This paper addresses the implementation of economic-oriented model predictive controllers for the dynamic real-time optimization of the operation of wastewater treatment plants (WWTP). Both the economic-optimizing controller (pure-EMPC) and the economic-oriented tracking controller (Hybrid-EMPC, or HEMPC) formulations are validated in the benchmark simulation model (BSM1) platform that represents the behavior of a characteristic activated sludge process. The objective of the controllers is to ensure the appropriate operation of the plant, while minimizing the energy consumption and the fines for violations of the limits of the ammonia concentration in the effluent along the full operating period. A non-linear reduced model of the activated sludge process is used for predictions to obtain a reasonable computing effort, and techniques to deal with model-plant mismatch are incorporated in the controller algorithm. Different designs and structures are compared in terms of process performance and energy costs, which show that the implementation of the proposed control technique can produce significant economic and environmental benefits, depending on the desired performance criteria.

In this paper, a control approach for improving the overall efficiency of a wastewater treatment plant (WWTP) is presented. It consists of a cascaded control system that uses a global performance indicator as the controlled variable to drive the plant to operating conditions that satisfies trade-offs involved in the WWTP operation, improving the global performance of the plant. The selected global performance indicator is the N/E index that measures the ratio between the amount of nitrogenated compounds eliminated (kgN) and the energy (kWh) required to achieve that goal. This index links the variables of the activated sludge process with the energy consumed in the whole plant, thus the control strategy takes actions based on plantwide considerations. An external Proportional Integral (PI) controller changes the DO set point according to the N/E index and the basic dissolved oxygen (DO) control scheme in the activated sludge process follows this reference changes varying the aeration intensity. An outer loop with an event-based controller is used to compute the index values when the DO concentration is driven to excessively low limits, preventing long operation periods in this undesirable condition. Simple proportional integral controllers (PI) are used to adapt the strategy to the automation systems available in WWTPs. The implementation in the Benchmark Simulation Model 2 (BSM2) demonstrates the potential of the proposed approach. The results show the possibilities of the N/E index to be used as an indicator of global performance of WWTPs. It provides a link between water line objectives and energy consumption in the whole plant that can be exploited to introduce plantwide considerations in alternative control strategies formulated to drive the plant to operating conditions that optimize the overall process efficiency.

In this work, deep reinforcement learning methodology takes advantage of transfer learning methodology to achieve a reasonable trade-off between environmental impact and operating costs in the activated sludge process of Wastewater treatment plants (WWTPs). WWTPs include complex nonlinear biological processes, high uncertainty, and climatic disturbances, among others. The dynamics of complex real processes are difficult to accurately approximate by mathematical models due to the complexity of the process itself. Consequently, model-based control can fail in practical application due to the mismatch between the mathematical model and the real process. Control based on the model-free reinforcement deep learning (RL) methodology emerges as an advantageous method to arrive at suboptimal solutions without the need for mathematical models of the real process. However, convergence of the RL method to a reasonable control for complex processes is data-intensive and time-consuming. For this reason, the RL method can use the transfer learning approach to cope with this inefficient and slow data-driven learning. In fact, the transfer learning method takes advantage of what has been learned so far so that the learning process to solve a new objective does not require so much data and time. The results demonstrate that cumulatively achieving conflicting objectives can efficiently be used to approach the control of complex real processes without relying on mathematical models.