Analysis of Natural Fractures in the Basal Zechstein Carbonates in the Dutch Offshore Area using Wireline Log Data

Analysis of Natural Fractures in the Basal Zechstein Carbonates in the Dutch Offshore Area using Wireline Log Data

Kooijman, D.A.

Luthi, S.M. (mentor)
Giovanni, B. (mentor)

Civil Engineering and Geosciences

Section Petroleum Engineering/ Applied Geology/ Applied gophysics and petrophysics/ Geo-Engineering/ Resources Engineering

2011-11-21

The primary gas reservoirs in the Southern Dutch offshore are in decline. Discoveries of new gas reservoirs are getting smaller and smaller reservoirs receive more attention. The three major carbonate beds in the Basal Zechstein (Z1C, Z2C and Z3C) are considered potential secondary gas reservoirs. To be able to develop the gas from these carbonates depends strongly on the presence of fractures. The fracture pattern of the Zechstein carbonates in the Dutch offshore is however not well known. Currently the most reliable methods to investigate fractures in the boreholes are borehole-wall imaging techniques. These imaging techniques are expensive and time-consuming and most of the time not included in the logging suite. Therefore, an alternative method of predicting fractures using basic wireline logs (Gamma ray, sonic, density, photoelectrics, resistivity and caliper) would be very valuable. In this study an artificial neural network (ANN) is developed in order to make fracture predictions based on wireline log data. Additional the ability of indicating fractures with a decision tree is explored for wells that have too few logs available for a ANN approach. Both the ANN and decision tree are trained with supervised learning. The models are calibrated using core data (rotary sidewall core and slabbed core data) of fractured and non-fractured Zechstein carbonate intervals. The ANN is trained with success, where the degree of success is directly related to the amount of fractures identified in the core. The DCT approach resulted in a moderate fracture prediction performance and is best used in combination with an ANN. With this combination a fracture prediction has been made on several wells that were excluded from training with these two methods. The resulting fracture predictions have been compared with indications of mud losses and productivity data, showing a good match.

fractures

http://resolver.tudelft.nl/uuid:f916c94d-fd73-43bb-bc71-4d7a199ab23b

Student theses

master thesis

(c) 2011 Kooijman, D.A.