Sensor Screening Methodology for Virtually Sensing Transmission Input Loads of a Wind Turbine Using Machine Learning...

Area of Science:

• Mechanical Engineering
• Renewable Energy Systems
• Condition Monitoring

Background:

• Increasing wind turbine (WT) size drives higher drivetrain loads, necessitating gearbox load monitoring.
• Direct load measurement is costly; virtual sensing using stationary sensors offers an economical alternative.
• Optimizing sensor number and placement is crucial for cost-effective virtual sensing solutions.

Purpose of the Study:

• To identify optimal sensor locations for virtual sensing of WT 6-degree of freedom (6-DOF) transmission input loads.
• To reduce the cost of virtual sensing systems by prioritizing essential sensor placements.
• To evaluate the impact of sensor selection on the accuracy of load prediction.

Main Methods:

• Utilized random forest (RF) models applied to simulated operational data from a Vestas V52 WT multibody model.
• Analyzed 6-DOF transmission input loads alongside signals from potential sensor locations (deformations, misalignments, rotational speeds).
• Employed a statistical test to rank sensor locations based on their impact on virtual load sensing accuracy.

Main Results:

• RF models successfully identified sensor locations with the highest influence on virtual load sensing accuracy.
• Model performance was assessed before and after reducing the number of input signals, demonstrating robustness.
• The study prioritized and reduced the number of necessary input signals for effective load monitoring.

Conclusions:

• The proposed method shows high promise for optimizing the cost of future virtual WT transmission load sensors.
• Screening sensor locations prior to real-world implementation can significantly improve cost-efficiency.
• Virtual sensing offers a viable and economical approach to monitoring intensified WT gearbox loads.