RISK ASSESSMENT FOR CO2 SEQUESTRATION

Transcription

1 RISK ASSESSMENT FOR CO2 SEQUESTRATION Yan Zhang Department of Chemical Engineering Carnegie Mellon University Panagiotis Vouzis Department of Chemical Engineering Carnegie Mellon University Nick Sahinidis National Energy Technology Laboratory Department of Chemical Engineering Carnegie Mellon University

2 INTRODUCTION TO CO2 SEQUESTRATION Effect of anthropogenic CO2 on global warming Carbon mitigation portfolio Improved efficiency vehicles Decarbonization renewable, nuclear Sequestration CO2 sequestration options Ocean large capacity, ecological impact Mineral permanently, energy penalty Geological: mature technology oil/ gas reservoir deep saline formation deep coal seam

3 RISKS IN CO 2 GEOLOGIC STORAGE Risks Massive release of CO 2 to the atmosphere Contamination of local environment: soil, water Risk = Probability * Severity (impact) Identification of events/scenarios Estimation of likelihood/probability Evaluation of impact/ consequences (IPCC, 2005) A. CO 2 gas pressure exceeds capillary pressure B&D. CO 2 leaks through faults C. CO 2 escapes through gap in caprock E. CO 2 leaks through broken wells F&G. CO 2 leaks through open aquifer

4 Approx. Depth (m) 300 PHYSICAL SYSTEM Radius ~0.1m Biosphere Aquifer 6 Aquitard 6 Aquifer 5 Aquitard 5 Upper Formations Aquifer 4 Aquitard 4 >800 Aquifer 3 Aquitard 3 Aquifer 2 Aquitard 2 Aquifer 1 Annulus cement Steel casing Oil Reservoir Caprock Low permeability Lower Formations CO 2 Source Pool Cement plug Simplified geological structure

5 PROCESS DESCRIPTION Viability of a sequestration system: Failed seals of wellbore Wellbore Caprock Leakage vs. Sequestration Celia et al., 2004 Open fractures and faults Dissolution of source pool Migration of CO2 into surrounding formations Mineralization dm dt = F i F o M: CO2 mass F i : inflow rate = 0 F o : outflow rate = leakage + sequestration

6 MOVEMENT THROUGH WELLBORE Non-penetrating well equation Wellbore Caprock Fluid pressure CO2 movement Buoyancy CO 2 source pool

7 MOVEMENT THROUGH CAPROCK Equation for drainage in tunnels Wellbore Caprock Fluid pressure CO 2 source pool Buoyancy

8 DISSOLUTION OF SOURCE POOL Dissolution of source pool to the formations below or above the source A layer of moving formation Diffusion through fluids a falling film A layer of stagnant formation fluids Diffusion through a semi-infinite plane CO 2 source pool CO 2 source pool Steady state dissolution rate = Diffusion through a semi-infinite plane + Diffusion through a falling film

9 MIGRATION THROUGH FAILED SEALS Migration of CO2 to surrounding formations through failed seals on the way of rising up Failure times are unpredictable If the cement annulus fails first If the cement plug fails first Physical model Heat conduction from a thin wire Aquifer with advection, dispersion and reaction Aquitard with diffusion and reaction Apply solution for temperature profile Get CO2 concentration in the formation Obtain steady state flux at the wall of the wellbore Migration rate = steady state flux * concentration at the wall * thickness

10 OVERALL MASS BALANCE The mass left in the reservoir pool as time goes by: Leakage through wellbores and/or faults Sequestration Dissolution + Mineralization Depends on the failure times of wellbore components

11 RESULTS FOR ONE SET OF PARAMETERS Loss rate due to migration Left in the reservoir Released to the biosphere Sequestered in the formations < 5% 500 years 400 years Loss rate due to migration Leakage rate through the wellbore Leakage rate to the surface CO2 mass fractions vs. time CO2 leakage rate (kg/a) vs. time

12 MODEL PARAMETERS Parameters: Main independent parameters Formation Brine water CO2 phase Wellbore Depth Thickness Permeability Porosity Tortuosity Mineralization rate Salinity Pressure Surface tension of the free phase Temperature gradient Darcy velocity Capillary pressure in reservoir Dispersion coefficients Mass Composition Thermo-property Permeability Effective saturation Cement seal failure times Casing failure times Permeability of degraded cement Effective crack diameter of degraded cement Surface tension Wellbore radius Main calculated parameters: formation fluid density, free phase density, viscosity, solubility, etc.

13 PROBABILISTIC RISK ASSESSMENT Input: Randomly Sampled from probability distribution function of independent parameters Monte Carlo simulation Output: Statistical Analysis Simulator sample # Generated inputs with independent parameters x 1 x 2 x n Permeability Porosity Model Outputs y 1 y 2 y m Mass left Mass leaked

14 MINIMUM MC SAMPLING SIZE 95% confidence interval for the average of the output: s X ±1. 96 n Algorithm: 1.96 n = B B 2 2 s 2 Start Current population n n > n? Yes End Add one sample No Effect of number of input parameters on minimum sampling size

15 RESULTS FOR MONTE CARLO SIMULATION Frequency histogram showing fraction released to the biosphere at 5000 years with normal (left) and lognormal (right) inputs

16 RESULTS FOR MONTE CARLO SIMULATION Frequency histogram showing various times when 1% of total injected CO 2 mass leakage occurs with normal (left) and lognormal (right) inputs

17 SAMPLING SIZE VS NUMBER OF PARAMETERS Number of uncertain parameters

18 SENSITIVITY ANALYSIS Spearman s rank correlation between the mass fraction of leakage and normally distributed independent input parameters

19 SENSITIVITY ANALYSIS Frequency histogram showing fractions released to the biosphere in 5000 years with different variances of the distribution of relative permeability

20 MUTIPLE LEAKY WELLS SIMULATION Frequency histogram showing mass fraction released to the biosphere in 5000 years with ten leaky wells under snormal (left) and lognormal (right) inputs

21 GPU PARALLEL COMPUTING Multi-core and many-core CPU CORE CORE CORE CORE Dynamic Random Access Memory GPU Lecture Notes, David Kirk/NVIDIA and Wen-mei Hwu,

22 FASTER THAN CONVENTIONAL CPU IMPLEMENTATION

23 CONCLUSIONS Developed a CO2 sequestration simulator Based mostly on the ideas of the CQUESTRA model Performed deterministic simulation and probabilistic risk assessment Over 90% of the injected CO2 is safely trapped underground after 5000 years in our simulations Parallel implementation of Monte Carlo simulation on GPUs 64 times speedups

24 Forma'on Deep saline aquifer Depleted oil/gas reservoir Time scale FUTURE WORK: CO 2 STORAGE IN DEEP SALINE AQUIFER Injection Mobile Post-injection Immobile Time (years) Trapping mechanisms Structural trapping Residual trapping Solubility trapping Mineral trapping (IPCC, 2005) (Kaldi, 2010) (Juanes et al., 2008)

25 FUTURE WORK: CO 2 STORAGE IN DEEP SALINE AQUIFER Forma'on Deep saline aquifer Depleted oil/gas reservoir Space Leaking scenarios A layer of target forma'on Con'nuous Leak when encountering broken wells Leak when encountering faults Individual reservoir(s) Discrete Leak through intersec'ng wells Leak through faults/fractures Injection well A broken well Injection well A broken well A fault A fault Immobilized CO 2 Static CO 2 source pool Brine water Mobile CO 2

26 Thanks! Questions?