Innovative Surveillance and Process Control in Water Resource Recovery Facilities

Water Resource Recovery Facilities (WRRF), previously known as Wastewater Treatment Plants (WWTP), are getting increasingly complex, with the incorporation of sludge processing and resource recovery technologies. Along with maintaining a stringent effluent water quality standard, the focus is gradually shifting towards energy-efficient operations and recovery of resources. The new objectives of the WRRF demand an economically optimal operation of processes that are subjected to extreme variations in flowrate and composition at the influent. The application of online monitoring, process control, and automation in WRRF has already shown a steady increase in the past decade. However, the advanced model-based optimal control strategies, implemented in most process industries, are less common in WRRF. The complex nature of biological processes, the unavailability of simplified process models, and a lack of cost-effective surveillance infrastructure have often hindered the implementation of advanced control strategies in WRRF. The ambition of this research is to implement and validate cost-efficient monitoring alternatives and advanced control strategies for WRRF by fully utilizing the powerful Internet of Things (IoT) and data science tools. The first step towards implementing an advanced control strategy is to ensure the availability of surveillance infrastructure for monitoring nutrient compositions in WRRF processes. In Paper A, a soft sensor, based on Extended Kalman Filter, is developed for estimating water-quality parameters in a Sequential Batch MBBR process using reliable and inexpensive online sensors. The model used in the soft sensor combines the mechanistic understanding of the nutrient removal process with a statistical correlation between nutrient composition and easy-to-measure parameters. Paper B demonstrates the universality of the soft sensor through validation tests conducted in a Continuous Multistage MBBR pilot plant. The drift in soft-sensor estimation caused by a mismatch between the mathematical model and process behavior is studied in Paper B. The robustness of the soft sensor is assessed by observing estimated nutrient composition values for a period of three months. A systematic method to calibrate the measurement model and update model parameters using data from periodic lab measurements are discussed in Paper B. The term SCADA has been ubiquitous while mentioning online monitoring and control strategy deployment in WRRFs. The present digital world of affordable communication hardware, compact single board processors, and high computational power presents several options for remote monitoring and deployment of soft sensors. In Paper C, a cost-effective IoT strategy is developed by using an open-source programming language and inexpensive hardware. The functionalities of the IoT infrastructure are demonstrated by using it to deploy a soft sensor script in the ContinuousMultistage MBBR pilot plant. A cost-comparison between the commercially available alternatives presented in Paper A and the open-source IoT strategy in Paper B and Paper C highlights the benefits of the new monitoring infrastructure. Lack of reliable control models have often been the cause for the poor performance of advanced control strategies, such as Model Predictive Controls (MPC) when implemented to complex biological nutrient removal processes. Paper D attempts to overcome the inadequacies of the linear prediction model by combining a recursive model parameter estimator with the linear MPC. The new MPC variant, called the adaptive MPC (AMPC), reduces the dependency of MPC on the accuracy of its prediction model. The performance of the AMPC is compared with that of a linear MPC, nonlinear MPC, and the traditional proportional-integral cascade control through simulator-based evaluations conducted on the Benchmark Simulator platform(BSM2). The advantages of AMPC compared to its counterparts, in terms of reducing the aeration energy, curtailing the number of effluent ammonia violations, and the use of computational resources, are highlighted in Paper D. The complex interdependencies between different processes in a WRRF pose a significant challenge in defining constant reference points for WRRFs operations. A strategy that decides control outputs based on economic parameters rather than maintaining a fixed reference set-point is introduced in Paper E. The model-based control strategy presented in Paper D is further improved by including economic parameters in the MPC’s objective function. The control strategy known as Economic MPC (EMPC) is implemented for optimal dosage of magnesium hydroxide in a struvite recovery unit installed in a WRRF. A comparative study performed on the BSM2 platform demonstrates a significant improvement in overall profitability for the EMPC when compared to a constant or a feed-forward flow proportional control strategy. The resilience of the EMPC strategy to variations in the market price of struvite is also presented in Paper E. A combination of cost-effective monitoring infrastructure and advanced control strategies using advanced IoTs and data science tools have been documented to overcome some of the critical problems encountered in WRRFs. The overall improvement in process efficiency, reduction in operating costs, an increase in resource recovery, and a substantial reduction in the price of online monitoring infrastructure contribute to the overall aim of transitioning WRRFs to a self-sustaining facility capable of generating value-added products