AI-Based Predictive Maintenance and Its Impact on the Efficiency of Wind Turbines
During the lifetime of a wind turbine, operation and maintenance can account for up to 30% of the levelized costs per kWh generated. Turbine failure usually results in downtime of a week or more – resulting in respective downtime costs. Transporting spare parts is expensive as turbine fields are commonly located in remote areas, resulting in more time in transportation than the repair itself. Estimates suggest that unplanned repairs account for over 70% of wind turbine generator (WTG) downtime. This is where AI and IoT could come into play to reduce logistics and maintenance costs.
European Union’s Horizon 2020 grant
Horizon 2020, the EU’s research and innovation funding program, aims to develop advanced technological solutions that help reduce the operation and maintenance costs of offshore wind power facilities. The ROMEO project, led by IBERDROLA RENOVABLES ENERGÍA, includes twelve experienced market players from six different EU member states and one associated country, such as IBM Zurich, Siemens, Adwen, and ElectricitéDe France. It is backed by the European Union’s Horizon 2020 grant. These companies are working jointly on an IoT predictive wind turbine maintenance system for 3 offshore farms: Teesside (United Kingdom), East Anglia ONE (United Kingdom), and Wikinger (Germany).
ReaLCoE is another project supported by Horizon 2020. It plans to integrate predictive and demand driven maintenance with sensor concepts, allowing evidence and need based maintenance of critical components of wind turbines. ReaLCoE will also implement a Digital Twin and Condition Monitoring System. This will improve the knowledge of operational data and increase the overall reliability of the turbine over the entire lifecycle by improving the adaptation of the controller. By enhancing the operational capacity of the regular offshore WECs to 14-16 MW, an electricity price of €35-50/MWh could be achieved, which is just a third of the baseline price of similar projects.
ONYX InSight – Wind Turbines
Romax Technology and Castrol jointly founded ONYX InSight on December 5, 2016 as a joint venture. It has partnered with 8.2 Monitoring GmbH to provide advanced sensing technologies, plant monitoring, and component life extension programs to the wind industry in Germany. They are combining multiple data streams collected via ecoCMS Advanced Sensing Technology to improve diagnostic accuracy. The ecoCMS hardware and fleetMONITOR, and ONYX InSight’s monitoring software, analyzes the turbine performance and health data.
Électricité de France S.A., known as EDF, founded on April 8, 1946, in Paris, France, is a French multinational electric utility company largely owned by the French state. ONYX InSight and EDF Renewables North America are partnering to deliver lower energy costs on more than 1,500 wind turbines. The cooperation aims to reduce turbine operational expenditure for EDFR and increase asset availability. EDFR integrates siloed data streams from multiple turbines fitted with different CMS hardware and other relevant data stream inputs. ONYX will combine these data streams into fleetMONITOR.
Berlin-based Turbit Systems, founded in 2017, develops condition monitoring software for wind turbines. The software compares real-time turbine-generated SCADA data with expected behavior to detect technical faults and recommend corrective measures. This insight will help wind farms improve their energy output by up to 5%.
The monitoring software comes with two settings. In general alarm settings, Turbit Detection KPIs measure the difference between simulated data from neural networks and actual data from the wind turbine. The second type of alarm settings for Turbit Modules, has shorter periods (30 Minutes) of detection KPIs for Power Monitoring analysis. Detection KPIs for the main components observe a longer time frame (5-10 days), which ensures high alarm relevance. Each automatic alert comes with an automatically generated report called an Event Card. The Event Card displays relevant plots, overlapping anomaly status codes, and nearby turbine benchmarks.
Other research activities
Siemens Gamesa published a case study showcasing their ability to monitor 10,000 remotely connected wind turbines globally. Every turbine transmits data from its 300 sensors to the Siemens Gamesa Remote Diagnostic Center for further analysis. In addition, the company uses digital twinning technology to improve spare parts forecasting with early failure prediction and enhance its asset management. The digital twins are based on historical vibration data collected since 2004 and real-time performance data. By using neural networks, the system compares the inputs and locates early damage indication patterns. Their algorithms can remotely detect 99% of drive-train damage, such as gear-tooth cracks or main bearing damage.
IBM Zurich, Switzerland, one of IBM’s 12 global research labs, has proposed a fault prediction and diagnosis solution for wind turbine generators using SCADA data.
The proposed solution has deployed and evaluated in two wind power plants in China. The experimental study demonstrates that the generators’ Remaining Useful Life (RUL) can be predicted 18 days ahead with about 80% prediction accuracy. When a fault occurs, we can diagnose the specific type of generator fault with an accuracy of about 94%.
Market studies predict that the global offshore wind market will scale to USD 56.8 billion by 2026, which is almost twice the amount of 2021. The International Renewable Energy Agency’s (IRENA’s) 1.5 °C scenario foresees an increase from 34 GW today to 380 GW by 2030 and more than 2,000 GW by 2050.
O&M costs are among the key barriers hindering wind energy production on a large scale. Estimates indicate that the median O&M cost for a US wind farm ranges between $42,000 and $48,000 per MW during the first 10 years of operations. However, with aging wind turbines, the O&M costs ramp up too. Predictions indicate that between 2016 and 2026, US operators will spend over $400 billion on maintenance activities.
Studies show that operators with better wind turbine data analytics save up to 16% on their OPEX costs. Connected sensors and predictive analytics algorithms have emerged as strong contenders to maximize the reliability and performance of wind turbines while reducing O&M costs.