VI Oil and Gas Production Optimization Workshop Dynamic

VI Oil and Gas Production Optimization Workshop Dynamic Modeling and Advanced Control for Offshore Platforms Argimiro R. Secchi Chemical Engineering Program – COPPE Universidade Federal do Rio de Janeiro Technological Center, Rio de Janeiro – RJ Solutions for Process Control and Optimization Rio de Janeiro 26 Apr 2017

Outline • Challenges for Offshore Production • Advanced Process Control • Model Predictive Control • APC in Offshore Platforms • Final Remarks 2

Challenges for Offshore Production Lula 250 km - Ultra-deep water - Faraway from coast - Pre-salt layer - High CO 2 content - Stringent regulations - High costs Source: PETROBRAS 3

Challenges for Offshore Production - Complex networks for well allocation, gas/water injection, and gas lift - Multiphase flow (gas, oil, water, sand, hydrate) - Lack of instrumentation - New technologies for subsea processing - Difficult maintenance - Automation and Remote operation Source: FMC Technologies 4

Challenges for Offshore Production oil flowrate - Time-varying process (gas-oil -ratio, water cut, gas/water coning) - Nonlinearities (variable gain, output saturation) - Severe disturbances (slugs) - High uncertainties gas lift flowrate flow instability Data from offshore platform 5

Challenges for Offshore Production Plenty of room for Modeling and Soft Sensors Advanced Process Control Process Optimization 6

Challenges for Offshore Production Targeting Ø Ø Ø Ø Maximize Production (oil & gas) Ensure Quality Specification (oil, water & gas) Minimize Losses (flare & TOG) Minimize Energy Consumption (heat & power) Minimize Operational Costs (maintenance & shutdowns) Minimize Process variability Ensure Process Stability and Safety Flow Assurance (prevent: hydrate, paraffins, asphaltenes, fouling) 7

Process Control Hierarchy Planning and Scheduling Plantwide computer Lack of room for computer in the platform Real-Time Optimization weeks hours Process computer Advanced Process Control minutes DSC Regulatory Control seconds Process 8

Advanced Process Control (APC): is a term that can include a range of methodologies, including model predictive control (MPC), fuzzy logic, statistical control, etc. The common objective is to find a way to manage complex interactions within a process better than traditional regulatory control. 9

APC & RTO Objectives • Maximize production • Ensure product specifications • Minimize energy and water consumption • Minimize process variability • Minimize loss of products • Respect process constraints • Safeguard environmental laws Constrained optimization problem 10

Challenges for Offshore Production Targeting APC+RTO Objectives Ø Ø Ø Ø Maximize Production (oil & gas) Ensure Quality Specification (oil, water & gas) Minimize Losses (flare & TOG) Minimize Energy Consumption (heat & power) Minimize Operational Costs (maintenance & shutdowns) Minimize Process variability Ensure Process Stability and Safety Flow Assurance (prevent: hydrate, paraffins, asphaltenes, fouling) 11

Open-Loop Optimal Control Plant Controller set-point input output r(t) u(t) y(t) measurements model constraints path constraints terminal constraints 12

(desired output) Open-loop optimal control problem: Find current and future manipulated inputs that best meet a desired future output trajectory. next sample time Model Predictive Control (MPC) Feedback nature: Implement first “control move” then correct for model mismatch. Major issue: disturbances vs. model uncertainty. * B. W. Bequette, 1998. Process Dynamics. Modeling, Analysis, and Simulation, Prentice Hall. 13

Open questions in MPC • Type of model for predictions? linear: state space, TF, step response, impulse response, ARX nonlinear: first principles, NN, Volterra, Wiener, Hammerstein, multiple model, fuzzy, NARX • Information needed at step k for predictions? outputs, state estimates, measured disturbances, model parameters • Objective function and optimization technique? quadratic (QP), absolute values (LP), economics (EMPC), nonlinear (NLP) • Correction for model error? additive output, additive input, disturbance estimation (KF, EKF, MHE) 14

Implementation of APC ü Control structure design ü Check instrumentation and retune regulatory control ü Pre-tests and design of inferences (soft sensors) ü Plant test and identification of dynamic models ü Controller configuration and closed-loop simulation ü Commissioning and tuning of the controller ü Monitoring the APC performance ü Training of operators and documentation 15

Retuning Regulatory Control Regulatory control is essential for the success of APC Gas Processing Plant * Campos, M. C. M. , Teixeira, A. F. , 2011. Advanced Control Systems, 7 th Int. Conf. on Integrated Operations, Trondheim, Norway. 16

Retuning Regulatory Control FPSO Control System (offshore oil & gas production plant) Many transients • Shutdowns • well re-alignments • Flow instabilities • compressors availability About 90% of PID control loops analyzed could have a better performance Adaptive tuning * Campos, M. C. M. , Teixeira, A. F. , 2011. Advanced Control Systems, 7 th Int. Conf. on Integrated Operations, Trondheim, Norway. 17

