Case Studies

Case Study: Controlling Product Composition

This customer needs to control product composition without knowing what the feed composition is an not having any control over the feed rate.

Company		Sarawak Shell (Shell Malaysia, a subsidiary of Royal Dutch Shell)
Product		Crude Oil
Location		Bintulu MLNG Processing Facility, North Shore of Borneo, state of Sarawak
Situation		Oil and gas comes onshore from platforms and the oil is sent to four stabilization towers. These stabilization units further remove gas from the raw, fizzy oil in preparation for storage and shipment by tanker to Japan. If the oil is stored with too much dissolved gas, the gas will come out of solution and tilt storage tank roofs, creating a significant problem. If intermediate hydrocarbons (light oil) are removed as a gas and sent to the gas processing facility, they are condensed and returned with a penalty processing fee. Additionally, there is an economic advantage to sell the intermediate hydrocarbons in the oil, as the value of oil is higher than gas. The feed rate and the composition to the towers is uncontrollable, at the whim of what comes from the ground out on the platforms. The process is reasonably complicated, containing high and medium pressure separators, feed splits, reboilers, recycles, etc. The customer wishes to produce oil at a "spot on" vapor pressure, +/- 0.25 psi of target but could only produce 53% of the product within specification.
Objective		Produce oil at a specified Reid Vapor Pressure (RVP) +/- 0.25 psi.
Method		Predictive system behavior models were built using historical data. These models predicted RVP ahead by 15 minutes using 14-20 process variables, including tower temperature, the selected control handle. Predictions were put on-line to create virtual sensors. Model-based optimization schemes were created to control each tower temperature, within constraints, simultaneously considering the dynamics of the 14-20 other process variables. The optimization schemes were implemented in closed-loop process optimizers to achieve the desired RVP 15 minutes in the future in real-time (multivariate predictive control). This temperature was sent to a Yokogawa distributed control system as a setpoint.
Result		Product conformance to specification increased from 53% to 95% as depicted in this graphic:
Reference Available?		No. Most staff have moved on. Some staff that were involved at the time are available for contact.