基于集成相关向量机的水质在线预测模型

Online prediction model of water quality based on ensemble RVM

摘要:

针对污水处理过程存在着强非线性和非稳态运行等特征,传统传感器维护成本高昂且无法快速准确地测量生化需氧量(BOD)等水质指标的问题,提出一种基于集成相关向量机的水质在线预测模型。该模型首先采用相关向量机(RVM)为弱预测器,利用改进的AdaBoost.RT算法将多个弱预测器集成为强预测器,实现了污水处理过程中水质的在线预测。仿真实验结果表明,该水质在线预测模型预测精度高,综合性能突出,克服了单一预测器随着异常点增多,模型泛化能力下降和鲁棒性不足的问题,能较好地实现了污水处理过程中的水质在线预测。

Abstract:

Wastewater treatment exists strong nonlinearity, unsteady operation and other characteristics, traditional hardware transducer are with huge maintenance problems and make it extremely difficult to obtain water-quality index quickly and accurately, such as BOD. Concerning the concert problems, an online prediction model of water quality based on ensemble RVM is proposed. Firstly, set RVM as weak predictor and then use improved AdaBoost.RT to embody several weak predictor into strong predictor. The simulation experiments demonstrated that this online prediction model has higher precision, better generalization ability, and overcomes the less effectiveness and robust problem of single predictor induced by increasing abnormal points. Therefore, the proposed model can meet the requirements of online prediction of water quality of wastewater treatment process.