Application of knowledge-driven data mining in the complex reservoir evaluation

Several ten techniques were introduced and further developed for the complex oil and gas reservoir evaluation. However, these techniques are found with both consistency and differences/contradictions. Given this, knowledge-driven data mining, combined with the identified data response characteristics of multiple complex reservoirs in the study area, was used to search for sensitive parameters from mud logging and well logging data of target layers of formation testing. Subsequently, factor analysis, multivariate discriminant analysis, and grey correlation analysis were performed to build the multi-parameter reservoir fluid property evaluation model. The application results showed that the presented multi-parameter evaluation model is objective and effective. It delivers effective evaluations of the bio-degraded oil reservoir, heavy oil reservoir, and mixed-source oil reservoir in the study area, and provides inspiring experience for the fine evaluation of complex reservoirs. It can be further promoted to relevant oilfields and makes contributions to improving the efficiency of oilfield exploration and development.