CONTROL SYSTEM

• US 20120065746A1
• Filed: 10/21/2011
• Published: 03/15/2012
• Est. Priority Date: 04/22/2009
• Status: Active Grant

• Current Assignee: Steag Powitec GmbH (KSBG Kommunale Beteiligungsgesellschaft GmbH & Company KG)
• Original Assignee: Powitec Intelligent Technologies GmbH
• Inventors: Wintrich, Franz, Stephan, Volker, Schaffernicht, Erich, Steege, Frank-Florian
• Granted Patent: US 8,340,789 B2
• Time in Patent Office: Days
• US Class Current: 700/48
• CPC Class Codes: F01K 13/02 Controlling, e.g. stopping or starting
• CPC Class Codes: F23G 5/50 Control or safety arrangements
• CPC Class Codes: F23N 2223/44 Optimum control
• CPC Class Codes: F23N 2223/48 Learning / Adaptive control
• CPC Class Codes: F23N 2900/05006 Controlling systems using neuronal networks
• CPC Class Codes: F23N 5/00 Systems for controlling combustion regulating fuel supply F23N1/00, regulating air supply or draught F23N3/00
• CPC Class Codes: G05B 13/027 using neural networks only
• CPC Class Codes: G05B 13/04 involving the use of models or simulators
• CPC Class Codes: G05B 2219/31354 Hybrid expert, knowledge based system combined with ann
• CPC Class Codes: Y02E 20/12 Heat utilisation in combustion or incineration of waste

Abstract

A control system (1) for a complex process, particularly for controlling a combustion process in a power plant, a waste incinerator plant, or a cement plant, has a controlled system (14) and at least one controller (36), wherein the control system (1) is divided hierarchically into various levels (10, 20, 30, 40). The first level (10) represents the complex, real process to be controlled and is implemented by the controlled system (14). The second level (20) represents an interface to the process and is implemented by a process control system. The third level (30) represents the control of the process and is implemented by the at least one active controller (36). The fourth level (40) represents a superordinate overview and is implemented by a principal controller (44).

Claims

1. A control system for controlling a complex, real process of a controlled system, comprising:

a process control system;

a set of controllers comprising

controllers which were previously active, and

at least one active controller;

a principal controller configured for

training simulators that are for simulating the controlled system,

developing the set of controllers,

testing the set of controllers on the simulators in order to find a best controller, and

comparing the best controller with the at least one active controller for determining whether to replace the active controller with the best controller,

wherein the control system is hierarchically split into a plurality of levels comprising

a first level that is implemented by the controlled system and represents the complex, real process that is to be controlled,

a second level that is realized by the process control system and represents an interface to the complex, real process that is to be controlled,

a third level that is realized by the at least one active controller and represents control of the complex, real process that is to be controlled, and

a fourth level that is realized by the principal controller and represents a superordinate supervisor.

2. The control system as claimed in claim 1, wherein the at least one active controller is based on a recurrent neural network.

3. The control system as claimed in claim 1, comprising a feature extractor configured for:

performing image processing for the controlled system, and

ascertaining features from the image processing.

4. The control system as claimed in claim 1, wherein the at least one active controller is configured for ascertaining a manipulated variable using control error between an actual value (y) from the controlled system and a setpoint value.

5. The control system as clamed in claim 1, wherein the principal controller is configured:

for individually training the simulators in evolutionary fashion using a recurrent neural network; and

so that, for each of the simulators,

the training comprises defining a fitness for being optimized, and

the defining of the fitness comprises determining a discrepancy in a forecast for an actual value from a real actual value.

6. The control system as claimed in claim 1, wherein the principal controller is configured so that the testing of the set of controllers on the simulators comprises a search being performed to identify a controller which, on average over the simulators best matches the forecasts of the simulators in a time profile to profiles of setpoint values.

7. The control system as claimed in claim 6, wherein the principal controller is configured so that fitness of a controller from the set of controllers is obtained from simulated control errors of the controller in closed control loops on the simulators.

8. The control system as claimed in claim 1, wherein the principal controller is configured so that the controllers which were previously active are excluded from evolutionary search strategies for new controllers in the set of controllers

9. The control system as claimed in claim 1, wherein the principal controller is configured for:

training the simulators in cycles, and

testing the set of controllers on the simulators.

10. The control system as claimed in claim 1, wherein:

the at least one active controller comprises a plurality of active controllers; and

the control in the third level is split among the plurality of active controllers into subareas which have associated subprocesses in the complex process.

11. The control system as claimed in claim 10, wherein the principal controller is configured so that the split of the control among the plurality of active controllers changes dynamically by way of automatic problem decomposition.

12. The control system as claimed claim 1, wherein artificial data points are inserted in a data space for the states of the controlled system in unoccupied or sparsely occupied regions.

13. The control system as claimed in claim 12, wherein artificial data points are also inserted in regions of untested control actions in the data space area with known process situations and/or in data space areas with totally unknown process situations.

14. The control system as claimed in claim 1, wherein the complex, real process of the controlled system is a combustion process in a power plant, a waste incineration installation or a cement works.

15. The control system as claimed in claim 1, wherein the principal controller is configured:

for searching for new controllers in the set of controllers; and

so that the controllers which were previously active are excluded from the searching for the new controllers.

16. The control system as claimed in claim 5, wherein the principal controller is configured:

for searching for new controllers in the set of controllers; and

so that the controllers which were previously active are excluded from the searching for the new controllers.

17. The control system as claimed in claim 8, wherein the principal controller is configured so that the controllers which were previously active are excluded from the search.