Incremental data driven modelling for Plug and Play Process Control

This paper studies the data driven update of a model for a system where the number of inputs or outputs increased. Often existing control systems are equipped with an additional sensor or actuator to improve performance. If a good model for the present system is available it is advantageous to only estimate the additional part while keeping the present model, compared to estimating the whole model from scratch. The capabilities with convex methods are investigated. It is shown that model updating for static sensor/actuators can be done consistently for the deterministic part. The stochastic part is far more complicated and here convex methods gives a approximate solution. The total solution is demonstrated by simulation to improve state prediction and control performance.