WHAT is SEPTIC?– Tool for on-line estimation and Predictive control
(MPC) WHY?
– Statoil R&D has several present and future projects within model based predictive control, including development of
ƒ fundamental mathematical modelsƒ control algorithmsƒ estimatorsƒ methods for Real Time Optimisation (RTO)
– Increased customer focus on state estimation– A standard tool reduces the costs with each
implementation, and supports knowledge transfer HOW?
– Software package developed in C++– User interface with configuration files
(measurements, inputs, outputs, parameters, couplings, tuning) and graphical presentation
– Makes use of external- or self developed models
Developed byTEK F&T OGF Process Control
SFA separation train:
Control hierarchy for typical process plant
Basic control loops(PID, FF,..)
Supervisory control/model based control (e.g. MPC)
Stationaryoptimization
(RTO)
Planning
PCDA
Supervisoryprocess controlcomputer
Trip on Snorre – one minute later this leads to lack of feed at SFA.
MPC at SFA is activated. New setpoints for the indicated FIC and LIC are optimised and calculated such that the sepration train avoids trip.
Benefits:– Approximately 20 SFA trips due to Snorre
trips before implementation. None afterwards.
– 1-2 MNOK increased income for each trip avoided
Mongstad refinery crude oil tower - a particularly complex distillation column:
An MPC simulation model of the tower, which when combined with measurements of pressure, temperature and liquid loading in the column, has helped to make the column more robust with respect to accepting varying raw material quality and other aspects that are detrimental to processing
Benefits:– Better quality control of the products (i.e. heavy naphtha,
kerosine and gas oil) which, in turn, is beneficial to further processing and the blending of end products.
– Increased flexibility to switch between various products
Model Predictive Control in Statoil
• In-house tool Septic, Statoil Estimation and Prediction Tool for Identification and Control
• 25 MPC applications with Septic within Statoil
• Experimental step response models, built-in functionality for model gain scheduling
• Flexible control priority hierarchy
• Quality control by inferential models built from laboratory data or on-line analysers
• DCS/PCDA interfaces currently in Septic: Honeywell TDC3000 (CM90 on Vax computer), ABB Bailey via InfoPlus (AspenTech), ABB Bailey via ABB OPC server, ABB Bailey via Matrikon OPC server, Kongsberg Simrad AIM1000 (integrated).
• Septic also supports mechanistic type models, generally nonlinear models, for applications with wide operating regimes.
Refining and Gas plant applications• RCCU reactor/regenerator section (1)
• RCCU main fractionator (1)
• CDU atmospheric distillation (1)
• CDU feed preheat heat exchanger network (MPC)
• CDU feed preheat heat exchanger network (RTO)
• CDU heater pass balancing
• Coker Unit main fractionator (2 drums, 24-hours cycle) (1)
• CokerNaphtha/CokerLGO splitter (1)
• Visbreaker feed maximisation (1)
• Visbreaker main fractionator (1)
• LPG/Naphtha splitters (4)
• C3/C4 splitters (3)
• iC4/nC4 splitter (1)
• Medium/Heavy Naphtha splitter (1)
• Light/Medium Naphtha splitters (3)
Hva oppnås med bedre regulering?
Ved redusert varians, kan settpunkt flyttes nærmere øvre spesifiserte verdi
MPC: Generelt system
prosessu
v
yxPådrag, MVR
Forstyrrelse, DVR
Regulert variabel, CVR
tilstand
med bibetingelser tidsvektorer:
MPC: Generelt system
Prediksjonshorisont
Nåtid t
Framtidige måling, "optimal" regulering(closed-loop respons)
Framtidig "optimalt" pådrag
Interpolert "optimalt" pådrag fra forrigeberegning
Framtidige måling basert på pådrag fra forrige beregning (open-loop respons)
Referanse
Control priorities
• MV ROC Limits
• MV High/Low Limits
• CV hard constraints
• CV soft constraints, CV setpoints, MV ideal values: Priority level 1
• CV soft constraints, CV setpoints, MV ideal values: Priority level 2
• CV soft constraints, CV setpoints, MV ideal values: Priority level n
• CV soft constraints, CV setpoints, MV ideal values: Priority level 99
Sequence av steady-state QP-solutions all priority levels handled
Then a single dynamic QP to meet the possibly adjusted steady-state goals
Open loop response is predicted by nonlinear model
–MV assumption : Interpolation of optimal predictions from last sample
Linearisation by MV step change
–One step for each MV blocking parameter (increased transient accuracy)
QP solver as for experimental models (step response type models)
Closed loop response is predicted by nonlinear model
–Compute linearisation error (difference open-loop + QP from simulated nonlinear closed-loop response)
ƒAbove threshold ---> closed-loop to "open-loop" and iterate solution
–QP solution ---> defines line search direction with nonlinear modelƒPossibly closed-loop to "open-loop" and iterate
MPC – Fundamental models (first principles)
MPCT101 – Crude tower Model Predictive Control application
MPCT101 – Crude Tower – Manipulated, Controlled and Disturbance Variables
MPCSPLT – T-108 (LPG/Naphta), T-112 (Light/Medium Naphta), T-113 (Light/Medium Naphta)
T-108 – Manipulated, Controlled and Disturbance variables
T-112 - Manipulated, Controlled and Disturbance variables
T-113 – Manipulated, Controlled and Disturbance variables
T-112 and T-113, bottom C6-, inferentials (red) and analyzers (AI063 and AI104)