Research On Extracting Useful Rules And Perceptual Reasoning For Intelligent Evaluation Of Oilfield Production

Currently, in view of “massive oilfield data and a little useful information”, it has been a hot issue widely concerned how to dig useful information and implement intelligent evaluation of oilfield production combined with intelligent control method. In addition, it has great research significance and application value. Extracting useful rules and perceptual reasoning are mainly investigated for intelligent evaluation of oil-field exploitation in the paper.At first, a scheme of oilfield production intelligent evaluation is proposed. Combined the data of oilfield production with the expert experience of reservoir engineers, oilfield production intelligent evaluation system is built according to computing with words based on interval type-2 fuzzy sets. The engine part of computing with words includes linguistic summarization used to extract useful rules and perceptual reasoning used to compute the output of computing with words engine.Then, linguistic summarization method based on interval type-2 fuzzy sets is used to generate IF-THEN rules from oilfield recovery database. Four quality measures, that is, truth, coverage, reliability and outlier to measure different character of each rule. Parallel coordinate method in data visualization technique is used to establish the visually fuzzy logical rule base in GUI interface of MATLAB. Quality measures of each rule are shown, logical curve graphs is drawn between antecedents and consequents of IF-THEN rules and the membership function graphs are plotted corresponding to antecedents and consequents. These logical rules construct initial rule base for perceptual reasoning through specialistic validation. The useful rules extraction method not only saves expert’s time but also helps improve rules considering the inconsistency between expert experience and field data. Hence, it provides rule basis for intelligent evaluation of oilfield production.Finally, perceptual reasoning based on firing intervals and perceptual reasoning based on similarity are researched according to the problem of intensely distorted output footprint of uncertainty of interval type-2 fuzzy logic system. Firing interval or firing level are computed and the outputs of firing rules are obtained using linguistic weight average method. Therefore, the map between input interval type-2 fuzzy sets and output interval type-2 fuzzy sets is established. Then, comparative analysis is carried for two different perceptual reasoning methods based on useful rule bases. The results of simulation show that results obtained from perceptual reasoning based on similarity are more similar with footprint of uncertainty of words in codebook, and more in accordance with oilfield production reality.