Monitoring the APC Performance MPC performance can degrade due to: Ø Changes in the unit operations objectives; Ø Equipment efficiency losses (e. g. , fouling); Ø Changes in the feed quality; Ø Problems in instruments and in soft sensors; Ø Lacks of qualified personnel for the controller's maintenance. 18

MPC in Offshore Platforms Industrial applications: ü Hocking & Caward (1999) – Honeywell RMPCT applied to production control 2% increase in production ü Godhavn et al. (2005) – Statoil in-house MPC controller (SEPTIC): slug control 3% increase in production Simulated: ü Plucenio & Djmgab-ah (2006) – NL-DMC applied to pressure control ü Willersrud et al. (2011) – NMPC with unreachable setpoint for production control ü Cota & Reis (2012) and Ribeiro et al. (2016) – MPC applied to production and quality control ü Miyoshi et al. (2012) and Mendes et al. (2012) – MPC applied to slug control ü Peixoto et al. (2015) – Extremum seeking for production control. 19

Challenges for MPC v Optimization problem - infinite prediction horizon - multiple objectives v Simplifying the model development process - plant testing & system identification - nonlinear model development - intensive use of dynamic simulators - model reduction techniques v State Estimation - Lack of sensors and sensor location for key variables v Reducing computational complexity - approximate solutions, preferably with some guaranteed properties - modern computation (sparse matrices, better numerical methods) v Better management of “uncertainty” - creating models with uncertainty information (e. g. , stochastic model) - on-line estimation of parameters / states - “robust” solution of optimization - self-tuning and adaptive MPC 20

APC in Offshore Platforms Conventional Platform • FPSO (Floating Production, Storage and Offloading) • Located on the Marlim Field 21

Conventional Platform Gas Lift Exportation Compression Fuel Gas Wells Oil Treatment Produced Water Treatment Purge 22

Environment for Modeling, Simulation and Optimization What can we do with EMSO? þ Steady-state simulations þ Dynamic simulations þ Steady-state optimizations (NLP, MINLP) þ Steady-state parameter estimations þ Dynamic parameter estimations þ Steady-state data reconciliations þ Process follow-up and inferences with OPC communication þ Build bifurcation diagrams (interface with AUTO for DAEs) þ Dynamic simulations with SIMULINK (interface with MATLAB) þ Add new solvers (DAE, NLA, NLP) þ Add external routines using the Plugins resource 23

Environment for Modeling, Simulation and Optimization EMSO Key Features ü Open source library of models ü Object-oriented modeling ü Built-in automatic and symbolic differentiation ü Automatic checking and conversion of units of measurement ü Solve high-index problem ü Perform consistency analysis (Do. F, initial condition) ü Integrated Graphical User Interface (GUI) ü Building blocks to create flowsheets ü Discrete (state and time) event handling ü Multitask for concurrent and real-time simulations ü Very modular architecture and support to sparse algebra ü Multiplatform: win 32 and posix ü Interface with user code written in C/C++ or Fortran ü Automatic documentation of models using hypertexts and La. Te. X 24

Conventional Platform 25

Pre-Salt Platform Exportation Injection Wells Exportation Compression Fuel Gas Main Compression Wells Oil Gas Dehydration Oil Treatment Dew-Point Adjustment Injection Compression CO 2 Removal CO 2 Compression Vapor Recovery Unit Produced Water Treatment Purge 26

Pre-Salt Platform 27

Thermodynamic Models - VRTherm High CO 2 content at high pressure and temperature • GERG 2008 • CPA with quadrupole Mixture 1 Mixture 2 Component Mixture 1 Mixture 2 Pressure (bar) 28

Production well model • Observer Design for Multiphase Flow in Vertical Pipes with Gas-Lift — Theory and Experiments, O. M. Aamo, G. O. Eikrem, H. B. Siahaan, B. A. Foss, Journal of Process Control 15 (2005) 247– 257. • Simplified model that aims to capture the casing heading phenomenon • Constant reservoir pressure • Two-phase flow in the pipeline, oil and water treating as a single phase 29

Riser model • A low-dimensional dynamic model of severe slugging for control design and analysis, Espen Storkaas, Sigurd Skogestad, John-Morten Godhavn, Multiphase'03 2003. • Simplified model that aims to reproduce severe slugs, capturing the main pressure dynamics in the pipeline 30

Three-phase separator model • Black Oil model • Mass balances: – Oil chamber – Separation chamber – Gas space 31

Compression cycle model • • • 1 turbine 1 turbo compressor per stage 1 heat exchanger per stage 1 flash drum per stage 1 heat exchanger at the end of the cycle 32

