基于PCA-ANFIS的油井注水有效周期预测研究

Prediction of water injection effective cycle for oil wells based on PCA-ANFIS

针对油田上注水有效周期不确定的问题,提出一个基于混合主成分分析与ANFIS的数值方法进行油井注水有效周期的预测研究.首先使用主成分分析对原始数据进行降维处理,然后应用ANFIS对降维后的数据进行训练与测试.实验使用油田上116口油井的真实注水统计数据检验混合主成分分析与ANFIS模型的正确性,测试的注水有效周期平均绝对误差为1.80个月,而未经过主成分分析处理的测试平均误差为4.33个月,混合主成分分析与ANFIS模型的测试精度得到大幅度提高,说明主成分分析与ANFIS的混合方法对预测油井注水有效周期是可行与有效的.

To solve the problem of uncertain cycle of water injection in the oil field,a numerical method based on PCA-ANFIS is proposed for the prediction of effective cycle of water injection. Using PCA to reduce the dimension of original data,ANFIS is applied to train and test the new data. The correctness of PCA-ANFIS model is checked by the real injection statisties data of 116 wells from an oil field,the av- erage absolute error of test is 1.80 months. However,the average error is 4. 33 months without process- ing by PCA,the test accuracy has been greatly improved by PCA-ANFIS. Therefore,PCA-ANFIS meth- od is reliable to forecast the effective cycle of water injection,and which can help oil field developers to design the water injection scheme.