Research Area 1: Monitoring and redundancy for high reliability and robustness

The biggest obstacle to the use of online optimizing control is the need to guarantee a reliable and safe operation of the controlled plant over long periods of time where unforeseen disturbances and gradual changes of the behavior of the plant will occur. Failure to realize the last 0.1% of the possible economic gain is by far not as important in practice as to avoid one upset in three months that is caused by a malfunction of the control scheme. Reliability and failure detection is also a crucial element of building trust on the side of the operators, and in RA3 we will also investigate how the operators can be enabled to contribute to the monitoring of the control scheme.

The issue of reliability and robustness is complicated because the possible causes of failures are manifold. Single elements of optimization-based control schemes such as the optimization algorithm, the state estimator, sensors and actuators or low-level controllers may fail, the constraints may be too tight so that a feasible solution does not exist, and the deviations between the models that are used in the model-based control scheme and the behavior of the plant or the presence of large disturbances may lead to malfunctions. We will critically review all elements of the overall control scheme from this point of view and develop schemes with a high resiliency due to built-in monitoring and correction capabilities. We will not deal with fault detection on the sensor and actuator level but assume in this research that information on sensor and actuator malfunctions is provided by the low-level control system and indicated to the optimizing controller which has to adapt online to cope with these errors.