gPROMS activities: Optimization
Finding the optimal answer directly rather than by trial and error
Optimization is a key technology for process organisations to create value and competitive advantage in process design and operations.
In particular, large-scale optimization based on high-fidelity models has the ability to create significant value from ‘already-optimized’ processes.
gPROMS® family products take full advantage of the optimization capabilities in the gPROMS platform, which make it possible to apply rigorous optimization to the design and operation of individual unit operations (e.g. reactors or distillation columns), plant sections (e.g. multiple distillation column sequences), entire plants integrating reaction, separation and utility sections, or even multi-site applications.
The power of gPROMS’s equation-oriented approach and resulting much faster solution times open the door to a wide range of advanced optimization applications that have not been feasible in the past.
Optimization – what does it mean?
“Process optimization” is a much-used term in the process industries. However, it is often used simply to mean ‘process improvement’, which is usually achieved by manually varying some equipment parameters and process conditions to improve the design or operation.
In the gPROMS context, optimization means formal mathematical optimization, where the optimizer searches the decision space for the combination of decision variables that give the best-possible results.
gPROMS platform optimization capabilities
Key capabilities of the gPROMS platform’s optimization facilities include:
- Steady-state optimization. Determine the optimal values of multiple decision variable such that the value of the objective function (typically an economic function) is maximized or minimised.
- Dynamic optimization (sometimes referred to as optimal control). Optimize the dynamic or transient behaviour of a system – for example, to minimise the time for a batch process to reach a certain state subject to constraints.
Because of the power of its equation-oriented framework and optimization solvers, gPROMS allows many decision variables to be varied simultaneously. Plant-wide optimization studies have included up to 50 continuous and integer decisions at a time; multi-site optimizations typically involve 50–100 decision variables.
gPROMS platform optimization – how it works
To define an optimization run in gPROMS, you need to create an objective function (or, alternatively, simply select a suitable variable from within the model) to maximize or minimise, then select decision variables and, optionally, define constraints.
Objective function
The objective function is defined by any equation of the form “obj_fun = expression“, where the expression can be composed of any variables in the model. The objective function is often an economic objective that sums the values and costs.
Decision variables
Any specified variable in the model can be used as a decision variable. Decision variables can be:
- Continuous variables, in which the value can vary in a continuous manner over the course of the optimization (e.g. a distillation column diameter)
- Integer variables, in which the decision variable may take only integer or discrete values (e.g. standard pipe diameters, or number of stages in a distillation column).
Constraints
It is possible define many different types of constraint, including final and interior-point constraints in dynamic optimization.