CO 2 removal model • Gas-separation membrane modules • Cocurrent operation • Permeation flowrate of each compound as function of the log-mean square of its partial pressure differences and permeability. Feed Retentate Permeate 33

Regulatory control • • • Oil level control Oil-water interface level control Flash drums level control Compressors capacity control Anti-surge control Flowrate control Pressure control Temperature control Anti-slug control 34

Anti-slug control • Topside choke valve: Havre & Dalsmo (2001), Eikrem et al. (2004), Godhavn et al. (2005), Sinegre et al. (2005), Storkaas & Skogestad (2008), Scibilia et al. (2008), Di. Meglio et al. (2012), Jahanshahi et al. (2012), Stasiak et al. (2012), Bendia (2013), Oliveira et al. (2015), Campos et al. (2015) • Gas-lift control: Asheim (1988), Hu (2004), Plucenio et al. (2012), Krima et al. (2012) • MIMO control: Pagano et al. (2008), Miyoshi et al. (2012), Abardeh (2013), Jahanshahi et al. (2013) 35

Anti-slug control Wellhead pressure (P 1) • Variable gain and unstable region Stable s. s. Min. pressure Max. pressure Unstable s. s. Average pressure Hopf bifurcation point Topside choke opening (u) 36

Anti-slug control separator topside choke well • PI controller with adaptive tuning (Bendia, 2013) 37

Anti-slug control 38

Wellhead pressure (bar) Anti-slug controller on Topside choke opening Time (s) 39

Anti-slug control Proportional gain PI proportional gain Time (s) 40

MPC in Offshore Platform • MPC for production and quality control (Ribeiro et al. , 2016) Anti-slug regulatory control 41

MPC in Offshore Platform Goals – Setpoint tracking of oil production flowrate – TOG ≤ TOGmax = 1000 ppm (Total of Oil and Grease) – BSW ≤ BSWmax = 20% (Basic Sediments and Water) 42

MPC in Offshore Platform 4 3 model identification EMSO-MATLAB/Simulink interface: 43

MPC in Offshore Platform Model fit = 95. 4% Identification with overlaped inputs signals sampling points (sampling time = 10 s) Planta (EMSO) Modelo Output 1: oil flowrate 44

MPC in Offshore Platform Produced oil flowrate setpoint change 45

MPC in Offshore Platform § Increase of water content due to reservoir aging, represented by a linear increase of BSW in well #91 from 30% to 40% during 4 hours 46

MPC in Offshore Platform Disturbance rejection (BSW increase in well #91) 47

Nonlinearity 48

NMPC in Offshore Platform Controlled Variables Unit Description HP liquid level (Simões, 2017) Manipulated Variables LP liquid level Unit Description Control valve of HP liquid level TO interface level Control valve of LP liquid level HP pressure Control valve of TO int. level LP pressure Control valve of HP pressure Control valve of LP pressure Compression system booster Production header Exportation oil Produced water treatment 49 49

NMPC in Offshore Platform Disturbance Description Unit HP inlet liquid flowrate LV 01 outlet pressure HP inlet gas flowrate PV 01 outlet pressure LP inlet liquid flowrate LV 02 outlet pressure LP inlet gas flowrate PV 02 outlet pressure LV 03 outlet pressure Pump outlet pressure HP temperature LP temperature Compression system booster Production header Exportation oil Produced water treatment 50 50

NMPC in Offshore Platform § Oil Treatment Plant NMPC model language HYSYS Simulation BRNMPC OPC 51 51

NMPC in Offshore Platform § Response to slug flow disturbance TO level control 0. 95 LP pressure control 55 0. 9 54 h 3 [m] SP m 0. 8 PID - PV m 0. 75 MPC - PV m 0. 7 Pressão [k. Pag] 53 0. 85 NMPC - PV m 0. 65 52 51 SP m 50 PID - PV m 49 MPC - PV m 48 NMPC - PV m 47 46 0. 6 0 5000 10000 15000 20000 0 25000 51 15000 20000 25000 54 49 52 45 43 PID - OP % 41 MPC - OP % NMPC - OP % 39 PV 02 [%] 47 LV 03 [%] 10000 Tempo [s] 50 48 PID - OP % 46 MPC - OP % 44 NMPC - OP % 42 37 40 35 0 5000 10000 15000 Tempo [s] 20000 25000 52

NMPC in Offshore Platform § Step response to setpoint change in the TO interface level 1. 2 47. 2 42. 2 1. 1 37. 2 1 h 3 [m] 0. 9 Set-point [m] NMPC - PV [m] 0. 8 LV 03 [%] 32. 2 27. 2 NMPC - OP [%] 22. 2 PID - OP [%] PID - PV [m] 17. 2 0. 7 12. 2 0. 6 7. 2 0. 5 0 5000 10000 15000 Tempo [s] 20000 25000 30000 2. 2 0 5000 10000 15000 20000 25000 30000 Tempo [s] 53

