基于遗传算法优化单类支持向量机的油田离心泵注水站异常检测

目前油田离心泵注水站多采用传统的人工巡检等方式进行异常检测，存在浪费大量资源且检测精度不高的情况。针对此问题，提出一种基于遗传算法优化单类支持向量机(genetic algorithm optimized one-class support vector machine, GA-OC-SVM)的注水站异常检测方法。首先，对注水站数据进行标准化、归一化处理以及特征提取；其次，使用遗传算法进行寻优得到最佳种群个体值作为单类支持向量机的参数，建立检测模型；最后，将GA-OC-SVM算法与孤立森林算法、局部离群因子算法等主流方法用于测试数据集的异常检测对比，并分析算法的精度，采用接受者操作特征(receiver operating characteristics, ROC)曲线进行模型评价。结果表明所提出的GA-OC-SVM算法更优，检测精度达到99%,同时能够节省大量的人力物力资源。

Genetic algorithm based optimization of one-class support vector machine for abnormality detection in oilfield centrifugal pump injection stations

At present, most of the oilfield centrifugal pump injection stations use traditional manual inspection and other methods for anomaly detection, which wastes a lot of resources and has low detection accuracy. To address this problem, this paper proposes a genetic algorithm optimized one-class support vector machine (GA-OC-SVM) based anomaly detection method for water injection stations. Firstly, the water injection station data are normalized, normalized and feature extracted, and then the genetic algorithm is used to find the best population individual values as the parameters of the one-class support vector machine and establish the detection model. Finally, the GA-OC-SVM algorithm is compared with the mainstream methods such as isolated forest algorithm and local outlier algorithm for anomaly detection of the test dataset, and the accuracy of the algorithm is analyzed and the ROC curve is used for model evaluation. The results show that the proposed GA-OC-SVM algorithm is superior, with a detection accuracy of 99%, and can save a lot of human and material resources at the same time.