Fuzzy inference analysis of petrophysical data

Abstract

Petrophysical data analysis is a major task in the oil industry and it can be time consuming, tedious and very expensive. Over the years, geoscientists have worked tirelessly to minimize the cost of interpreting these data. Several methods such as the geophysical inversion techniques, statistical methods, interval inversion method, and global optimization method have been used to interpret these data. This paper investigates the possibilities of using rule-based Fuzzy Inference method to analyze petrophysical data. In this study, the well logs data used was provided by Shell Petroleum Development Company, Nigeria. The exploration well logs data is clustered into sixteen groups using unsupervised neural network. The rule-based containing six fuzzy lithology rules is developed from the training data sets, and the rule strength is weighted. A Mamdani inference system and the centroid of area defuzzification method are used for fuzzy inference. Six fuzzy rules was proposed based on neural network output. The lithology was determined as Non- reservoir, oil, gas, water and low reservoir quality (LRQ) respectively. It was observed that fuzzy inference systems used provides a fast and comprehensive detail of the lithology and fluid content of the subsurface structure of the petrophysical data interpreted. The results obtained from fuzzy inference analysis of the data tallies with the descriptive lithology of the data obtained from Shell Petroleum Development Company.