Review of Modern Methods for Identification, Forecasting, and Intelligent Control of Wastewater Biological Treatment Processes

This article presents a comprehensive review of contemporary approaches to the automation and intelligent control of wastewater biological treatment processes. Particular emphasis is placed on the digitalisation of wastewater treatment plants, ranging from the implementation of automated process control systems (APCS/SCADA-based solutions) to the application of predictive algorithms and the development of digital twins of bioreactors. Special attention is devoted to mathematical models that underpin the control of bioprocesses. The evolution of the most widely used activated sludge models—ASM1, ASM2d, and ASM3—is examined, as these models describe key processes such as microbial community growth, nitrification, denitrification, and phosphorus removal. It is shown that modern model parameter identification techniques, including extended and ensemble Kalman filters as well as Bayesian approaches, enable accurate real-time estimation of process characteristics. This, in turn, contributes to more precise and robust process control. A separate focus is placed on intelligent soft sensors that integrate physicochemical models with machine learning algorithms or neural network-based solutions. These sensors are capable of predicting key wastewater quality parameters—such as chemical oxygen demand (COD), ammonium and nitrate concentrations, and suspended solids—and play a crucial role in adaptive control frameworks. Their outputs serve as inputs for more advanced predictive controllers. In addition, the article reviews a range of intelligent control strategies, from auto-tuning PID controllers and Model Predictive Control (MPC) systems to fuzzy and neuro-fuzzy controllers, as well as approaches based on Reinforcement Learning (RL). The latter are increasingly integrated into digital twins of wastewater treatment plants, creating the foundations for fully autonomous control systems. A comparative analysis of the reviewed methods demonstrates that the application of intelligent control can reduce energy consumption for aeration by an average of 15–30%, maintain dissolved oxygen concentrations within ±0.2 mg/L, and enable adaptive adjustment of technological operating regimes without operator intervention. The review covers more than 45 scientific publications from the past decade and clearly illustrates the transition of the sector from isolated solutions towards integrated digital control platforms. Future developments are associated with the creation of hybrid models combining mechanistic principles and artificial intelligence, the implementation of economically oriented MPC systems, the deployment of learning RL agents, and ultimately the development of self-learning control systems for biological treatment processes.