The BP Amoco group uses multi-variable controls extensively in its refineries, but successful application to upstream operations offshore is just now breaking new ground. Honeywell Robust Multivariable Predictive Control Technology (RMPCT) was installed in March 1999 on BP Amoco Norway's Ula field, becoming one of the world's first offshore platforms to implement advanced control optimization.
The project has generated a sustained production increase of 2%, and enabled BP Amoco to maintain optimal fuel gas to Ula's turbines, resulting in a savings in carbon dioxide tax of $1 million/year.
Ula is located in the southwestern corner of the Norwegian Continental Shelf and was commissioned in 1986. At its peak, it achieved crude oil production of 150,000 b/d. It is now in decline and today produces 30,000 b/d. However, with new production and reservoir techniques plus tiebacks from surrounding fields, Ula is expected to remain economic through to 2010.
One of the features of late-life operation is that produced water, which comes up from the reservoir with the oil and gas, has to be cleaned and disposed of. Ula currently produces 120,000 b/d of water. Previously, the only disposition method was to make this water as clean as possible and dump it overboard. However, in 1995, BP Amoco Norway engineers began re-injecting produced water into the reservoir. In fact, Ula was one of the first offshore operations in the world to do so.
In the early 1990s, the Norwegian government decided to impose an environmental tax based on emissions of carbon dioxide to the atmosphere. The tax is levied on the volumetric measurement of fuel and flare gas. This tax has today risen to a level which makes it a significant part of Ula's operating budget, presently costing around $8 million/year.
The above factors create complex interactions between process and environ mental goals, which are difficult to control using conventional techniques. To improve productivity a new integrated control and safety system supplied by ABB was installed on ULA in 1995. Although this system helped, the control and production engineers felt that more could be achieved using advanced specialized control techniques.
In 1996, the then-BP Group launched major programs aimed at finding new technologies that could help its upstream businesses. One initiative, dedicated to promoting the opportunities presented by advanced control, was spearheaded out of the BP Research Centre in Sunbury, England.
Ula was perceived as an ideal test bed for advanced control and a feasibility study was performed in 1997. This identified ways of increasing the oil production rate and hence the profitability of the platform, because the application of multivariable predictive control would allow the process to be run closer to the limits set by the complex interactions. In addition, it was predicted - correctly - that the project could also contribute to reducing the cost of carbon dioxide tax paid on fuel gas.
Honeywell Hi-Spec Solutions was awarded the contract to deliver advanced control solutions to BP Norge (later BP Amoco Norway). The project team, consisting of engineers from Norway, Sunbury, and offshore staff worked together with Honeywell for over a year to develop the system and interfaces.
The producing wells on Ula are connected to the wellhead manifold and then to either the test separator or the HP separator. For the greater part of the time, the test separator is used as a normal production vessel. Separation of the produced water from the oil occurs mainly in the HP separator, with the water being routed through the hydrocyclones (to reduce the oil levels further) to the de-gassing drum. Oil separated in the hydrocyclones flows back to the closed drain drum, from where it is pumped back to the MP separator.
Oil from the HP separator flows under level control to the MP separator. Again, any emulsified or entrained water is separated at this point and passed through the MP hydrocyclones to the de-gassing drum. Oil separated in these hydrocyclones also flows back to the closed drain drum. The oil then flows to the MOL booster pumps from where it is pumped to Ekofisk.
The produced water in the de-gassing drum then passes through a set of coolers and is mixed with de-oxygenated seawater, subject to a temperature constraint, before being re-injected back into the wells.
Gas evolved in the HP separator passes through a cooler and into the HP scrubber. Condensate knocked out at this point flows back, under level control, to the MP separator. The gas then passes through the dehydration unit and onto the HP compressor from where it is either used to top up the fuel gas main, used for gas lift or is passed onto the WAG (water alternate gas) compressor for re-injection back into the well.
Gas evolved in the MP separator is passed through the MP gas cooler and into the MP gas scrubber, where condensate is knocked out and pumped back into the MP separator. The gas is then compressed and passes, under back-pressure control, to either the fuel gas main or the HP scrubber. Due to the Norwegian legislation on carbon dioxide tax, the use of "heavier" gas for the supply of fuel gas is preferable, since the tax is levied upon the normalized quantity of gas combusted, not based upon its molecular weight.
