PO134

Hitting target reliability levels by incorporating fundamental physics to improve predictions

Mark Spring 1, Marco Sepúlveda2, Peter Davies1, Jeff Bonnell1
1Lloyd's Register, London, UK, 2IDCORE, Edinburgh, UK

Abstract

At 30% of levelised cost of energy (LCoE), operations and maintenance activities constitute a central focus for the ambitious commitments for wind with regards reaching price parity with other forms of electricity production.

Currently maintenance tasks which are pre-planned and scheduled in advance comprise just 60% of hours worked on turbines. There is ample scope for substituting predictive for reactive repairs and component replacements.

With over 10 years' experience using methods based on reliability targets to prioritise maintenance tasks, the team at Lloyd's Register (LR) has chosen operation of offshore wind farms as the focus for application of proven statistical models, now incorporating cutting-edge physics-based models of key failure modes specific to wind turbines, identified previously using a rigorous failure modes and effects analysis (FMEA). By quantifying all components of risk, FMEA has identified the most vulnerable systems and has ranked the following systems as priorities, in agreement with industry experience, verified through close collaboration with maintenance teams at operating wind farms: frequency converters, gearbox, pitch system and yaw system. FMEA has been undertaken on radically different configurations of wind turbines. This paper reports results from wind farm pilot projects, using LR software Axxim, fully embedded within CMMS and incorporating new algorithms from the author's doctorate research. Comparisons are made using key performance indicators arising from models of logistics and costs. The benefits are demonstrated arising from these early pilots and further investment in more comprehensive models is justified.

Method

Everything known about failure mechanisms, consequences and detection has been embedded within computerised maintenance management systems (CMMS). Real-time visualisation of the risk profiles associated with failure modes and maintenance tasks enables degradation and failures to be predicted more accurately. This supports rational decision regarding maintenance tasks, which are scheduled in advance of any failures in a cost-effective manner. The objective is to reduce as far as possible system downtime, expensive logistical support and spare parts inventories, whilst simultaneously maximising component utilisation and duration of service.

Results

Combining published, proprietary and project-specific models of costs associated with wind farm operation has enabled quantitative performance comparisons between business as usual and new predictive, reliability-based maintenance task-planning.

Conclusions

Cost savings across areas comprising operation and maintenance of offshore wind farms have been estimated based on the results of the pilot projects described. These have been equated to reductions anticipated in LCoE resulting from full implementation of the techniques described across all constituent assets. New areas of investigation have been proposed in order to develop physics-based models of degradation and failure for aspects not already covered in the pilot.

Objectives

Combining statistical and physics-based models enables more accurate risk-profiles in real time.  Maintenance tasks can be scheduled intelligently based on predicted changes in these risks.  Displaying these clearly to wind farm operators and fully embedding within CMMS allows costs associated with downtime, maintenance logistics and spare parts inventories to be reduced.  Models of project costs have been included to ensure consistent comparison of different maintenance strategies.