F1 - Analysis of remote sensing and condition monitoring data
F - Operations, Management, Maintenance and Safety
As the use of remote sensing and remote condition monitoring through digital twin technology becomes more commonplace, improved techniques for analysis of the data using AI and machine learning will be needed.
New analysis tools are needed to be developed, demonstrated and benchmarked for analysis of the data generated from remote sensing and remote monitoring using autonomous systems.
Context And Need
Effective and efficient prognostic condition monitoring techniques are needed in order to utilise the state of the art computational capability in its full breadth for the ORE sector.
The size and quantity of data from ORE assets steadily increases. In order to make best and efficient use of this information, tools and standardised processes are needed to handle and interpret these important data sets.
The increasing quantity and value of ORE assets requires and Improve understanding and estimation of O&M activities/cost/scheduling. Risk-based approaches can help to estimate the likelihood and consequence.
To minimise offshore operations and inform O&M strategies, remote sensing and remote condition monitoring through digital twin technology are important developments that have the potential to enable remote resets and repairs and significantly reduce the cost of O&M offshore. Improved analysis tools and treatment of big data generated by remote sensing, proven by smarter benchmarking will be needed to unlock the potential for use of AI and machine learning and present opportunities for advanced asset monitoring and management, potentially lowering OPEX cost, whilst increasing availabilities.
These techniques have the potential to reduce OPEX whilst maintaining acceptable safety levels
- Reduced OPEX (short-term)
- Reduced CAPEX (long-term)
If O&M uncertainties can be reduced the overall OPEX cost and production losses can be reduced.
Digital twins used in other industries
AI/Machine learning established in computational science, but emerging as an offshore application
Isolated initiatives to collect and generate benchmark data, e.g. ORE Catapult SPARTA project for reliability
Several O&M models in use, many are deterministic / empirical and would benefit from probabilistic methods.
Active research projects:
- HOME-Offshore: Holistic Operation and Maintenance for Energy from Offshore Wind Farms (EPSRC, EP/P009743/1) - This project investigates the use of predictive modelling, robotics, advanced sensors and big data techniques to target interventions and thus improve safety, and reduce the cost, of the operation and maintenance of offshore wind farms. This will also help address the increasing shortage of skilled workers in this field.
Supergen ORE Hub - Flexible Funding Research
- V-SCORES (Validating Surface Currents at Offshore Renewable Energy Sites)
Lead Institution: University of the Highlands and Islands
Marine current measurements are vital for tidal resource estimation and resilient design criteria for all Offshore Renewable Energies (ORE). In-situ measurements are costly, and retrieval of seabed mounted equipment is not guaranteed. Moreover, in many potential ORE locations globally, suitable field survey campaigns may not be viable. Until now, most data for model validation and impact assessment have focused on temporal variability from single-point measurements, yet spatial variability is of critical importance. Additionally, most oceanographic current measurements are sub-surface; the near-surface zone is largely unknown due to instrument limitations (e.g., surface interference making the top few “bins” of ADCP data unusable). The development of low-cost and low-risk surface current mapping tools, and translating this knowledge to flow at depth, is therefore a key challenge in ORE development. Surface current maps would provide high-resolution detail needed to measure spatial heterogeneity, understand realworld wakes and the relationship between flow and animal behaviour when combined with ecological surveys. A better understanding of surface currents will also improve resilience of floating ORE and yield of floating tidal turbines. The aim of V-SCORES is comprehensive validation of unmanned aerial vehicle (UAV) techniques for surface current spatial mapping, demonstrated at tidal stream sites. Field campaigns will be conducted at contrasting commercial sites (Pentland Firth, Scotland & Ramsey Sound, Wales) under different environmental conditions (wave exposure, operational turbines installed, etc.).
- Demonstrating a machine learning system to integrate metocean data, sensor networks, and model output for improved coverage and accuracy:
Lead Institution: University of Exeter
Machine Learning for Low-Cost Offshore Modelling (MaLCOM) will develop a modelling methodology to provide rapid, accurate nowcasts and forecasts of the wave conditions at a regional scale using limited input data and requiring drastically reduced computational power. MaLCOM will use the historical outputs of a physics-based model and in-situ measurements to build a statistical representation, termed a surrogate model, between measurements and modelled conditions throughout a region using a machine learning approach. The developed surrogate model will provide two key benefits: (1) immediate spatial assessment of conditions with very little computational power required, such that it could be deployed on a mobile device or autonomous vessel and (2) improved accuracy of metocean forecasts through integrating in-situ measurements. This project will allow refinement and demonstration of this resource modelling and forecasting concept for marine energy sites.
- A hybrid and scalable digital twin for intelligent direct drive powertrain condition monitoring
Lead Institution: University of Strathclyde
As larger wind turbines with newer powertrain technologies are introduced in the offshore wind sector, state-of-the-art machine learning techniques that use past field data are no longer directly applicable. Operational alarms based on physical models of older turbines are often no longer valid with new powertrain technology. This represents a key vulnerability in the offshore wind sector. This project will develop a hybrid digital twin combining transfer learning and physical modelling approaches that will be able to model normal and abnormal behaviour for new turbines before operational data is available. As turbines move further offshore, operators are motivated to reduce the number of turbine visits for cost and safety reasons. The hybrid models proposed in this application could be used to reduce the number of powertrain inspection and service visits. The requirement for visits will be reduced through the digital twin providing additional health indicators and recommendations to the operators, and by adding confidence to the use of existing health indicators provided by SCADA and monitoring systems
Links to Industry Priorities:
- Also Offshore Wind Innovation Hub - O&M and Windfarm Lifecycle innovation priorities
- Offshore Wind Innovation Hub Roadmap Data: Machine learning deep learning from big data
We would also like to invite UK researchers and industry stakeholders within ORE to submit links to research projects, both past and present, for inclusion within the landscape.
Therefore, if you have a UK-based research project within an area of ORE that you feel is relevant to a specific research theme or challenge within the Research Landscape, click HERE to submit your research project to the research landscape
PhD projects in Offshore Renewable Energy
In order to better understand the breadth of ORE research currently being conducted in the UK, the Supergen ORE Hub has collated from its academic network, UK Centres for Doctoral Training and Industrial partners, a list of PhDs currently being undertaken in ORE.