NMPC in Offshore Platform § NMPC presented better performance; § Level control: § MPC and NMPC: effects of the upstream vessels were taken into account; § MPC and NMPC: presented similar results; § Pressure control: § Nonlinear behavior; § NMPC: superior performance than MPC and PID, with anticipatory actions; § 5 PID controllers were replaced by one (N)MPC (safety with watchdog). 54

MPC in Compression System recycle 1 st relief valve u 12 Feed d 01 -d 02 (Thomaz, 2017) u 14 1 st train y 01 -y 05 Heat exchanger u 01 Heat exchanger u 03 1 st header u 09 2 nd train y 06 -y 10 Dehydration Heat exchanger u 02 u 15 2 nd relief valve recycle u 13 2 nd header u 10 y 21 Heat exchanger u 05 1 st train Heat exchanger u 07 3 rd header u 11 2 nd train y 16 -y 20 u 17 recycle Heat exchanger u 04 recycle u 18 recycle u 16 1 st train y 11 -y 15 Sweetening Heat exchanger u 06 2 nd train u 19 recycle Exportation d 03 Heat exchanger u 08 output 55

MPC in Compression System nz = 305 transfer functions Ø Not considering functions with small static gains (< 10 -3): result in 305 transfer functions. Controlled variables Original system: 22 21 = 462 Manipulated variables 56

Interface EMSO-MATLAB® Ø EMSO: virtual plant with regulatory control Ø Simulink®: MPC measured disturbance unmeasured disturbance d reference trajectory MPC Toolbox u EMSO (regulatory control and process model) y output feedback 57

MPC in Compression System Main Compressor Ø Step response to +5% disturbance in feed flowrate Less work with MPC Less energy with MPC 58

MPC in Compression System Export Compressor Ø Step response to +5% disturbance in feed flowrate Fast response and constraint satisfaction Less work with MPC Less energy with MPC 59

MPC in Compression System CO 2 removal Ø Step response to +5% disturbance in feed flowrate Less impact in separator efficiency with MPC 60

MPC in Compression System Main compressor Ø Step response to 3% disturbance in 2 nd train compressor efficiency 1 st train 2 nd train 61

MPC in Compression System Ø Significant economic gains were obtained with MPC, constraints were respected increasing the uptime of the plant and less maintenance of the equipment; 62

Final Remarks • Offshore plants are becoming more complex, requiring APC and RTO strategies • MPC and RTO are mature industrial technologies • Robustness and remote operation are still very demanding for offshore APC/RTO • Monitoring and diagnosis are open issues for feedback information • First principles models are even more needed • Formation and training of engineers in these advanced tools are crucial Lots of work for Process System Engineers!! 63

Team Professors: Adilson E. Xavier Argimiro R. Secchi Bruno Capron Frederico W. Tavares Maurício B. de Souza Jr. Paulo L. C. Lage Príamo A. Melo Jr. Pós-Docs: José Mauel G. T. Perez Leonardo S. Souza Rodrigo G. D. Teixeira Simone C. Miyoshi Engineers: Carlos R. Paiva Jéssi C. Heck Rafael M. Bendia Secretariat: Rosemary Cezar Ph. D students: Ana K. Muniz Alex F. Teixeira Ataíde S. Andrade Caio F. C. Marcellos Daniel M. Thomaz Eliza H. C. Ito Felipe C. Cunha Guilherme A. A. Gonçalves Gustavo V. K. Campos Jacques Niederberger Jeiveison G. S. S. Maia Leonardo D. Ribeiro Rafael B. Demuner Rafael R. L. Britto Reinaldo C. Spelano Roymel R. Carpio Sergio A. C. Giraldo Thamires A. L. Guedes MSc students: André F. F. Souza Joaquin Lujan Mario G. Neves Nt. Maria Rosa R. T. Goes Otávio F. Ivo Pedro C. N. Ferreira Saul Simões Nt. Thiago C. Dávila Undergrad. students: Carlos M. M. Fonseca Isabella Q. Souza Lucas Marques Luis A. V. Carapeto Pedro Delou Rodrigo Moyses Silvio Cisneiros Nt. Thales S. M. Gama Victor C. Gomes Vinícius C. C. Plácido 64

. . . thank you for your attention! http: //www. enq. ufrgs. br/alsoc Solutions for Process Control and Optimization Process Modeling, Simulation and Control Lab • Prof. Argimiro Resende Secchi, D. Sc. • Phone: +55 -21 -3938 -8307 • E-mail: arge@peq. coppe. ufrj. br • http: //www. peq. coppe. ufrj. br/Areas/Modelagem_e_simulacao. html 65