr 17, 2019
Priority date	Apr 19, 2018
Publication date	Sep 7, 2021
Grant date	Sep 7, 2021

Abstract

A control device (70) of a plant which controls operation amounts of a plurality of types of equipment constituting the plant, the control device includes a prediction model deriving unit (71) that derives a prediction model from which a monitoring target value is output, an adjustment amount calculating unit (72) that calculates adjustment amounts of operation amounts in the case where the monitoring target value becomes a desired value and calculates differences between the operation amounts of existing control and the calculated operation amounts as first adjustment amounts, a constraint condition setting unit (73) that sets constraint conditions based on a measurement value and operation conditions of the plant 1, and an operation amount calculating unit (74) that calculates second adjustment amounts to which the constraint conditions are applied and calculates a plurality of the adjusted operation amounts based on each of the calculated second adjustment amounts.

First claim

What is claimed is:

1. A control device of a plant which controls operation amounts of a plurality of pieces of equipment constituting the plant, the control device comprising:
a prediction model deriving unit that derives a prediction model to which a measurement value measured at the plant is input, and from which a monitoring target value which is a value to be monitored is output;
an adjustment amount calculating unit that calculates the operation amounts of a plurality of pieces of the equipment in a case where the monitoring target value which is to be output becomes a desired value, and calculates differences between the operation amounts of an existing control and the calculated operation amounts as first adjustment amounts;
a constraint condition setting unit that sets a constraint condition of the operation amounts of a plurality of pieces of the equipment based on the measurement value and an operation condition of the plant when the measurement value is measured; and
an operation amount calculating unit that calculates second adjustment amounts obtained by correcting the first adjustment amounts by applying the set constraint condition for each of a plurality of pieces of the equipment and calculates a plurality of the adjusted operation amounts based on each of the calculated second adjustment amounts,
wherein the operation amount calculating unit calculates a plurality of the adjusted operation amounts based on a plurality of the second adjustment amounts and operation amounts of the existing control and inputs a test signal to the operation amounts of the existing control.

2. The control device of the plant according to claim 1 ,
wherein the prediction model deriving unit derives the prediction model through multiple regression analysis using a linear prediction expression including partial regression coefficients and explanatory variables,
the adjustment amount calculating unit calculates the first adjustment amounts of a plurality of the operation amounts based on the partial regression coefficients of the linear prediction expression of the prediction model, and
the constraint condition setting unit sets the constraint condition such that a sum of the first adjustment amounts of the operation amounts is fixed.

3. The control device of the plant according to claim 2 ,
wherein the prediction model deriving unit derives the prediction model by selecting one of the explanatory variables having higher correlation than correlation between other explanatory variables and deleting the other of the explanatory variables.

4. The control device of the plant according to claim 2 ,
wherein the prediction model deriving unit calculates a distribution of the partial regression coefficients in a predetermined interval and calculates a credible interval of the prediction model from the distribution of the prediction model based on the distribution of the partial regression coefficients,
in a case where a width of the credible interval is equal to or less than a predetermined threshold, the adjustment amount calculating unit calculates the first adjustment amounts using the prediction model, and
in a case where the width of the credible interval exceeds the predetermined threshold, the prediction model deriving unit derives a new prediction model in a next predetermined interval.

5. The control device of the plant according to claim 1 ,
wherein the operation amount calculating unit sets the test signal so as to be a value within a predetermined range in which the test signal does not become a disturbance to the existing control.

6. A plant comprising the control device according to claim 1 .

7. A control method of a plant which controls operation amounts of a plurality of pieces of equipment constituting the plant, the control method comprising:
a step of deriving a prediction model to which a measurement value measured at the plant is input, and from which a monitoring target value which is a value to be monitored is output;
a step of calculating the operation amounts of a plurality of pieces of the equipment in a case where the monitoring target value which is to be output becomes a desired value, and calculating differences between the operation amounts of an existing control and the calculated operation amounts as first adjustment amounts;
a step of setting a constraint condition of the operation amounts of a plurality of pieces of the equipment based on the measurement value and an operation condition of the plant when the measurement value is calculated; and
a step of calculating second adjustment amounts obtained by correcting the first adjustment amounts by applying the set constraint condition, for each of a plurality of pieces of the equipment, and calculating a plurality of the adjusted operation amounts based on each of the calculated second adjustment amounts,
wherein the step of calculating the operation amounts calculates a plurality of the adjusted operation amounts based on a plurality of the second adjustment amounts and operation amounts of the existing control and inputs a test signal to the operation amounts of the existing control.

8. A non-transitory computer-readable storage medium storing a control program of a plant which controls operation amounts of a plurality of pieces of equipment constituting the plant, the control program of the plant causing a computer to execute the method comprising:
a step of deriving a prediction model to which a measurement value measured at the plant is input, and from which a monitoring target value which is a value to be monitored is output;
a step of calculating the operation amounts of a plurality of pieces of the equipment in a case where the monitoring target value which is to be output becomes a desired value, and calculating differences between the operation amounts of an existing control and the calculated operation amounts as first adjustment amounts;
a step of setting a constraint condition of the operation amounts of a plurality of pieces of the equipment based on the measurement value and an operation condition of the plant when the measurement value is measured; and
a step of calculating second adjustment amounts obtained by correcting the first adjustment amounts by applying the set constraint condition, for each of a plurality of the equipment, and calculating a plurality of the adjusted operation amounts based on each of the calculated second adjustment amounts,
wherein the step of calculating the operation amounts calculates a plurality of the adjusted operation amounts based on a plurality of the second adjustment amounts and operation amounts of the existing control and inputs a test signal to the operation amounts of the existing control.

Assignees

• Mitsubishi Heavy Ind Ltd

Inventors

• Ide Kazunari
• Hiroe Takaharu
• Sase Ryo
• Ikawa Yoshikatsu
• Ishiwaki Toshio
• Doi Tomohiro
• Kitayama Hiroshi

Classifications

• G05B23/024 Primary

  Quantitative history assessment, e.g. mathematical relationships between available data; Functions therefor; Principal component analysis [PCA]; Partial least square [PLS]; Statistical classifiers, e.g. Bayesian networks, linear regression or correlation analysis; Neural networks · CPC title

• G05B19/042 Primary

  using digital processors (G05B19/05 takes precedence) · CPC title

• G05B13/048

  using a predictor · CPC title

• G06N20/00

  Machine learning · CPC title

• G05B2219/2639

  Energy management, use maximum of cheap power, keep peak load low · CPC title