一种基于差分演化算法的规则提取及其在石油储层识别中的应用

Differential Evolution for Rule Extraction and Its Application in Recognizing Oil Reservoir

提出了一种基于差分演化算法的规则提取方法(DE-Rule)。规则的表达形式为IF-THEN模式,规则前件中的连接词为AND,规则后件为识别对象的类别。通过编码将规则表示为差分演化算法中的种群个体,然后通过差分变异和二项式交叉对种群进行演化,对最优个体进行解码即可得到对应的最优规则集。最后,用DE-Rule、RS-Rule、ANN-GA-Cascades-Rule等3种规则提取方法对江汉油田某区块进行石油储层识别(干层、水层、差油层和油层),其中,oilsk81油井数据作为训练数据,oilsk83油井数据作为测试数据,结果表明,在综合考虑规则集的识别准确率和解释性上,DE-Rule方法都优于其他两种方法。

We propose a method for rule extraction based on differential evolution （DE-Rule）, with the rule form as IF-THEN, where the antecedent of rule is AND and the consequence of rule is a class label. The rules are encoded and expressed for individuals in a population of differential evolution. The population evolution is through differential mutation and binomial crossover and the optimal rule set is obtained after decoding the optimal individual. DE-Rule, RS-Rule and ANN-GA-Cascades-Rule are used to recognize oil reservoir （dry layer, water layer, inferiority layer and oil layer） in the Jiangban Oil Field. Data of the well ＂oilsk81＂ is used as the training data and data of well ＂oilsk83＂ as testing data. The computational results show that in terms of the accuracy and the interpretability, DE Rule is the best.