"VENTUS"

In the “VENTUS” project, we apply current research in the fields of physics-informed AI and probabilistic-causal AI to drastically optimize the operation and maintenance of wind turbines. Based on an analysis of failure cases and performance degradation conducted in collaboration with relevant stakeholders, we will strive to develop an explainable AI system that has the potential to reduce losses due to downtime and maintenance by 50%.

MORE PROJEKT DETAILS

SUSTAINABILITY

Key Facts

Program: FFG (Austrian Research Promotion Agency)
"Digital Technologies, AI for Green 2023"

Duration: 01.09.2024 - 31.08.2027 (36 months)

Project stakeholders and roles:

University of Technology Vienna - coordination
University of Technology Graz - AI development
DiLT Analytics FlexCo - AI development

Associated Partners

user view & industrial context with LOI partners
knowledge transfer with international research institutions

Updates about the project can be found on our LinkedIn channels!!

About the project

GOALS

The challenge

The transition to renewable energy sources is one of the biggest challenges facing Europe (European Environmental Agency, 2019). The energy sector accounts for 78% of total greenhouse gas emissions within the EU and 73.2% worldwide (European Commission, 2019). The EU is therefore aiming to reduce emissions by 55% by 2030 and achieve climate neutrality by 2050. Of all renewable energy sources, wind energy currently has the largest share in the EU (37.5%) (EuroStat, 2024). The expected growth in both onshore and offshore wind energy is enormous:

The EU's “Fit for 55” package increases the target for renewable energies from 32% to 40% by 2030, requiring an increase in installed wind power capacity from 180 GW (2021) to 451 GW (2030) (WindEurope, 2021).
Austria's target for the expansion of wind energy by 2030 (10 TWh share of the electricity generation mix) requires a newly installed wind power capacity of 500 MW/year (Wien Energie, 2022).

Share of electrical energy generated by wind in Europe

Number and geographical distribution of wind turbines in Europe

The project content

VENTUS focuses on the development of probabilistic and physically based ML frameworks tailored to wind energy systems, providing improved predictions with uncertainty quantification and causal relationships. These are then applied in European, industry-relevant scenarios of wind turbine operation (using real data from our project partners), in particular the predictive maintenance of mechanical components. Finally, contributions to the open data community are integrated, in particular the publication of codes and selected data sets.

The innovation

VENTUS builds on the latest developments in the field of physics-based and probabilistic/causal machine learning methods. While progress has been made in both areas in recent years, the combination of both approaches represents a new direction in research. This enables:

... the extension of traditional learning-based techniques with (physical, causal) background knowledge.
... dramatically higher data efficiency and transferability to new scenarios.
... a significantly higher degree of explainability than conventional AI systems.

This innovation in the form of an explainable AI system has the potential to reduce losses due to downtime and maintenance in wind turbines by 50%.

The project goals at a glance

Main objective: Explainable AI for automated fault detection and predictive maintenance (TRL 3-4)

Goal areas

#1 - “Development of theoretical foundations for combining physics-based ML and probabilistic/causal ML”
#2 - “Development of an ML-based model for automated diagnosis and predictive maintenance of wind turbines” (target improvements: 10% better prediction accuracy, 10x less training data, ...)
#3 - “Scientific publications (publications/presentations in high-ranking journals & conferences)”
#4 - “Open-source contributions (open-source repository for models and datasets)”

Sustainability aspects of the project

VENTUS is directly involved in European and national efforts to transition to renewable energy sources and thus addresses the following SDGs: “7 – Affordable and Clean Energy,” “8 – Decent Work and Economic Growth,” and “9 – Industry, Innovation, and Infrastructure.”

In the context of knowledge transfer, explainability, and security of our AI application, we also address the following three SDGs: SDG 4: In our dissemination activities, we focus on knowledge transfer; SDG 5: No different usage patterns between women and men; SDG 12: Trustworthy ML applications can be better managed in the long term.
Resource savings in relation to climate targets:

At VENTUS, we are developing an explainable AI system with the aim of reducing losses due to downtime and maintenance work by 50%, thereby directly addressing the EU target of increasing wind power capacity from 180 GW in 2021 to 451 GW by 2030.
The AI system we have developed uses physically based and causal techniques that are significantly more data-efficient than common deep learning approaches, resulting in a low environmental footprint.
In addition, the methods are developed at the academic level and in the SME sector, consuming far fewer resources than AI systems used by large companies. Therefore, the potential environmental impact of AI resources and their efficiency can be rated as excellent.
Possible rebound effects, such as job losses, are rather limited, as the AI developed will mainly serve as a decision-making aid and will not operate autonomously.