Determining In-Situ Stress Profiles From Logs . Shahab D. Mohaghegh - PowerPoint PPT Presentation

SPE 90070 Determining In-Situ Stress Profiles From Logs . Shahab D. Mohaghegh West Virginia University Andrei Popa , Chevron-Texaco – Razi Gaskari , WVU – Steve W olhart , Pinnacle Technologies – Robert Siegfreid , GTI & Sam Am eri , WVU 1 SPE Annual Technical Conference, Houston, TEXAS, September 2004

SPE 90070 OUTLINE � Introduction � Methodology � Results & Discussion � Conclusions 2 Shahab D. Mohaghegh

SPE 90070 Introduction � All hydraulic fracturing simulators require information regarding in-situ stress. � An important factor in designing successful hydraulic fractures. 3 Shahab D. Mohaghegh

SPE 90070 Introduction � The contrast between different formation’s in-situ stress is the controlling factor for the shape of the hydraulic fracture. 4 Shahab D. Mohaghegh

SPE 90070 Introduction 5 Shahab D. Mohaghegh

SPE 90070 Introduction � Lack of accurate in-situ stress values during the design of a hydraulic fracture can result in as much as 50% error in the actual fracture length upon implementation. 6 Shahab D. Mohaghegh

SPE 90070 Introduction � Currently there are two methods of in-situ stress determination: � Direct measurements (Expensive) � Calculate from logs (inaccurate) 7 Shahab D. Mohaghegh

SPE 90070 Introduction � The inaccuracy of calculating in-situ stress from logs has its roots in lack of incorporating tectonic stress components in the calculation � Not an easy and inexpensive way to measure it � Not sure how to incorporate it, even if available. 8 Shahab D. Mohaghegh

SPE 90070 Introduction � This paper introduces a new method for incorporating tectonic stress when modeling in-situ stress from logs. 9 Shahab D. Mohaghegh

SPE 90070 OUTLINE � Introduction � Methodology � Results & Discussion � Conclusions 10 Shahab D. Mohaghegh

SPE 90070 Methodology � It was identified that formation lithology is an important input for the model. � This was also shown in previous studies such as “ABC” method. 11 Shahab D. Mohaghegh

SPE 90070 Methodology � To deduce formation lithology from logs a fuzzy system was developed. � The fuzzy system uses three logs and 27 rules in order to decide on the formation lithology. � Gamma Ray � Deep Induction � Spontaneous Potential 12 Shahab D. Mohaghegh

SPE 90070 Methodology Fuzzy system rules for determining formation lithology. 13 Shahab D. Mohaghegh

SPE 90070 Methodology � In order to incorporate tectonic stress component in our analysis two parameters were identified as surrogate: � Spectral Response Acceleration � Seismic Zone Factor 14 Shahab D. Mohaghegh

SPE 90070 Methodology Seismic Zone Map of the United States of America 15 Uniform Building Code – 1991Edition Shahab D. Mohaghegh

SPE 90070 Methodology Spectral Response Acceleration 16 Shahab D. Mohaghegh

SPE 90070 Methodology � These parameters are borrowed from structural engineering where tectonic activities play an important role in design. � They have been used as an indicator of tectonic activities in our study. 17 Shahab D. Mohaghegh

SPE 90070 Methodology � Neural Networks were used to develop an in-situ stress model. � Model inputs are: � Logs � Formation Lithology � Tectonic Stress Component 18 Shahab D. Mohaghegh

SPE 90070 Methodology � Inputs to the Neural Network Model. 19 Shahab D. Mohaghegh

SPE 90070 OUTLINE � Introduction � Methodology � Results & Discussion � Conclusions 20 Shahab D. Mohaghegh

SPE 90070 Results � Existing models usually perform reasonably well when applied to the wells that were used for these model’s development. � The real test would be their performance on wells that were not used during the model development. 21 Shahab D. Mohaghegh

SPE 90070 Results � Four existing models are tested: � Staged Field Experiment (SFE) # 1 � Staged Field Experiment (SFE) # 2 � Staged Field Experiment (SFE) # 3 � Staged Field Experiment (SFE) # 4 � These models were developed for GRI in the past decades. 22 Shahab D. Mohaghegh

SPE 90070 Results All models applied to SFE # 2 measurements. 23 Shahab D. Mohaghegh

SPE 90070 Results All models applied to measurements from Ashland SFOT and AT Mast wells. 24 Shahab D. Mohaghegh

SPE 90070 Results All models applied to measurements from Hogsback and Anderson Canyon wells. 25 Shahab D. Mohaghegh

SPE 90070 Results Neural model applied to measurements from a well that was not used during its development. 26 Shahab D. Mohaghegh

SPE 90070 Results Neural model applied to measurements from a well that was not used during its development. 27 Shahab D. Mohaghegh

SPE 90070 OUTLINE � Introduction � Methodology � Results & Discussion � Conclusions 28 Shahab D. Mohaghegh

SPE 90070 Conclusions � A new model for determining in-situ stress from logs has been developed and presented. � The new model incorporates tectonic stress component for in-situ stress determination 29 Shahab D. Mohaghegh

SPE 90070 Conclusions � The new model is demonstrated to be superior to the existing techniques. � Neural networks and fuzzy logic were the main techniques used to develop the new model. 30 Shahab D. Mohaghegh