DigiWind – The digital twin of a wind turbine
The big effort of transforming our entire energy system puts many challenges upon the wind energy industry. Large quantities of power generating units have to be built, operated and maintained. This can be facilitated by implementing a digital twin. However, wind energy systems pose a unique combination of challenges due to their long operating time, the large number of loosely coupled units, and the modularity of the system architecture. These aspects are not covered by general approaches to the development of a digital twin but require special methods.
The DigiWind research project is developing an open-source digital twin platform specifically tailored to the wind energy sector. The DigiWind framework builds on the modular, ever changing system structure of wind energy systems and translates it into a similarly evolving and adapting digital representation. Users of the DigiWind platform are enabled to define services for their assets and link them to already existing business logic and processes.
For the optimal and flexible operation of wind turbines, it is important to know the remaining useful life of the individual components of a wind turbine. The evaluation of the lifetime requires the regular estimation of the condition of components to make maintenance decisions. To ensure reliable results, the models used for remaining useful lifetime calculations must always match the components in the real system. If a part is replaced during maintenance, the corresponding model needs to be adjusted. To accommodate for changes in the wind turbine, we use the model management and assembly service on the digital twin platform. In this way, the remaining service life - service always has access to the latest model of the wind turbine.
Furthermore, the remaining useful lifetime can be used in an additional service for optimal operation planning with the help of the modular concept of the DigiWind - platform. We at the industrial energy systems research group are investigating how reinforcement learning algorithms can be used to determine the optimal operating strategy of a wind farm. This involves incorporating the provided estimate of remaining useful life with additional predictions of available wind resources and estimates of electricity prices. Using the digital twin approach, this optimal operating strategy can be directly fed back into the wind energy system.
The DigiWind platform, opens an external URL in a new window is available as an open source GitHub repository (DOI: 10.5281/zenodo.7994327).