Periodic Reporting for period 2 - DARROW (DRIVING THE FUTURE OF WATER RESOURCE RECOVERY FACILITIES THROUGH DATA INTELLIGENCE)
Período documentado: 2023-09-01 hasta 2025-02-28
The wastewater sector is going through a profound transformation whose first steps began a few years ago when energy efficiency and resource recovery became key priorities in wastewater treatment plants (WWTPs) and these installations started to be perceived as Water Resource Recovery Facilities (WRRFs). Furthermore, WRRFs aim to generate products of sufficient value to make the process economically viable and enhance resource productivity while simultaneously remediating wastewater to an acceptable quality. WRRFs are focused on the conversion of the organic carbon, nitrogen, phosphorus and associated trace nutrients in the wastewater stream to added value products while providing clean or ‘fit for use’ water as a product.
The concept of a WRRF contributes to valuing wastewater treatment as an integrated component of a wider system rather than a unit process for ‘end-of-pipe treatment’. About 13 % of the fertilizer nutrient demand in EU could be supplied by WRRFs. The annual production of methane in EU WRRFs can reach about 9.5 billion m3 with a caloric value of 100 billion kWh. which can serve electricity to 25 million households. With an estimated EU production of 15 million t of sludge waste in WWTPs, resource recovery strategies could result in annual sludge waste reductions of 4 million t. Finally, the treatment of municipal wastewater amounts to 34 billion kWh/year in the EU, meaning 34 % of the recoverable energy from WRRFs. Thus, the implementation of new design patterns and operational strategies can allow WRRFs to be self-sufficient.
Based on this context, DARROW aims to build and demonstrate an innovative, optimised, modular, and flexible data-driven AI solution to make existing WWTPs more autonomous, more energy efficient and better prepared for their transformation into Water Resource Recovery Facilities (WRRFs). This AI solution will be tested in Tilburg WRRF. DARROW will take advantage of existing AI & Data analysis techniques to analyze, improve, particularize and demonstrate them with the final objective of contributing to a greener planet, thus maximizing contribution to the Green Deal.
The concept of a WRRF contributes to valuing wastewater treatment as an integrated component of a wider system rather than a unit process for ‘end-of-pipe treatment’. About 13 % of the fertilizer nutrient demand in EU could be supplied by WRRFs. The annual production of methane in EU WRRFs can reach about 9.5 billion m3 with a caloric value of 100 billion kWh. which can serve electricity to 25 million households. With an estimated EU production of 15 million t of sludge waste in WWTPs, resource recovery strategies could result in annual sludge waste reductions of 4 million t. Finally, the treatment of municipal wastewater amounts to 34 billion kWh/year in the EU, meaning 34 % of the recoverable energy from WRRFs. Thus, the implementation of new design patterns and operational strategies can allow WRRFs to be self-sufficient.
Based on this context, DARROW aims to build and demonstrate an innovative, optimised, modular, and flexible data-driven AI solution to make existing WWTPs more autonomous, more energy efficient and better prepared for their transformation into Water Resource Recovery Facilities (WRRFs). This AI solution will be tested in Tilburg WRRF. DARROW will take advantage of existing AI & Data analysis techniques to analyze, improve, particularize and demonstrate them with the final objective of contributing to a greener planet, thus maximizing contribution to the Green Deal.
During this period, while working in parallel on management (WP1), dissemination and communication (WP8) topics, we worked on the technical WPs (WP3, WP4 and WP5) and the integration WP (WP6). In addition, WP7 (demonstration) has just started. The work in the technical and integration WPs has been impacted due to two reasons beyond the control of the partners:
• The Tilburg WRRF has been rebuilt (finished December 2024) and the availability of stable and representative data has been limited.
• The Network and Information Systems 2 (NIS2) EU directive has classified WRRFs as critical infrastructures. This has created new needs in terms of cybersecurity for the DARROW digital solution.
