Jul-2020

Cross-unit APC boosts downstream performance

Manipulating advanced process control in a crude unit reduced constraints in a downstream diesel hydrotreater

AZURA Binti AZAHAR and Siti Sarah Ahmad Nadzri, Petronas
Y ZAK FRIEDMAN and SEUNGYUN NAM, Petrocontrol

Article Summary

How difficult is it for advanced process control (APC) to manipulate CDU (crude unit) parameters in order to alleviate constraints of a downstream diesel hydrotreating unit (DHT)? Theoretically not too complicated, but practically difficult because these units are operated by two different operators. The CDU operator’s first priority is to handle CDU constraints while maximising production of the more valuable products (see Figure 1). Indeed, there was an APC application in place to help accomplish such economic objectives. As it happens, at Petronas Melaka refinery such a strategy in isolation may cause problems in the DHT unit, forcing a throughput cut, costing the refinery dearly in lost premium diesel production. Where are the APC benefits then?

Management asked the Melaka APC team to mitigate this conflict, adding DHT feed constraints to the CDU APC application, and this article is about how such an order could be accomplished.

Problem statement
Melaka’s 180000 b/d CDU2 is a high sulphur crude unit, feeding a downstream DHT, a hydrocracker, and a delayed coker. This modern high conversion complex (see Figure 2) produces mostly gasoline, jet-fuel and low sulphur diesel oil. It is important to keep all of these units working seamlessly, or conversion or throughput may suffer. Figure 3 shows the connection of interest here. Light gasoil (LGO) goes to the DHT either directly or indirectly via an intermediate storage tank. Coming into the DHT, LGO is filtered to protect the catalyst. There are two filters in parallel, one active while the other is being backwashed and then in standby. Filter switching should take place no more than once per day. However, feed flow from the tank is often contaminated by a slurry of rust, plugging up the filter, speeding up filter switching sometimes to three times a day, to the point that DHT throughput must be cut in order to reduce switching frequency. Not only is frequent filter switching a major operating inconvenience, throughput reduction also costs dearly in lost premium diesel production.

Can we just simply increase the DHT hot feed draw and reduce tank rundown? It turns out that there are hydraulic constraints. The hot feed valve becomes saturated, and trying to maximise total LGO draw reduces the pump head, worsening the problem. The way to alleviate this situation is to reduce total LGO draw, sending the excess diesel material either down to HGO and to the vacuum unit or to kerosene, provided kerosene is still on specification. Sometimes we are forced to increase the flow of overhead naphtha, an undesirable product, to reduce LGO yield. This is a complex multivariable constrained optimisation problem.

That was the driving force for configuring the CDU APC to consider DHT constraints. Ideally, the CDU APC should continue to maximise diesel production, but without running much of it down to the cold tank.

Adding to the complexity is light coker gasoil (LCGO), which is also fed to the DHT. LCGO is only 15% of DHT feed, but it fluctuates with coker drum switches, sometimes dropping from 15% to 10% of feed for two hours or so, and that shortage is made up by dirty cold feed from tank. This scenario begs for a coker APC drive to minimise LCGO fluctuations. Indeed, we have implemented such control logic, and managed to reduce the fluctuations by letting coker unit levels rise and fall. Nonetheless, there is not enough LCGO inventory volume in the coker to eliminate the fluctuation completely. That coker APC scheme would be the subject of another article.

APC history
Melaka has invested heavily in APC and presently all major units are under APC control. There is a sizable team in place to handle day to day maintenance as well as the occasional APC revamp or new applications. APC is much appreciated by operators and refinery management, and it typically runs at 90% service factor.

CDU2 APC was implemented in 2003 and has remained in closed loop almost continuously since then. It runs on RMPCT (Honeywell’s multivariable controller) plus inferential control models based on a Petrocontrol CDU package called GCC. GCC works to identify the true boiling point (TBP) curve of the crude being run from column measurements, and from the crude TBP curve it estimates product properties. GCC is a reliable, well tested, first principles inferential package, and using this package permits the APC to continue working during crude switches. We would not describe GCC further here except to say that several papers have been published about its performance,1-9 and one of these5 describes our initial CDU2 APC implementation in Melaka. Over the years, CDU2 underwent revamps and process changes, and the APC has also been revamped to keep it current.

DHT APC was implemented in 2013, primarily in order to control diesel flash point, again a Petrocontrol inferential model. DHT APC was not the main carrier of this cross-unit optimisation drive and hence it will not be covered further in this article.
 
Control and manipulated variables
The manipulated variables are typical of CDUs with one addition:
-    Top temperature, controlling naphtha cut point
-    Side draw flows, controlling side product cut points
-    Stripping steam, controlling kerosene flash-point
-    Other MVs because the actual application is more complicated than our current description
-    The DHT hot feed flow is added; this can be manipulated below the point of valve saturation.

The control variables are also typical of CDUs with some additions:
-    Product 95% points
-    Kerosene freeze point
-    Kerosene flash point
-    Top temperature NH4Cl sublimation point; top temperature must be kept above sublimation point to keep the column top trays clean. To some extent, controlling sublimation point causes kerosene flash point giveaway.
-    Other CVs because the actual application is more complicated than this current description
-    DHT hot feed valve position; saturation is to be avoided.
-    LGO rundown to tank, to be minimised
-    And most important, actual DHT cold feed flow, which directly affects filter switching frequency