October 1, 2021 – September 30, 2024
ARRS - Slovenian Research Agency, L2-3166, co-financing from Public Utility JP Centralna čistilna naprava Domžale-Kamnik d.o.o., Public Utility Komunala Kranj, javno podjetje d.o.o., Public Utility Komunala Novo mesto d.o.o., Kolektor Sisteh d.o.o.
Wastewater treatment plants (WWTPs) are public facilities purifying urban and industrial wastewater before its release to the environment. In Slovenia, large investments were made in construction and modernisation of sewage infrastructure and WWTPs in the last decade. It is of vital importance that after (re)construction, the plants will operate efficiently both from the environmental and economic point of view. Besides treating wastewater, WWTPs are becoming important also as a source of energy (chemical energy in wastewater organic fractions) and resource recovery (phosphorous, nitrogen). This is an important issue in view of the societal transition to renewable energy sources and circular economy. Energy positive sewage treatment, wastewater reuse, nutrient recovery, carbon neutrality are some of the new challenges, which are transforming WWTPs into Nutrient+Energy+Water factories and water resource and recovery facilities (WRRFs). By so expanding their functionalities, WWTPs are becoming complex and demanding to operate. They consist of a large number of complex biological processes and other process units, interacting with each other via recycling loops. Besides, they pursue a variety of mutually conflicting operational goals. Therefore, their operation should be addressed from a plant-wide perspective. In the past years, a plant-wide consideration motivated a significant advancement in plant-wide modelling and the design of plant-wide Benchmark Simulation Models. On the other hand, new developments in plant-wide control and optimization are much slower. In addition, a significant gap is noticed between simulation studies and full-scale applications. The project aims to design and test a supervisory control system for plant-wide optimization, which will optimize plant performance criteria on a top layer. Plant operational parameters will be adjusted on-line with the purpose to optimize clean water production, energy generation and nutrients recovery. Optimization will be performed by extremum seeking control. As a non-model based real-time optimization approach, this presents a new paradigm in this area. While models are indispensable in the analysis and understanding of WWTP operation, so far they have presented less successful in WWTP control applications. The reason is that models are inaccurate due to many non-measurable disturbances, complex non-linear process dynamics and slow process changes. A non-model optimization approach is also easier to adjust to different wastewater technologies and plant configurations. The developed approach will be compared with model-based solutions. Scientific contributions of the supervisory control system design will be the following: definition and mathematical formulation of WWTP performance criteria, identification of optimization variables that optimize the performance, and the design of an on-line extremum seeking optimization algorithm. The algorithm will have a penalty function for handling constraints, a solution for including available theoretical and expert knowledge to improve the convergence rate, low-pass filtering for operation under non-stationary conditions, and a stochastic optimization to avoid local minima. The research will be undertaken by an interdisciplinary project group with complementary skills in control technology, optimization, knowledge technologies, wastewater treatment and ICT. Research results will support activities within IWA (International Water Association) Specialist Group on Modelling & Integrated Assessment and Task Group on Good Modelling Practice in WRRFs, as well as impact the development of WWTP plant-wide control, which is still in its infancy. Besides conceptual design and testing of the approach in simulation, an important impact is also testing and potential implementation on real plants. By implementing the designed solutions at users it is envisaged to improve the treatment performance and reduce operational costs by about 10%.
HVALA, Nadja, KOCIJAN, Juš. Input variable selection using machine learning and global sensitivity methods for the control of sludge bulking in a wastewater treatment plant. Computers & chemical engineering. [Print ed.]. 2021, vol. 154, str. 107493-1-107493-10. ISSN 0098-1354. DOI: 10.1016/j.compchemeng.2021.107493. [COBISS.SI-ID 74445571].
GRANCHAROVA, Alexandra, VALKOVA, Ivana, HVALA, Nadja, KOCIJAN, Juš. Distributed predictive control based on Gaussian process models. Automatica. [Online ed.]. 2023, vol. 149, str. 1-9, ilustr. ISSN 1873-2836. DOI: 10.1016/j.automatica.2022.110807. [COBISS.SI-ID 135932419].