Navigating the future: AI-enabled digital twins for cost-effective offshore wind farms

Artificial intelligence (AI) is rapidly evolving, with a growing number of applications in the offshore wind industry. AI is currently employed in wind resource assessment, environmental impact assessment and predictive maintenance. Looking ahead, AI-enabled digital twins represent a promising future application that has the potential to significantly reduce developer costs.  

A digital twin is a numerical representation of a physical asset. As a virtual replica of an actual system, a digital twin can be used for anomaly detection, condition monitoring, performance optimisation, predictive maintenance and design improvement. 2H, Acteon’s Engineering consultancy, has developed and implemented structural digital twins for offshore energy projects. 

Why should developers opt for a digital twin

A digital twin tracks loads and stresses at critical structural hot spots on a real-time basis using the least number of physical sensors above water. This is done by designing a fully coupled finite element model to generate input and output data for training a digital twin. A digital twin can be used for replicating different types of offshore wind structures such as foundations, towers, wind turbine generators, cables and moorings.   

A digital twin helps in learning about the past and predicting the future, identifying anomalies, extending the life of an asset and optimising subsea inspection intervals. The data collected facilitates actionable insights, reduces the decision-making period for the developer, improves approval rates from government agencies and brings down the overall capital and operating expenditure of a project. It also helps in enhancing the safety of the personnel by reducing the number of required offshore trips. 

The challenge: long-term structural monitoring of a wind farm

Long-term structural monitoring of a large wind farm poses many hurdles to the developers.   

• High number of turbines: In a large wind farm, each turbine experiences varying operational conditions. Managing fatigue accumulation across numerous turbines becomes complex.
• Economic feasibility: Outfitting every turbine with monitoring equipment is cost-prohibitive. Balancing accurate data collection with financial constraints is crucial.
• Costly sensors: Placing sensors below water for underwater turbines is expensive.

The solution: digital twin monitoring

• Benchmark turbine: One benchmark turbine is selected within the wind farm and heavily equipped with inclinometers, accelerometers and strain gauges. This benchmark turbine serves as the reference for system training and design validation. 
• Machine learning training: The digital twin model is trained using simulated data and machine learning. Simulated values are compared with real-world data, allowing the model to improve over time. 
• Virtual sensors: Minimal sensors are deployed on other turbines to reduce costs. Instead, virtual sensors, which extrapolate data based on the benchmark turbine’s behavior, are developed to gather data. A digital twin is useful to estimate the structure’s response at unmeasured locations where instrumentation is impractical. 
• Platform for data visualisation (iCUE): Operators access historical data, key performance indicators (KPIs), trends, and forecasts through the iCUE visualisation platform. iCUE is currently utilised by offshore oil and gas operators and Siemens Gamesa for the Coastal Virginia Offshore Wind fixed wind development.  

The solution: digital twin monitoring

In the UK, 2H has been awarded two projects, including the development of a digital twin on the TetraSpar wind demonstration project. See our case study here. 2H plans to further develop and test the digital twin for real-time monitoring of floating wind.  

The digital twin model will create a virtual replica of the turbine and its components and will enable remote monitoring of performance, early detection of faults, and predictive maintenance planning for the substructure components, including foundation, tower and mooring lines. Machine learning technology coupled with a finite element analysis model will be used to build a digital replica of an existing floating wind platform with embedded pre-trained algorithms. 

Conclusion 

A successful digital twin combines optimised sensor placement and advanced analytical tools. It serves multiple purposes, including initial design validation, integrity management, fitness-for-service assessment and remaining life estimation. By reducing instrumentation costs, optimising inspection intervals, and minimising unplanned downtime, digital twins enhance asset value and extend operational life. While they will not replace inspections entirely, digital twins complement them, allowing asset managers to strategically space out maintenance activities.