Fault Detection Of Dissolved Oxygen Sensor In Wastewater Treatment Plants

Wastewater treatment has attracted more and more attention with the increasing pollution of water environment and the shortage of water resources.The increasingly stringent effluent discharge standards require wastewater treatment plants（WWTP）to be equipped with advanced automatic control systems.And the effective operation of the control systems depends on the reliability of the sensors.The measuring instruments are easy to fail in the very bad environment of the WWTP.Although they can be maintained regularly by high quality sensors and professionals,but there is no lack of failure.It is necessary to configure sensor fault detection and diagnosis system in WWTP in order to reduce the negative impact of sensor failure because sensor failure will reduce the effluent quality and increase energy consumption.Dissolved oxygen is an important factor affecting the quality of effluent in the process of wastewater treatment and monitoring dissolved oxygen sensor is the key to ensure the quality of effluent.However,the dissolved oxygen sensor works in the control loop and it is difficult to detect it.At present,there are few researches on the fault detection of dissolved oxygen sensors in WWTP.And because of the complexity of the control system involved in the previous fault detection methods of dissolved oxygen sensors.This not only increases the cost of sensor,but also makes the detection time of fault detection longer.And the threshold is set by human experience.Therefore,a new fault detection method for dissolved oxygen sensor is proposed in this paper.The control system used in this method is simple and the time delay of fault detection is small.And the calculation of confidence intervals does not require any statistical characteristics.The main contents of this paper are summarized as follows:（1）Fault analysis of dissolved oxygen sensor: To verify the difficulty and importance of dissolved oxygen sensor failure detection,the effects of dissolved oxygen control process and dissolved oxygen sensor failure were analyzed respectively.The experimental results show that when the dissolved oxygen sensor fails,its failure is not seen from its signal itself.However,the oxygen transfer coefficient（KLa5）of the fifth unit deviates from the fault-free situation.And will make the WWTP outlet water quality is not up to the standard and operating energy consumption increase phenomenon.Therefore,the model of KLa5 can detect dissolved oxygen sensor failure,and the study of dissolved oxygen sensor fault detection is important.（2）Introduction and improvement of fault detection frontier methods of dissolved oxygen sensor: In the process of slow drift fault detection of adaptive fuzzy inference method,the threshold that needs to be defined is too much and the problem of complex detection process is improved.The experimental results show that the improved method can effectively detect the fault of dissolved oxygen sensor and is simpler than the detection process before improvement.However,both methods have the problems of long fault detection delay and high sensor cost,and the essential problem is that the definition of threshold is artificially set according to experience.（3）The establishment of interval prediction model is introduced: The core idea of interval prediction modeling in this paper is to model nonlinear systems using the approximation ability of RBF neural networks.Then use the AIC criterion to determine the order of the model and the number of hidden nodes in the network.The set description of the output weights of the network is obtained by using the set identification algorithm.This model can predict confidence intervals for system output by a single step.Confidence interval is an expression of the residual valve value.The calculation of this confidence interval does not require the statistical properties of any variables.Experimental results show that the nonlinear model order determined by the AIC criterion method accords with the real situation of the system,and the network structure is more compact.In the above case,the confidence interval obtained by the ensemble identification algorithm is less conservative.（4）Fault detection of dissolved oxygen sensor based on KLa5 interval prediction model: In view of the problems of airflow ratio and adaptive fuzzy reasoning,the control system is complex,the fault detection delay is long,and the definition of threshold is not clear.This paper presents a new fault detection method for dissolved oxygen sensors.By establishing a KLa5 interval prediction model to design the fault detection strategy for dissolved oxygen sensors.Finally,three fault scenarios are considered for simulation experiments.The experimental results show that the proposed method can effectively detect the fault of dissolved oxygen sensor,and the control system of this method is simple.The calculation of fault detection delay and confidence interval does not require any statistical characteristics.