On-line pulp mill production optimization

We consider the planning of pulp production for large sulphate and sulfite mills. The production planning problem is formulated as a non-linear program (NLP) given a process model of the mill as constraint. The objective is to minimize the usage of expensive chemicals and to minimize the (squared) deviation from specified set-points for selected variables, e.g. production, tank-level and chemical composition of the cooking liquor. The problem formulation also considers upper and lower limits on variables and limitations in the derivative of production related variables. The NLP, which involves several tens of thousands of variables, is solved using algorithms for large-scale optimization. To provide a correct initial state of the process model, a moving horizon estimation is done to estimate the current state of the process.