Dynamic simulations suggest that the three control modes are each stable for a wide range of process disturbances despite the cogent process non-linearity and bidirectionality. Thus, the proposed integrated control scheme allows for the column operation to adequately reflect a changing economic environment that could require the process objectives to be regularly revised.
The control objectives were derived from the process objectives in a manner which maximised the process linearity and provided effective control loops, and the resulting control structure was shown to be capable of maintaining the process objectives within acceptable limits following perturbations in the feed rate and composition. Although accurate regulatory control of neither the ETBE product purity nor the reactant conversion was achieved (primarily due to measurement difficulties and the process bidirectionality), a mechanism was provided to manually increase or decrease the desired qualities.
Control mode 1 provided control of both the ETBE purity and the isobutene conversion via two composition control loops. Strong loop interactions were predicted by the high RGA values and this reduced the dynamic responsiveness of the control system. However, key process disturbances were suitably rejected by the controller and the process could be manipulated satisfactorily via the available set-points. The second control mode maximised the reactant conversion while retaining control of the purity. This was achieved by operating the reflux control loop in manual (at a value near the equipment constraints) and resulted in a much more responsive control system as the loop interactions were eliminated. Control mode 3 successfully maximised both the ETBE purity and the conversion, although purity loop was given a higher priority. This had a slightly detrimental effect on the conversion that was achieved because the operating point for maximum purity and maximum conversion do not coincide exactly. Process fluctuations also increased since the purity set-point was automatically adjusted with feed rate changes. However, the controller meets the process objectives.
At this point (i.e. after a satisfactory regulatory control system has been implemented), more advanced control applications can be considered to further improve the control performance. The process is a good candidate for multivariate predictive controllers (e.g. dynamic matrix control) as the control objectives can be changed continuously according to an overall economic optimisation. This type of controller also allows process constraints to be directly incorporated into the control scheme and can be extended to provide optimisation for interacting process units. However, the process bidirectionality must be considered at all times as the poor selection of control loops and variable pairings could easily result in an unstable controller.
It is important to realise that the effectiveness of the MPC is directly related to the effectiveness of the underlying linear control system which itself is strongly influenced by the choice of controlled and manipulated variables and the control structure. MPC is not effective for systems with: order of magnitude changes in process gains; sign changes in process gains; and highly unusual step responses. Reactive distillation commonly displays all of these attributes so that the performance of a MPC could be severely retarded by poor implementation. MPC is becoming increasingly common in the oil and gas industries so that it is often regarded as a panacea for all control problems. It is not! Often, comparable performance is achievable with well designed and well tuned PI controllers.
Was this article helpful?