废水处理过程的降维方法综述

Review on Dimensionality Reduction Methods in Wastewater Treatment Processes

废水处理过程变量存在强非线性、强耦合、易受干扰等问题,数据驱动的水质监控系统需要利用采集到的大量过程数据进行故障诊断和软测量建模。对废水处理过程数据进行降维预处理可以提高建模精度以及避免数据维度过高导致的“维数灾难”问题。综述了基于数据降维技术的故障检测、故障诊断和软测量建模方法在废水处理过程中的应用,归纳了高维数据降维的方法与类别,提出了数据降维方法在废水处理过程应用中存在的问题和相应的解决方案,最后指出了基于数据降维的废水处理过程建模方法的未来发展方向。

Wastewater treatment processes(WWTPs)have some problems,such as strong non-linearity among variables,multivariable coupling,frequent internal and external disturbances.A large number of parameters in WWTPs are measured and monitored to ensure effluent quality.Fault diagnosis and soft sensor modeling are carried out by using a huge amount of process data.Data dimensionality reduction techniques are widely applied in WWTPs to avoid the curse of dimensionality caused by high dimension of process data and improve the accuracy of process models.Dimensionality reduction methods for fault detection,fault diagnosis and soft sensing modeling used in WWTPs are reviewed.The existing problems and corresponding solutions in the research field are discussed.The future directions of modeling methods based on dimensionality reduction methods in WWTPs are pointed out.