Data-Driven Analysis of Stimulation Treatments Using Association Rule Mining

Abstract

Effective analysis of data collected during the well’s operational cycle is crucial to well performance, economic evaluation, and good decision-making in the upstream oil industry. Generally, the analysis of huge volumes of data stored in databases is beyond the power of traditional methods, such as curve-fitting and statistical hypothesis testing. Data mining is the practice of analyzing large databases to identify patterns, anomalies, and correlations, within the data, leading to new, hidden, and valuable knowledge that would support decisions. This article proposes a data-driven methodology for analyzing the stimulation operations data in oil/gas wells to identify the underlying rules or patterns that lead to successful operations. Association rule mining (ARM) is used in this research for rule induction purposes. The proposed approach aims to mine the frequently occurring rules, within the collected database, that guarantee the success of stimulation operations with a high degree of confidence. Finally, the proposed approach is evaluated against a set of real data from an Iranian oil field. On the basis of past stimulation operations, these extracted rules show the conditions that are most likely to lead to a successful operation. The rules identified by the proposed approach are compared against the rules that can be generated by the decision tree (DT) technique using the same data set. As the reliability of the rules is controlled by setting the minimum thresholds on support and confidence, more significant and useful rules could be derived from ARM compared to the DT technique. Using the identified rules and generated information can support the operational decisions by assisting in the design of due stimulation jobs or in selecting the appropriate candidates for future operations.