‘Fundamentally, wind turbine performance is hard to assess’
OPINION | The potential is huge for automation and AI to drive up wind farm performance and drive down technical risk – starting with the rotor blades, writes Gareth Brown
The wind industry has grown incredibly quickly in a short period of time. From small turbines on Denmark’s hillsides in the 1980s to the machines of today installed on every continent on the planet, with large offshore wind turbines now passing the 10MW threshold.
This recent progress has been driven by the ever-decreasing cost of energy, with subsidy free onshore wind winning competitive bids in open markets and offshore wind making great strides. Yet even with this growth there are still key challenges to ensure the assets perform to their potential.
Fundamentally, wind turbine performance is difficult to assess as it is the only power source where the inflow conditions are not well defined, and spread over tens if not hundreds of turbines all with varying inflow and operating states. The lack of understanding of inflow conditions means large variations of normal production performance hide underlying systemic issues from the owner’s view.
The issue is compounded by an industry obsession with availability rather than performance, the application of broad power industry SCADA data structure not suited for wind farm performance management, and turbine design standards that have not kept pace with the growth.
Leading edge erosion of turbine blades has been an increasingly critical concern for the industry and one that has been difficult to address because of these aforementioned issues. In particular, the rapid growth of our sector means design standards have not kept pace. Wind turbine standards were developed over time and rely on information derived from smaller turbines manufactured earlier in the industry. Learning from these early-stage small turbines do not necessarily translate well to ultra-large modern turbines and the different environments they are deployed in.
Wind turbines are now designed with less margin for error as we have better tools to design with and can use lighter, more advanced materials.
The problem we face is that the underlying assumptions of environmental conditions that drive loading are outdated. So, where a 50-metre-diameter rotor had a lot of room in the design for error – such that after 20 years of operational life you might be able to refurbish the blades, put in a new gearbox and update the controller to potentially enable the turbine to run for another 15 years – can the same be said for a new 100-metre-diameter rotor built today which has been designed right to the edge?
Many turbine blades are suffering from early life leading-edge degradation as they are now designed to be much lighter ,to enable larger sweep areas with minimum room for error in the manufacturing process or in the operational environment. Tip speeds on modern wind turbines have also increased as the rotor size increases leading to wind turbine design standards not accurately considering the impact loads of rain droplets on leading edge erosion at these higher tip speeds.
The good news is the issue has already been solved in aviation and our industry has started to take note, with some designs implementing the learnings in the manufacturing process.
For new and existing assets, the impact of leading edge erosion can be monitored and measured through visual inspection and/or through data science . The hard part of detecting leading edge in an automated manner is creating a statistically significant result from noisy data when the inflow conditions are not easy to infer. However, we have seen some fantastic results in the industry as the latest data science, machine learning and more broadly AI are being applied with the latest domain expertise.
"The potential is huge for automation and AI to get the information to the industry's fingertips."
It means wind farm data streams can be put in context of the turbine’s operational condition along with inflow conditions and what drives them such as atmospheric stability, wake effects and forestry, terrain and bodies of water.
If owners take advantage of more advanced data architecture and enrichment then the new forms of analytics, machine learning and AI, have the potential to capture leading-edge erosion issues early. That leads to not only better-supported warranty claims with manufacturers, where erosion caused by poor design may be more difficult to break apart from general wear and tear if the claim is made at the end of a service or warranty period, but also stronger business case evidence to support undertaking the expensive repairs and retrofits.
The potential is huge for automation and AI to get the information to the industry's fingertips, driving up wind farm performance and managing technical risk.
Gareth Brown is CEO of Clir Renewables, an AI software company developing cloud-based tools for wind farm asset management