RESERVOIR CHARACTERISATION USING WIRELINE/LWD & DRILLING AND HYDROCARBONGAS DATA & ADVANCED COMPUTATIONAL MATHEMATICS
Andrew Hurst1,2, Ravi Arkalgud1, Gerold Tischler1 & Steve Cuddy2 1Northlight
Geoscience and 2University of Aberdeen
CONTENT 1. Background 2. Goals of technology development 3. Work flow 4. CASE STUDY: prediction of porosity using wireline data 5. CASE STUDY: prediction of porosity & saturation in an HPHT well using HC-gases and drilling data 5. Conclusions
BACKGROUND: formation evaluation
•
Definition of the non-dynamic characteristics of reservoirs is strongly dependent on borehole data (wireline, LWD, core)
•
Borehole data are in the main accurate and precise measurements that are possible to make (natural gamma-radiation, electron density etc) but do not represent what we actually need to know (porosity, bulk-volume hydrocarbons etc)
•
Formation evaluation utilises empirical relationships to define the required characteristics from borehole data
•
In general formation evaluation uses statistical methods (MLR, FL, GA, NN) that employ distribution functions of various inputs and outputs and implicity assume that what is known (borehole data) allows one to predict what is unknown (formation characteristics)
•
HOWEVER - statistical methods can only “predict” what is already known
SCIENTIFIC BACKGROUND
• NGeo’s premise is that hydrocarbon-gases measured from the mudstream and drilling parameters recorded at well sites are closely related to the physical properties of formations and their connate fluids thus provide a valuable basis for the prediction of formation character • application of an appropriate mathematical approach allows HG-gas and drilling data to be used to predict formation characteristics thus complementing conventional borehole measurements – this is an opportunity to optimise value in subsurface interpretation by cutting costs and creating value
BACKGROUND: alternative approach
• Mathematical methods exist that do not rely on statistical methods and are relevant to the predictive problems of interest • Focus on making the best possible prediction of reservoir characteristics rather than multiple realisations of individual characteristics • Hydrocarbon-gas and drilling data are physically related to the required reservoir characteristics but are used sparingly (DXC, wetness ratios etc) • Fast and precise method for reservoir characterisation that integrates all available data and extracts all possible value from input data
GOALS OF TECHNOLOGY DEVELOPMENT
•
Reduce rig time and cut costs
•
Data insurance – tool failure
•
Data assurance - complementary, independent evaluation – difficult borehole environments, difficult reservoirs
•
Optimisation of value – extraction of value from all available data
•
Real-time evaluation and decision making – formation evaluation at a very early stage of the drilling process
WORK FLOW
• DECIDE WHICH PREDICTIONS ARE REQUIRED • TRAIN ALGORITHM ON RELEVANT ANALOUE DATA TO OPTIMISE PERFORMANCE • QC INPUT DATA (e.g. wireline, LWD, hydrocarbon gases, drilling data, drilling reports etc.) • RUN ALGORITHM IN WELLS/FORMATIONS WHERE PREDICTIONS ARE REQUIRED
CASE STUDY 1: porosity prediction from wireline logs without empirical formulae (LOGDIRECT)
• evaluation of a siliciclastic reservoir in which four reservoir intervals are present • available wireline logs include, GR, RHOZ, NPHI, DT-p, DT-s and 4 resistivity logs • direct predictions are made using all data, all data minus DT-s, and without using acoustic data • predictions are compared with an operator’s evaluation of porosity
CASE STUDY 1: porosity prediction from wieline logs without empirical formulae - ~500 ft section 0.
GR (GAPI) RHOZ (G/CC) DTCO (US/F) AT10 (OHMM) 150. 1.95 2.95 140. 40. 0.2 200. 0.5 NPHI (ft3/ft3) DTSM (US/F) AT20 (OHMM) 0.45 -0.15 240. 40. 0.2 200. 0.5 AT30 (OHMM) 0.2 200. AT60 (OHMM)
PHIEDN (FRAC)
0. 0.5
PHIE_LD1 (dmsl)
PHIEDN (FRAC)
0. 0.5
PHIE_LD2 (dmsl) 0. 0.5
PHIEDN (FRAC)
0. 0.5
PHIE_LD3 (dmsl) 0. 0.5
0. 0.5
PHIE_LD1 (dmsl)
0.
PHIE_LD2 (dmsl) 0. PHIE_LD3 (dmsl)
0.5
0.
CASE STUDY 1: comparisons between predictions of porosity (LOGDIRECT)
20
UNIT 1
15
10 5
0 0.
0.3 POROSITY CURVES BASE CASE (OPERATOR) LOGDIRECT – ALL DATA LOGDIRECT – MINUS DTs LOGDIRECT – NO ACOUSTICS
CASE STUDY 1: comparisons between predictions of porosity (LOGDIRECT)
35 30 25 20 15 10 UNIT 2
5 0 0.
0.3 POROSITY CURVES BASE CASE (OPERATOR) LOGDIRECT – ALL DATA LOGDIRECT – MINUS DTs LOGDIRECT – NO ACOUSTICS
CASE STUDY 2: porosity and saturation prediction from HC-gas and drilling data – HPHT well (LOGPREDICT)
• evaluation of a HPHT siliciclastic reservoir in which the reliability of LWD data is questionable • input data include 7 HC-gas measurements, total HC gas, ROP, WoB, Tmud, RPM, TORQUE, Mud density, drilling reports • predictions are made using all available QC’d data • predictions are subsequently compared with the operator’s conventional evaluation
CASE 2 – input data – HPHT well
THE TRAINED ALGORITHM IS TASKED WITH PREDICITON POROSITY AND SATURATION IN A HPHT RESERVOIR INPUT DATA INCLUDES HC-GASES (7 + TOTAL GAS) & DRILLING DATA (ROP,WOB, TORQUE, RPM, T MUD, DEN MUD) DATA QC - INCLUDES REFERENCE TO DAILY DRILLING REPORTS
LOGPREDICT®
CASE 2 – blind predictions input data & output
LOGPREDICT PAY CURVES
LOGPREDICT®
CLIENT CPI FOR VALIDATION
CONCLUSIONS
•
CASE 1 Porosity prediction made without using empirical formulae produces similar results to operator interpretation – additional information about formation characteristics is derived from iterative tests
•
An alternative method for formation evaluation that is fast and robust
•
CASE 2 Pay (Ø & Sw) definition and delineation in the HPHT well compares favourably with the client’s independent evaluation
•
HC-gas and drilling data are amenable to the prediction of reservoir characteristics – they complement or may replace conventional data
CONCLUSIONS
•
When using data available at well sites predictions of reservoir character can be made in “real time”
•
Where LWD tools fail or are unavailable the method described are appropriate to repair/simulate in the missing sections – using HC-gas and drilling data this can be done in “real time”
•
Robust (and reasonable) relationships are inferred between drilling and HC-gas data formation characteristics that open opportunities for applications of these data in other aspects of formation evaluation (e.g. pore pressure, rock mechanics)