Research On Soft-sensor Modeling And Its Application To Wastewater Treatment

As a key part in the water environmental protection chain,the safe and stable operation of wastewater treatment plant is the prerequisite to ensure the quality of effluent.A major requirement for achieving this goal relies on the availability of real-time monitoring of key or primary process indicators.These indicators reflect the important information of wastewater treatment processes,but most of them are hard-to-measure or not easy to measure.Their realtime availability is often associated with expensive capital and maintenance costs,as well as being characterized by time-delayed responses that are often unsuitable for real-time monitoring.In order to deal with the online real-time acquisition of these variables,the softsensing techniques provide a feasible and effective solution.This dissertation surveys and discusses the modeling of data-driven soft-sensing techniques in municipal wastewater treatment plants.The main contributions are described as follows.1.The online prediction performance of the commonly used data-driven soft sensor modeling methods for wastewater treatment is studied.There are many kinds of data-driven soft-sensor modeling methods used for wastewater treatment,showing satisfactory results in their special applications.However,few literature has carried on the inductive research on the application effect of the commonly used basic modeling methods.Wastewater processes have some unique features compared to other process industries.The practical performance of these modeling methods is not only helpful to deepen the understanding of the characteristics of wastewater processes,but also providing a basic reference for the subsequent research of soft sensor modeling.Therefore,the online prediction performance of four commonly used datadriven soft-sensor modeling methods for wastewater treatment is compared and analyzed through a real data collected from a WWTP.Some conclusions are obtained for the reference of the subsequent research.2.A deep neural network soft-sensor model for wastewater treatment is proposed.Supervised neural network methods are one of the most popular data-driven soft-sensor modeling methods in wastewater treatment.However,most of them are shallow architecture,which are incapable of performing tasks effectively when got stuck in extremely complex situations,such as severe weather conditions.One of the potential solutions is resort to the neural networks with multi-layer structure,i.e.the deep neural networks.However,deep architecture neural networks suffered the difficulty of training,poor generalization ability,and so on,and thus be few successful application cases in early period.Furthermore,wastewater treatment has its unique practical matters,leading to the difficulty of collecting adequate data.Therefore,few literature devoted deep neural networks to the soft sensor modeling for wastewater treatment.In response to this situation,a deep neural network soft-sensor model for wastewater treatment is constructed based on a deep learning model: stacked auto-encoder.Meanwhile,a genetic algorithm searching strategy is proposed to determinate the number of nodes in hidden layers.The strategy can search for proper values in a specific range.Finally,the proposed model is validated by two case studies.3.A multi-output adaptive soft-sensor model based on deep neural network for wastewater treatment is proposed.In the wastewater treatment,the existence of several important but hardto-measure process variables hinders not only the monitoring of productive processes,but also the adjustment or optimization of process control strategies.Even though the soft-sensor models are reasonably constructed,which also suffer the degradation problem,resulting in high maintenance cost.Additionally,the proper secondary variables not only facilitate the modeling,but also reduce the cost of on-line instruments.Therefore,a multi-output adaptive soft sensor model based on deep neural network is proposed,which used for simultaneous online prediction of multiple target variables in wastewater treatment.The deep neural network is constructed on the basis of a stacked auto-encoder,displaying satisfactory online prediction performance under extremely complex scenarios.In order to deal with the degradation problem and select proper secondary variables,a time difference modeling method and VIP(Variable importance in projection)method are assimilated in modeling.Finally,the proposed model is validated through a case study.4.To address the ineffective monitoring of sludge bulking,a multiple-step-ahead softsensor prediction model is proposed.Filamentous sludge bulking is a common but intractable obstacles in the activated sludge processes,which seriously affects the wastewater treatment,even paralyzes the operation of the facilities.The soft-sensing technology based on the estimation of current time is incapable of predicting the sludge bulking ahead of time,and thus miss the best time to deal with it.For this reason,a multiple-step-ahead soft-sensor model is proposed by extending soft sensing to the multiple-step-ahead prediction scheme,which is capable of making ahead prediction of targets.It is achieved based on a indirect multi-step prediction strategy by constructing the same step of secondary variables with the predicted variables to reduce the prediction error,and thus improves the prediction accuracy.Finally,the proposed model is verified through a real data collected from the field influenced by filamentous sludge bulking.Results showed that the proposed model could make 7 days ahead prediction of SVI(Sludge Volume Index)stably,which is capable of monitoring the sludge bulking.At last,the main research work and corresponding results of this dissertation are briefly concluded,and the future researches are discuessed.