Digital Twin & Asset Management

Developing methodologies to elevate the capability level of Digital Twins from 0-2 to 3-5 (autonomy).

Work Package 4 is led by Adil Rasheed from NTNU and Kjetil Johannessen from SINTEF Digital. Its focus areas are Digital twin for wind energy, optimal farm control and asset management.

Anticipated results:

  • High level of physical realism in DTs
  • Max. power production with min. asset degradation
  • Probabilistic predictive maintenance

Hypothesis

Digital Twin with high level of realism will be a paradigm shift for informed decision making

Methodology and research tasks

Digital Twin adapted for wind energy

Framework development, optimised sensor placement, novel modelling paradigm.

Optimal Park Control

Holistic Modelling, Combined optimisation for energy production, asset degradation and its integration into the power grid, evaluation of farm control strategies.

Asset Management

Condition Monitoring (CM), Predictive Maintenance and improved strategies for life extension.

Informed public engagement

Augmented reality will be employed to inform the public of future impacts (WP5.3) of the upcoming wind farm.

Innovations

Main results from 2022

Digital Twin adapted for wind energy

Data consolidation: The Bessaker wind farm and Zefyros wind turbine data have been acquired and pre-processed to build standalone, descriptive, and predictive digital twins, accurately representing operation, and providing insights into future performance. The following databases have been established.

  • Data collected from the Bessaker Wind farm, cleaned, and utilised in building a standalone and descriptive digital twin.
  • Data for the Zefyros case collected, cleaned, and utilised in building a standalone and descriptive digital twin.

 

Novel modelling paradigm: This work involved developing a novel modelling paradigm, hybrid analysis and modelling (HAM), that combines the strengths of both physics-based modelling and data-driven modelling. Two approaches have recently been developed, known as Non-Intrusive Reduced Order Models (NIROM) and Corrective Source Term Approach (COSTA). NIROM employs Long Short-Term Memory (LSTM) networks to accelerate computational fluid dynamics (CFD) simulations of turbulent flows, while COSTA models elasticity problems in the presence of incomplete physics, uncertainty in physical parameters, and simplifying assumptions. This work has resulted in the following manuscripts:

  • A paper titled: Hybrid deep-learning POD-based parametric reduced order model for flow around wind-turbine blade published in the Journal of Physics: Conference Series
  • A masters’ thesis titled “Corrective source term approach for improving physics-based models” was completed and a journal article based on the thesis is under review

 

Asset management

Predictive maintenance and decision support for wind energy applications: Offshore wind turbines (OWTs) are important for wind power generation due to their high electricity output and low land use. However, the harsh environment and remote locations in which they are installed mean maintenance is difficult, making predictive maintenance (PdM) a compelling strategy. PdM relies on failure prognostics, which predict an asset’s remaining useful life (RUL) based on condition monitoring (CM). This work presents a systematic review of failure prognostic models for wind turbines, categorising them into data-driven, model-based, and hybrid models. The findings suggest that developing hybrid models that combine the advantages of data-driven and model-based models is promising. Meanwhile, a separate investigation examines offshore wind turbine failures and proposes four hypotheses on failure features. The work is compiled in the form of the following papers:

  • Statistical analysis of offshore wind turbine failures, Submitted to ESREL 2023
  • A paper titled “A review of failure prognostics for predictive maintenance of offshore wind turbines” published in the Journal of Physics: Conference Series, 2022

Informed public engagement

Development of tools for visualising datasets: A generalised and extendable virtual reality-based digital twin framework has been developed to demonstrate its standalone, descriptive, and predictive capabilities in the context of an onshore wind farm and an offshore wind turbine. The work will appear in the proceedings of the following conferences.

  • Standalone, Descriptive, and Predictive Digital Twin of an Onshore Wind Farm in Complex Terrain, Deep Wind 2023
  • Demonstration of a standalone and descriptive digital twin of a floating offshore wind turbine, Accepted in OMAE 2023

 

Previous results

Digital Twin adapted for wind energy

Conceptualising Digital Twins for wind energy:

  • A technology watch on the topic of digital twins was conducted to establish the state-of-the-art in the technology. 15 industry partners provided detailed feedback which has been processed and sent back to them for their approval for circulation and publication.
  • Work on a perspective article titled “A digital twin framework for wind energy” is ongoing.

 

Novel modelling paradigm:

  • A conference paper titled “Machine-learning based non-intrusive parametric reduced-order model for flows around aerofoils and wind turbine blade” was submitted to DeepWind.
  • Ongoing work on a conference paper titled “Hybrid analysis and modelling as a digital twin enabler for wind energy”.

 

Farm flow modelling

Multiscale wind modelling:

  • A multiscale wind prediction model has been set up for the Bessaker Wind Farm.
  • Work on a paper titled “Reinforcement learning for stabilising IEA 15MW offshore wind turbine under varying wind conditions” is underway.
  • Nord2000 noise models applied to Bessaker use case for idealised wind conditions.
  • Reduced Order Modelling of wakes using a hybrid approach.
  • A journal article titled “A hybrid POD approach for the solution of transient turbulent flow problems” is ready for submission.

 

Asset management

A novel Corrective Source Term Approach (CoSTA) was developed as a HAM tool for structural health monitoring.

  • First draft of a journal article titled “Physics Guided Neural Network-assisted Corrective Source Term Approach to Hybrid Analysis and Modelling” is ready for submission.
  • Predictive maintenance and decision support for wind energy applications.
  • A conference paper titled “A review of fault prognostic models for predictive maintenance of offshore wind turbines” was submitted to Deep Wind.

 

Informed public engagement

Development of tools for visualising datasets:

  • A first version of the Augmented Reality tool was demonstrated.

 

Picture showing a room with an augmented reality overlay and two hands also overlayed.
Holo Lens application for augmented reality inspection of wind farms