基于区间预测模型的污水处理厂传感器故障检测

Fault Detection of Sensors in Wastewater Treatment Plants Using Interval Predictor Models

污水处理厂配备许多传感器用于监测出水水质。传感器的正常工作与否对保证出水水质至关重要。给出了一种污水处理出水变量传感器故障检测方法。该方法根据入水和出水数据,采用径向基函数神经网络构造出水变量预测模型;使用参数线性集员辨识算法得到网络输出权值的集合描述,从而使预测模型能够给出出水变量的置信区间;以此置信区间为基础获得传感器的故障检测策略。由于置信区间描述了出水变量的存在范围,当传感器测量值超出置信区间,则可推断传感器发生故障。此外,在设计传感器故障检测策略时还考虑了污水处理过程异常的影响。实验结果证实所提方法的有效性。

A wastewater treatment plant(WWTP)is equipped with a number of sensors for monitoring the quality of the effluent.The normal operation of the sensors is critical to ensuring the quality of the effluent.This paper presents a fault detection method for sensors measuring the effluent variables in the WWTP.According to the available influent and effluent data,the radial basis function neural network is used to construct the predictor models for the effluent variables.Linear-in-parameters set membership identification algorithm is used to obtain a description of the uncertain set of the vector representing the output weights of the neural network,so that the predictor model can give a confidence interval of the effluent variable.Basing on this confidence interval,the sensor fault detection strategy is obtained.Since the confidence interval describes the existence range of the effluent variable,when the measurement result exceeds the confidence interval,it can be inferred that the sensor has been faulty.Besides,the influence of abnormal operation of the WWTP is considered in designing the sensor fault detection strategy.The experimental results confirm the effectiveness of the proposed method.