Fault Diagnosis Of Wastewater Treatment Plant Based On Interval Prediction Model

Nowadays,with the rapid development of industrialization,urbanization and population growth,wastewater discharge has also increased and water pollution has become a topic of concern.In order to better solve the problem of water pollution,it is imperative to strengthen the research work related to wastewate purification treatment.However,the wastewater treatment process often has a relatively complicated biochemical reaction process,which involves a large number of detection and monitoring equipment,and is susceptible to many factors such as seasons and water flow,and failures often occur in wastewater treatment plant.Therefore,accurate and effective fault detection and diagnosis of the entire wastewater treatment plant is particularly important.Due to the adverse effects of these factors,it is difficult to establish a precise mechanism model for the wastewater treatment process,but this process will generate a large amount of historical data containing process information.Therefore,in recent years,data-driven wastewater treatment plant fault diagnosis methods have been extensively studied.For the fault diagnosis methods based on datadriven,multivariate statistical method and neural network model are widely used.They all use the residual statistical properties to diagnose faults.Considering the influence of many uncertain factors,such as disturbance and noise in the wastewate treatment process and so on,the residual statistical characteristics are difficult to obtain.Therefore,a new fault diagnosis method is proposed in this paper,which combine the set membership identification with the radial basis function(RBF)neural network to establish an interval prediction model,and constructs a fault diagnosis strategy based on this interval prediction model.The confidence interval is an expression of the residual threshold,the calculation of this interval does not require the statistical properties of any variable,so it has good practicability.This method has been applied to the fault detection and diagnosis of wastewate treatment plant.And the simulation platform and experimental results analysis confirm that the method mentioned in this paper has good reliability and effectiveness.The main research contents of this article are summarized as follows:(1)Wastewater treatment benchmark simulation model construction : Briefly introduce the process flow of activated sludge wastewater treatment and analyze its characteristics in detail;then,it is done for the benchmark simulation model(BSM1)jointly developed by the European Union Scientific and Technical Cooperation(COST)and the International Water Association(IWA).Further in-depth study,fully understand the important information of BSM1 internal structure,biochemical reaction pool and secondary sedimentation tank;finally,build the experimental platform in Matlab environment,establish a simulation model for simulation experiment analysis,and conclude that the existing data is used for verification.The model fully proves the reliability and effectiveness of the model,and provides a simulation platform for the fault detection and diagnosis strategy of the wastewater treatment plant.(2)Interval prediction model:For systems with unknown but bounded noise interference,the RBF neural network is used to approximate it.The set membership identification algorithm is used to estimate the output weight of the RBF neural network,and the output weight set is described.In the process of system operation,the model can predict the actual system's prediction interval.Taking a nonlinear system as an example,the simulation experiment is carried out,and it is compared with the least squares identification RBF neural network output weight interval prediction method.It is concluded that the method mentioned in this paper has significant superiority.It lays a solid theoretical foundation for the fault detection and diagnosis application of wastewater treatment plants.(3)Application of interval prediction model in fault detection and diagnosis of wastewater treatment plant : Simulate three common fault scenarios in the actual wastewater treatment plant on BSM1: process variable sensor failure,effluent water quality variable sensor failure and wastewater treatment process anomaly;combine the data obtained by BSM1,establish interval prediction model and construct fault detection and diagnosis of wastewater treatment plant Strategy;Finally,the simulation experiment is carried out and it is concluded that the method mentioned in this paper can effectively complete the fault detection and diagnosis of the wastewater treatment plant sensor.