This situation presented both a challenge and an opportunity for the partners, as many AI solutions for the water sector are expected to encounter similar issues when implemented in real-world facilities. To minimize the impact, several actions were taken, including monthly consortium meetings that helped align the partners' efforts towards achieving the objectives. These efforts focused on developing the necessary methodologies flexibly, ensuring that once stable operation data is available and cybersecurity is guaranteed, the solutions can be implemented seamlessly. Moreover, a great effort is being made to increase the available data, by installing new sensors and performing specific experimental campaigns.
In WP3, WP4 and WP5, the methodological approaches have been developed. Concerning WP3, several machine-learning techniques have been tested for generation of diverse type of data (e.g. inflow and laboratory information). In addition, an offline development of machine learning models is being carried out for anomaly detection on sensor measurements. A biomass observer has been developed, key to understand the process status in the secondary treatment of a WWTP. Work is also ongoing on the development of software sensors for N2O and NH4. In WP4, the use of reinforcement learning has been tested with simplified virtual versions of the Tilburg WRRF. Regarding WP5, first versions of reduced order models are available, which will help in the monitorization and decision-making in the WWTP. All the methodologies and algorithms developed in these WPs are prepared to be adapted to the Tilburg case study when the stable data is available.
The work in WP6 is focused on integrating the AI solutions in a single platform. A significant number of sensors have been already ingested and a unified API from which ingested data and AI model output can be retrieved has been created. Moreover, dashboard and the generalized user interface has been also developed. While the final version of most AI tools developed in WP3, WP4 and WP5 is still not available, integration tests are being done by onboarding preliminary versions of the models into a first end-to-end solution.
Although WP7 has just begun, efforts have already been made to anticipate the requirements for meeting NIS-2 standards, ensuring the feasibility of the DARROW solution. Additionally, Dommel is coordinating with the other partners how the training of the operators will be done, in order to facilitate their acceptance of the developed solutions.
• The Tilburg WRRF has been rebuilt (finished December 2024) and the availability of stable and representative data has been limited.
• The Network and Information Systems 2 (NIS2) EU directive has classified WRRFs as critical infrastructures. This has created new needs in terms of cybersecurity for the DARROW digital solution.
This situation presented both a challenge and an opportunity for the partners, as many AI solutions for the water sector are expected to encounter similar issues when implemented in real-world facilities. To minimize the impact, several actions were taken, including monthly consortium meetings that helped align the partners' efforts towards achieving the objectives. These efforts focused on developing the necessary methodologies flexibly, ensuring that once stable operation data is available and cybersecurity is guaranteed, the solutions can be implemented seamlessly. Moreover, a great effort is being made to increase the available data, by installing new sensors and performing specific experimental campaigns.
In WP3, WP4 and WP5, the methodological approaches have been developed. Concerning WP3, several machine-learning techniques have been tested for generation of diverse type of data (e.g. inflow and laboratory information). In addition, an offline development of machine learning models is being carried out for anomaly detection on sensor measurements. A biomass observer has been developed, key to understand the process status in the secondary treatment of a WWTP. Work is also ongoing on the development of software sensors for N2O and NH4. In WP4, the use of reinforcement learning has been tested with simplified virtual versions of the Tilburg WRRF. Regarding WP5, first versions of reduced order models are available, which will help in the monitorization and decision-making in the WWTP. All the methodologies and algorithms developed in these WPs are prepared to be adapted to the Tilburg case study when the stable data is available.
The work in WP6 is focused on integrating the AI solutions in a single platform. A significant number of sensors have been already ingested and a unified API from which ingested data and AI model output can be retrieved has been created. Moreover, dashboard and the generalized user interface has been also developed. While the final version of most AI tools developed in WP3, WP4 and WP5 is still not available, integration tests are being done by onboarding preliminary versions of the models into a first end-to-end solution.
Although WP7 has just begun, efforts have already been made to anticipate the requirements for meeting NIS-2 standards, ensuring the feasibility of the DARROW solution. Additionally, Dommel is coordinating with the other partners how the training of the operators will be done, in order to facilitate their acceptance of the developed solutions.
This period has been focused on the technical development of the AI solutions and their integration. While the final models are still not available, the partners have made several advancements, both in the methodological approaches and the development of tools. The obtained results are available in the technical progress-oriented sections.