TETRA

Challenge

Artificial intelligence (AI) is a promising, dynamic technology with many applications. Constant and rapid improvements in algorithms, coupled with the increasing availability of data, are broadening the scope and effectiveness of AI applications. The ability to adopt and use AI technologies efficiently is crucial for maintaining and expanding German and French innovation and competitiveness in almost all areas. As different sectors have different requirements for AI solutions, TETRA aims to provide tools and methods for successfully implementing AI projects in specific areas, such as water management (WM).

During the project, the TETRA partners will develop tools and methods to be made publicly available to German and French stakeholders, enabling collaboration through a shared approach to AI projects. To ensure the tools and methods are applicable, they will be evaluated using two use cases: monitoring river water quality and river restoration. By creating a framework for integrating and executing AI algorithms and standardising data storage, TETRA will establish a foundation for a shared ecosystem of AI-based water projects. German and French SMEs will be able to use the developed tools to reduce costs and implementation time. This ecosystem will serve as an enabler and accelerator for AI projects focusing on water data.

Once the project is complete, participants in this ecosystem will benefit from the TETRA methodology, which provides information on integrating AI algorithms into a runtime environment and executing an AI-based project.

Motivation:

Water is an essential and immensely valuable resource for humans and the environment. To ensure its long-term availability, we need modern tools to monitor water efficiently and reliably. This is evident in the recent fish deaths in the Oder River, for example. To improve river protection, the TETRA consortium is proposing a project blueprint to simplify and accelerate the use of artificial intelligence (AI) in water management. The tools provided will reduce the time and cost required for AI solutions. Authorities and SMEs will be supported in maintaining the resilience and health of rivers. TETRA is developing an AI toolbox supported by a suitable methodology. The tools will be evaluated using two use cases on the German-French border of the Rhine: qualitative river monitoring and renaturation measures. River monitoring will be carried out using image and sensor data. Additional research is being conducted into how AI algorithms can run on edge AI devices. In the area of renaturation measures, AI is being used to promote sustainability. The accompanying methodology will be based on the existing industry model for AI engineering, PAISE. Where necessary, PAISE will be adapted to align with the toolbox developed and provide a template for subsequent projects.

The method and basic tools developed will be freely available, forming the basis for a European AI ecosystem. These results can be utilised by German and French authorities, as well as SMEs, thereby opening up opportunities for cross-border collaboration in the field of AI.

Solution

The development of multivariate anomaly detection methods based on convolutional neural networks (CNN) and long short-term memory (LSTM) networks for conspicuous events in time series of water quality parameters using artificial intelligence (AI) methods has already begun. This means that correlations between different parameters can be taken into account. However, the interactions between water quality parameters are complex. This knowledge will be modelled in an ontology and used to improve event detection in water quality algorithms. The current technology readiness level (TRL) is 3 and will reach TRL 5 by the end of the project through the integration of water expert knowledge using semantic search in the water quality domain ontology, thereby improving the results achieved so far.