(113d) Improved Distillation Composition Control

The traditional means of controlling a distillation column product composition is via a PID-to-PID cascade, analyzer to temperature. Another approach is to use Model Predictive Control (MPC) to control the product analyzer value with a column temperature controller as the primary manipulated variable, i.e., with the temperature controller in cascade to the MPC.

Natural questions arise: “Are there simpler control options when an MPC is not justified, e.g., when constraints are not a significant issue” and “Can we do better than an analyzer to temperature cascade?”

As with any cascade loop, there is the requirement that the inner, secondary loop (the temperature controller) be sufficiently faster than the primary loop (the analyzer controller). Unfortunately, the dynamic response between the manipulated variable (or disturbance) and temperature and between the temperature and analyzer are not sufficiently different, making it difficult to achieve this requirement in practice (without significantly detuning the analyzer controller). As a result, when a composition disturbance occurs, the temperature controller responds first to correct for the temperature deviation and when the analyzer responds later (due to slower dynamics and analyzer cycle time) the analyzer controller responds by changing the temperature setpoint, which if often unnecessary and prolongs the return to steady state.

An alternative approach is to develop a model between the controller temperature PV (not SP) and the analyzer PV. In using this model for the controller the temperature SP is only changed when model offset occurs or when the analyzer SP is changed. The model and controller, which is akin to internal model control, can be implemented fairly simply in a DCS or PLC.

The details for developing and implementing this scheme are described along with guidance for selecting the control temperature(s).