Soft Sensors for Primary Clarifier in Industrial Effluent Treatment Plant

Abstract

The industrial effluent treatment plant or the wastewater treatment plant is a facility in which physical, chemical and biological processes are used to remove pollutants from industrial wastewater. Sedimentation is one of the main techniques of physical processes which removes insoluble and heavy particles from the wastewater. Activated sludge, biological filters, rotating biological contactors are the techniques to remove organic substances from the wastewater. Chemical processes like coagulation, precipitation, oxidation, neutralization etc. are used alongside biological and physical processes to achieve various water standards. Monitoring and control of wastewater treatment plant is important to meet the standards set by the government. The measurement of various pollutants is carried out by experimental analysis or online instrumentation. The experimental analysis procedure consumes much time while online instrumentation is expensive. One alternative solution is to develop and use model-based virtual or soft sensor to estimate or predict amount of pollutants present in the wastewater. The soft sensor is a mathematical model of the process or system used as a virtual sensor to estimate the variable or parameter of interest. The mathematical model can be derived based on the process knowledge (first-principles or white box model) or with the help of measured input and output variables (data-based or black box model) or combination thereof (hybrid or greybox model). Soft sensors for various wastewater and water treatment plant systems, sub-systems and processes have been found well researched and published in the literature over last couple of decades. However, the focus has been mainly to develop soft sensors for whole wastewater plant, water treatment process units or biological processes like activated sludge process, sequence batch reactor, membrane bioreactor, moving bed biofilm reactor, reverse osmosis, filtration processes etc. Also, the major focus has been on soft sensors based on various data driven techniques mainly based on multivariate statistics or artificial intelligence. But, the investigations on soft sensing techniques for upstream side physical processes like primary and secondary sedimentation or clarification in the wastewater treatment plants are neglected or least addressed. Although, mathematical models and computational fluid dynamics models are found to be used in design and sizing of such sedimentation process units. Based on literature study, the following few research gaps are identified as the areas of interest for this research work:

• • Lack of TSS (total suspended solids) soft sensors for effluent of primary clarifier in wastewater treatment plant. • • Use of first principle model for TSS soft sensors is not reported. • • Use and comparison of popular data driven soft sensor methods like fuzzy inference system, multivariate statistical techniques etc. are not available. • • Lack of lab-scale experimentation in order to carry out controlled dynamic study to help such research. The focus of this research work is to address the above research gaps through development, simulation, validation and investigation of the various soft sensor models for specific water-quality e.g. TSS in the primary clarifier process. To make this research work more relevant to the industrial wastewater treatment plants, the data from real-life industrial common effluent treatment plant (CETP) Vatva located at Ahmedabad, Gujarat, India were used to develop and validate the soft sensors. Also, lab-scale experimental clarifier setups are developed as a part of this research work in order to carry out detailed controlled experiments. Such set-up can be extended also to fellow researchers to carry out further research investigations. The main contributions and outcome of this research work are as follows: Contribution-1) Mechanistic and Fuzzy-inference based models for industrial CETP: Two types of soft sensors are developed for the prediction of effluent TSS of primary clarifier using – i) mechanistic model of the sedimentation process, and ii) fuzzy inference system (FIS). These soft sensors predict the effluent total suspended solids (TSSe) of the primary clarifier process based on the measurement of influent flow rate (Qf), influent total suspended solids (TSSin) and the clarifier design parameters. For both these types of soft sensors the range of Qf and TSSin were considered as 800-900 m3/h and 300-600 mg/L respectively based on the operating conditions of a real-life industrial CETP plant. The root mean square error (RMSE) of mechanistic model and fuzzy inference system were 30 mg/L and 48 mg/L respectively. The percentage mean accuracy for mechanistic model was 88%. The difference between predicted and actual measured values of RMSE and percentage mean accuracy of the mechanistic model was found promising considering the challenges and uncertainty in the data of real-life CETP plants. The mechanistic model, in addition to effluent TSS, also provides information about TSS in the bottom retentate and TSS-profile (i.e. TSS present at different height of the sedimentation or clarifier tank). Contribution-2) Investigations on multivariate statistics models for industrial CETP: Investigation on the use of multivariate statistical techniques like PCA (Principal Component Analysis), PLSR (Partial Least Squares Regression) and PCR (Principal Component Regression) were used to develop data-driven soft sensors for effluent COD (Chemical Oxygen Demand), BOD (Biochemical Oxygen Demand) and TOC (Total Organic Carbon) for the industrial primary clarifier. Although, the main objective of the primary clarifier is to remove the settable solids and TSS present in the influent wastewater, other pollutants like pH, COD, BOD, TOC, TDS, ammonical nitrogen(NH3-N) were also measured for the influent and effluent wastewater of primary clarifier for the CETP performance monitoring. The PCA was used to evaluate the collinearity among the influent variables. The collinearity among influent COD, BOD, TOC and TDS was captured by PCA. The data driven techniques, PLSR and PCR were adopted to develop the soft sensors for effluent COD, BOD and TOC as a function of measured influent variables. The performance of both regression models (i.e. PLSR and PCR) were evaluated using metrics like R2, normalized RMSE (root mean square error) and MAPE (mean absolute percentage error). The prediction performance for both these models (i.e. PLSR and PCR) were found similar. But, PLSR models required lesser number of components (or input variables) as compared to the number of components (or input variables) required by PCR models. Contribution-3) Lab-scale experimental clarifiers set-up and soft-sensor investigations: The motivation to develop lab-scale circular and rectangular primary clarifier setup was to enable capability of performing rigorous/detailed dynamic experiments in controlled operating conditions to help develop and evaluate soft sensors for primary clarifier. This is helpful, knowing that certain dynamic, controlled experiments are generally not possible or not viable to perform in a large operational industrial clarifier units like at CETP. Hence, during this research work, the lab-scale experimental circular and rectangular clarifier setups were developed, built and used to perform such experiments. These set-ups were used to develop TSS soft sensors for circular and rectangular primary clarifiers and to also compare performance of circular versus rectangular clarifiers. The mechanistic model developed as a soft sensor for the primary clarifier of CETP was also applied to these lab-scale circular and rectangular clarifiers to assess the performance of both these lab-scale clarifiers at different operating conditions. The experiments were carried out for three operating conditions (low, medium and high solid concentrations) using lab-scale setup. The simulation was performed in order to generate profile of TSS for both these clarifiers. The average R2 value 0.97 was observed between measured and simulated TSS present along the height of the circular clarifier. sections like laboratory scale set up overview, experiments of laboratory scale circular and rectangular clarifiers, simulation results and discussion of laboratory scale primary clarifiers. Finally, Chapter-6 includes concluding remarks drawn from this research work including the list of 5 publications as outcome of this research work. It also proposes some potential future scope to extend this soft sensing research and applications for primary clarifiers in common effluent treatment plants. Finally, the list of publications referred and cited in this research work are listed as References. Key words: wastewater, clarifier, soft sensor, total suspended solids, mechanistic model.