Performance monitoring for model predictive control maintenance

An economically motivated performance measure is proposed for use with model predictive control applications. The typical case handled is that you have a trade-off situation between two conflicting goals: (i) the need to be close to a constraint, since this will give good production economy, and (ii) the need to avoid constraint violation, since that would mean producing an inferior quality, exceeding environmental constraints (thereby requiring a fee), or having some other costly drawback. The proposed performance measure uses consumption and production rates in the process and the corresponding cost and benefit factors, combined with the risks of constraint violations and their associated costs. There is a discussion on the hard task of choosing threshold values, and the approach is illustrated by applying it to a couple of examples. One of the examples is the Autoprofit benchmark pulp digester model.