Modelling and optimisation of wastewater coagulation process
S. S. Rathnaweera · BIBSYS Brage (BIBSYS (Norway)) · 2010
The efficiency and economics of chemical coagulation, one of the most robust wastewater treatment processes, is critically dependent on the optimal coagulant dosage which is strongly related to the influent quality. The process basics indicate that the optimal coagulant dosage is dependent on several parameters such as flow, particles, phosphates and pH which are not proportional to each other, thus cannot be represented by a single parameter alone. However, the usage of two or more parameters in wastewater coagulation control is yet to be seen in full scale applications as a common strategy. A survey among Norwegian wastewater treatment plants indicates that over 80% only used flow, or sometimes combined with pH, as the dosing control strategy. The recent developments in the on-line water quality sensors have brought about the practical possibilities to use them in treatment plants. Despite this, active usage to unveil the hidden potential of these instruments and control concepts is still scarce. The main challenge now lies in establishing robust and accurate mathematical models to describe the dosage vs influent & effluent parameters, as well as concepts to make the online measurements more accurate and valid. A multiple-parameter and multiple-model based coagulant dosage control (XCDC) concept has been developed, tested and elaborated in this thesis. The basis for this concept was preliminarily evaluated by Lu (2003), where a single model based basic modelling results were presented. The present study elaborates an expanded investigation and results by selecting the best suitable multivariate calibration system for model development, a system to validate and manage online water quality monitoring data, a multiple-model system to manage non-validated values, full scale tests under various conditions and finally contributing to the restructuring of the software enabling universal implementation at wastewater treatment plants (WWTPs). The studies were conducted at four WWTPs in both Norway and China. Full scale tests were conducted in NRA WWTP, Lillestrøm, Norway, HIAS WWTP, Hamar, Norway and Xiao Hong Men WWTP, Beijing, China. Furthermore, a pilot-scale study was conducted in Gaobeidian WWTP, Beijing, China. Coagulation is a well defined process, which can be mathematically described (Ratnaweera et al. 1994). However, the construction of a conceptual model has been a challenge due to the complex nature of the process. Few successful attempts to construct relationships between coagulant dosage and the water quality parameters are reported. In the present study, three different multivariate analytical methods, MLR, PCR and PLSR, were evaluated to find the best suitable regression analytical method for the purpose. Considering the fact that most related studies report the use of these systems to describe the historical relationships, this thesis focuses on the validity of these models’ predictability of future situations enabling usage in online process control. A robust and accurate error recognising and validating system for online measurements is crucial when they are used in process control as they may create critical conditions. The commonly available hardware and set-point based error detection methods are inadequate in real time process control. The experimental coagulant dosing control studied is evaluated with an efficient and accurate error recognising and validating system. The concept is based on a software based floating error detection system developed using multivariate calibration systems. When the measurements are validated and errors are identified, a management system to minimise their impacts on the real time process control is necessary to secure the accuracy of the process. A robust multiple model based strategy was integrated to the multi parameter based experimental coagulant dosage control system. The concept is based on a set of models with a variable number of water quality parameters, enabling the activation of the best suitable dosage estimation equation at all times. Though the experimental coagulant dosing control system required complicated programming structure, it was possible to integrate it in to a simple, commercially available Programmable Logical Controller (PLC). The PLC could then integrate in to the treatment plants’ main supervisory control and data acquisition (SCADA) systems. The system at NRA is successfully running with over 12% of coagulant saving with considerable reduction of the sludge production. Further studies suggest the ability of saving 16% or more compared with historical data. The HIAS system was successfully run until a shock-loading due to two very different influent types required usage of two model sets to manage variations. The experiments showed the possibility to save 5% to 15% compared with the traditional coagulant consumption. The treatment plants in Beijing, China showed savings up to 31% while maintaining the same effluent qualities. Further studies showed the possibility to reduce the coagulant demand 2.4 to 7.8 times by changing the present dosage strategies combined with use of better coagulants.