基于双层支持向量机的污水处理系统健康状态评估

为有效保障高速公路服务区污水处理系统的健康运行，结合污水处理系统的构成和系统运行特征，设计了一种用于评估污水处理系统健康状态的双层支持向量机（support vector machine，SVM）模型，并将健康状态分为健康、亚健康、故障和异常四个等级。首先，分析提升泵、进水泵、自吸泵等6个子系统的运行特征，使用随机函数和欠采样处理构建特征向量集。其次，采用SVM构建下层各子系统健康状态评估模型，并使用ThunderSVM算法来提高模型训练速度。然后，融合下层各子系统健康状态评估的结果，构建上层SVM系统评估模型，得到整个污水处理系统的健康状态。最后，依据青兰高速山西段服务区污水处理系统进行验证，结果表明所提出的双层SVM模型相较树类模型和单层SVM模型，不仅可评估整个污水处理系统的健康状态，还可评估各子系统的健康状态，且模型训练速度和判断精度均有提升。

Health Status Assessment of Sewage Treatment System Based on Double-layer Support Vector Machine Model

In order to effectively ensure the healthy operation of the sewage treatment system in the expressway service area, a double-layer support vector machine (SVM) model was designed to evaluate the health status of sewage treatment system based on the constitution and operation characteristics of sewage treatment system, and the health status was classified into four levels: health, sub-health, failure and anomaly. Firstly, the operating characteristics of six subsystems, such as lift pump, intake pump and self-priming pump, were analyzed, and feature vector sets were constructed by using random functions and undersampling processing. Secondly, the SVM was used to build the health status evaluation model of each sub-system in the lower layer, and the ThunderSVM algorithm was used to improve the training speed of the model. Then, the results of the health status assessment of the sub-systems in the lower layer were combined to construct the upper layer SVM system evaluation model to obtain the health status of the entire sewage treatment system. Finally, based on the verification of the sewage treatment system in the Shanxi section of Qinglan Expressway, the results show that the proposed double-layer SVM model than the tree model and single layer SVM model, can not only evaluate the health status of the whole sewage treatment system, but also evaluate the health status of each subsystem, and the model training speed and judgment accuracy are improved.