Transfer Learning in Wastewater Treatment Plant Control Design: From Conventional to Long Short-Term Memory-Based Controllers

Abstract

In the last decade, industrial environments have been experiencing a change in their control processes. It is more frequent that control strategies adopt Artificial Neural Networks (ANNs) to support control operations, or even as the main control structure. Thus, control structures can be directly obtained from input and output measurements without requiring a huge knowledge of the processes under control. However, ANNs have to be designed, implemented, and trained, which can become complex and time-demanding processes. This can be alleviated by means of Transfer Learning (TL) methodologies, where the knowledge obtained from a unique ANN is transferred to the remaining nets reducing the ANN design time. From the control viewpoint, the first ANN can be easily obtained and then transferred to the remaining control loops. In this manuscript, the application of TL methodologies to design and implement the control loops of a Wastewater Treatment Plant (WWTP) is analysed. Results show that the adoption of this TL-based methodology allows the development of new control loops without requiring a huge knowledge of the processes under control. Besides, a wide improvement in terms of the control performance with respect to conventional control structures is also obtained. For instance, results have shown that less oscillations in the tracking of desired set-points are produced by achieving improvements in the Integrated Absolute Error and Integrated Square Error which go from 40.17% to 94.29% and from 34.27% to 99.71%, respectively.

Long short-term memory-based modeling and controlling method for wastewater treatment process

“…27–29 It is widely used in sentiment analysis, 30–32 image processing, 33,34 language modeling, 27 data forecasting, 35,36 and other fields. 37 Liu et al 38 proposed a data-driven method using LSTM network to provide accurate future capacities prediction and reliable uncertainty management for Li–ion batteries. Aiming at the nonlinear relationship between water quality indicators, Yan et al 39 used the time series prediction ability of the LSTM network to predict the lack of water quality indicators at a certain point in time.…”

Intelligent Control of Wastewater Treatment Plants Based on Model-Free Deep Reinforcement Learning

“…Machine learning in wastewater treatment plants [25,26] stands out for predicting the risk of violating pollution legal effluent limits. Thus, supervised networks usually participate as predictors in control strategies focused on avoiding these violations [27,28].…”

Advances in the Monitoring, Diagnosis and Optimisation of Water Systems

“…Work presented in [ 5 , 6 , 7 , 8 ] discusses the water-sanitation side. In this field, there is a growing interest in the adaptation and use of technologies related to the circular economy which promote environmental sustainability, where resource recovery is a key issue for industrial and environmental processes and involves a wide spectrum of study possibilities.…”

“…Different nature-inspired optimisation algorithms are compared in the performance of this task, providing potential dramatic improvement when compared with actual nonoptimised operation. The improved operation of WWTP is also sought in [ 5 , 7 ] by means of improving the controllers involved in the operation of certain key processes of the WWTP, e.g., the aeration process of biological reactors. Classic proportional-integral (PI) controllers have been traditionally considered as the control strategy for such processes; however, improved performance may be achieved with more complex structures and techniques, e.g., model predictive-control (MPC) schemes or artificial neural network (ANN) approaches.…”

“…ANN schemes may provide an alternative to this issue since they are well suited to deal with such processes. In this line of work, [ 7 ] considers transfer-learning (TL) methods to train ANN nets supporting control operations in WWTPs, and compares this approach with traditional control schemes, providing improved control performance while reducing control-design complexity and time invested in the ANN training process, which can be considerably time-demanding. Last but not least, in this Special Issue collection, a soft-sensing approach to predict key performance indicators (KPIs) in water-quality monitoring and control of WWTPs—such as effluent biochemical oxygen demand (BOD) or ammonia nitrogen (NH3-N)—is presented in [ 8 ].…”