Oxygen Vacuum Swing Adsorption (O2VSA) is a relatively new process for the
production of 90-95% O2 from air using adsorbents in a pressure swing
cycle. Many adsorbents such as molecular sieve zeolites preferentially
adsorb nitrogen leaving the un-adsorbed oxygen to be recovered as product from
the exit of a packed bed. Removing the nitrogen is done by reducing the
bed pressure to below atmospheric, purging with some oxygen product, and
repressurising the bed with either product or feed or both. Thus a
sequence of steps is used with multiple adsorbent beds to produce a
"pseudo" continuous product stream of oxygen. During normal
operation of a O2VSA, the product composition, flow rate, and pressure, all
oscillate to a small extent. These oscillations are buffered by using an
oxygen product tank so that the customer receives a relatively steady flow rate
of constant purity oxygen at a relatively steady pressure. When such a
condition occurs, the system is said to have achieved cyclic steady state.
Unfortunately, cyclic steady state is an idealisation which seldom
occurs. A large variety of disturbances can and do affect the performance
of an O2VSA such as changes in ambient temperature, changes in the customer
demands (ie. attempting to take more or less oxygen), changes in valve
coefficients and therefore flow rates as the valves wear, and changes in
machinery performance. As a result of these disturbances, the product
composition, flow rate and pressure often start to diverge from their set points
and control must be taken to restore them to their required values. This
can be an exceedingly difficult task: there are a large number of variables
which can be manipulated (step times, valve positions) all or some of which may
affect the product composition, flow rate, and pressure. It is the goal of
this research project to understand the dynamic response of an O2VSA to
disturbances in customer demand and develop control strategies to allow
efficient return of the VSA to cyclic steady state at the desired setpoints.
To achieve this goal, both experimental and modelling strategies have been
adopted. An experimental O2VSA plant has been built to duplicate the
expected large scale behaviour. To achieve this, it is important that the
unit approaches adiabatic operation since the presence of large temperature
gradients in the bed could significantly affect the dynamic behaviour and
required control. The VSA built is a 2 bed unit, each bed consisting of
insulated PVC 10cm piping of height 1.8m. The beds contain two layers: a
prelayer of inert alumina and a main bed of LiX zeolite. It is the
presence of this prelayer which causes significant non-isothermal
behaviour.
The VSA is fully instrumented and PLC controlled and a wide
variety of possible cycles can be programmed. In our study, the feed
valve, product
valve, and purge valve are the main three variables to be manipulated to control
product purity, pressure, and flow rate. The response of the system to
load disturbances and the ability of these manipulated valves to restore the
system to its set point has been the focus of our study. A variety of
control schemes have been developed and tested. For more information on
this project see the publications or contact Paul Webley.