The platform currently is well limited and running with the production chokes wide open on all wells. Choke by-pass valves also are open where provided. The current limitation on the process is not dictated by the topside process capacity, but rather on the ability of the well to produce.
Prior to implementation of the multivariable control, the control strategy was focused on single loop basic control. Primarily, loops had been tuned for load disturbances, with relatively few control cascades. Any optimization of the process was performed by the operator, in response to his knowledge of the process and conditions at the time.
After a study performed by BP on the Ula platform, opportunities were identified for increasing the oil production rate and hence the profitability of the platform. This could be achieved by decreasing the overall operating pressure of the process, and due to its interactions, this would be an ideal application of multivariable control. It was estimated that this could yield increases in oil production rates in the region of 1-2%.
However, in order to achieve these benefits, there were process configuration changes required to enable the application of multivariable control to proceed. These changes were made during the Ula platform shutdown in August 1998.
For the RMPCT control scheme to be implemented, some changes were required on the basic proportional integral derivative (PID) controllers on the distributed control system (DCS). The Ula platform is controlled using the ABB Master DCS. To enable the correct changeover between RMPCT control and operator control, the Pidcon block was utilized. This enabled the operator to have full normal control over the block, but also enabled the use of a cascade mode, to which RMPCT would write when in control.
The RMPCT software runs on a 400 MHz Windows NT 4.0 machine. This is interfaced to the Honeywell process history database (PHD) which connects via an Ethernet connection to the ABB Master control system using a UNIX based IMS as the gateway.
Time was spent tuning the basic controllers and ensuring that the basic control philosophy was correct. Once this was completed, step-testing activities commenced with the operating pressure of the process manually lowered by the operator. This gave a clear idea of where the constraints on the operation would lie, when the RMPCT controller was continually trying to push the operating pressure down.
The step-tests were carried out whenever the process was steady enough to identify responses. The PHD was used to store the data.This was then extracted ready for the model identification at the end of the step-testing period. During the step-tests, potential manipulated variables were step-tested up to eight times, depending on the clarity of response observed.
The models were identified using the Honeywell advanced process control ( APC ) development environment and the RMPCT Identifier Release 170.00. The configuration was then built using the RMPCT point builder and these files were then used to generate the PHD tagload file. Once the PHD points were built, the RMPCT controller took two days to commission.
Models were commissioned in isolation, so that the RMPCT controller action on each model could be identified. Having the RMPCT controller built onto the PHD greatly aided the commissioning activities, since the controller predictions could be compared to the actual process value with no additional configuration work. Tools such as the Honeywell process trend package made this task easier and more intuitive.
The wealth of information generated by the RMPCT controller made building succinct, meaningful DCS resident schematics difficult. In the end, it was decided that the information available in the Honeywell profit viewer schematics would be much more meaningful and also gave built-in features such as access levels and multiple user access. The profit viewer schematics being designed around the existing RMPCT schematic format, the amount of training required for the operators and engineers was minimal.
The schematics combine the best of the proven schematic format together with Windows features such as the "Tooltips" text box feature, double clicking on fields and as such, users can navigate the schematics with little training. Because the schematics are dynamic (they change as the controller configuration changes), adding and removing manipulated or controlled variables to the controller is a minor task. This assisted in reducing the commissioning time for the controller.
The advanced controller was commissioned in March 1999 and has proven to be a great success. The project has not only given a sustained production increase of some 2% but has also enabled Ula to maintain optimal fuel gas to its turbines, resulting in a saving of carbon dioxide tax of some $1 million/year.
Experience so far is that the technology has been well received by the offshore staff. They find the profit viewer both intuitive and easy to operate. It has required only minimal training of operators who have had no previous experience of advanced control. The profit viewer has also given operations staff the opportunity to better understand the process goals possible and which constraints are stopping them achieving these.
Another possibility could be to extend the profit viewer to the onshore support office, via a new high bandwidth fiber optic link. This will enable the land based production engineers to monitor the profit controller's performance. Honeywell's support staff will also be able to maintain and modify the system from shore. The project team also believes the techniques used on ULA are directly applicable to the whole of BP Amoco's upstream business.