基于遗传算法优化支持向量机模型的套损预测

Casing Damage Prediction Model Based on GA-SVM Method

为快速有效预测油田套管损坏情况,基于实际开发资料分别选取水井和油井的影响因素作为模型训练数据,通过采用遗传算法(GA)进行参数优化的支持向量机(SVM)作为模型,选用大庆油田南三区1 220口油井及648口水井作为数据集,其中80%作为训练集,20%为测试集,最后将GA-SVM模型预测结果和SVM模型、BP神经网络模型预测结果进行对比分析,结果表明,将遗传算法优化后的支持向量机应用到套损预测中可以得到较高的预测精度,在油田套损预测方面具有一定的参考价值。

To predict casing damage timely and efficiently,the influencing factors on the water well and oil well based on actual developed data were selected as the training data of the model,support vector machine(SVM)the parameters of which were optimized by genetic algorithm(GA)are used as the model,and 1,220 oil wells and 648 water wells of the south third zone of Daqing Oilfield were taken as a data set,80%was used as the training sets,and 20%as the test set parameters,and then the predicted results of GA‐SVM model were compared with those of SVM model and BP neural network model.The comparison results showed that the support vector machine optimized by genetic algorithm obtained high prediction accuracy,and it had great valuable reference in predicting casing damage in oil field.