Wind energy is one of the fastest growing sub-segments in the renewable energy industry today. An International Renewable Energy Agency (IRENA) analysis suggests that wind power saw a 17% rise in 2021, and significant investments in wind energy are under way as industries and governments pursue NetZero targets.

While rapid growth is certain, wind turbine operation and maintenance remains a major cost center in wind energy generation. These costs are even higher for offshore wind farms – constituting up to 16-25% of the cost of electricity. What’s more, wind turbine maintenance can pose severe risks for maintenance personnel, in addition to causing significant output loss due to downtime.

All these factors call for an advanced approach to wind turbine maintenance. Take a look at wind turbine maintenance challenges, and how predictive maintenance can help overcome them while lowering costs, downtime, and safety hazards for technicians.

A closer look at wind turbine failures

Wind turbines are a sophisticated system of electromechanical components that are usually installed hundreds of feet above the ground or surface of the ocean. These systems are composed of gear-type or direct drivetrains, high and low-speed shafts, yaw systems, mechanical brakes, cooling systems, generators, rotor blades, and other electronic components. 

Wind turbine failure modes

According to an IRENA study, gearboxes, electric systems, and blades and pitch systems account for the highest downtime per failure annually, whereas hydraulics, shafts, and drivetrains are associated with less frequent failures. Electric and control systems fail the most frequently per turbine per year

While rotor blades are not prone to failures, they suffer erosion and damage from lightning strikes, and as a result, detecting wear over time is crucial. Moreover, because wind patterns are stochastic, diagnosing faults in the gearbox can prove challenging. 

Challenges in detecting root cause of turbine failure

Most wind turbines in use today are equipped with provisional sensors and controllers that route data to SCADA systems. These systems do not offer fault detection and failure mode analysis capabilities. As a result, each failure necessitates dispatch of a team of technicians, sometimes via helicopters, to detect the cause of failure, followed by potential round trips for severe failures. 

As a result, reactive maintenance strategies increase the cost of repair, and add to the downtime of wind turbines. Similarly, a preventive maintenance strategy increases maintenance costs due to over-maintenance

This is where predictive maintenance can prove invaluable to wind farm maintenance operations.

Optimizing wind turbine maintenance with predictive analytics

Predictive maintenance strategies leverage sensor data to monitor the condition of an equipment in real-time. They enable operators to undertake an intelligent, data-driven approach to wind turbine maintenance. Here’s how novel predictive strategies are helping detect major failure modes of wind turbines.

Power curve anomaly detection to pinpoint failure modes

Wind turbine manufacturers will typically provide curves that indicate the power generated by a wind turbine against a certain wind speed. Based on these, it is possible to model real-world power curves for wind turbines, which are then combined with existing SCADA data points like temperature, voltage, etc. to train ML models for detecting fault/no-fault operation. The model is also trained to classify failure modes for the various subsystems of the wind turbine, at least one hour in advance. This provides insight into the exact cause of failure and enables technicians to resolve failures with precision.

Anticipating gearbox failure with multi-class neural networks

Load zone reverals, transient loads, and uneven load sharing due to suboptimal bearing settings are some of the most common causes of gearbox failure. To predict gearbox failures, both SCADA and vibration data can be used – and neural networks prove the most effective for predicting remaining useful life (RUL). However, vibration data can provide an accuracy of up to 100% with multi-class neural networks, and offer a lead time of 5-6 months.

Detecting rotor blade erosion with computer vision

Leading edge erosion of wind turbine blades is known to drop the energy yield of turbines by up to 4%. When they are deployed in harsh environments like deserts and oceans, wind turbine blades will typically wear out faster, owing to erosion due to harsh substances. To this end, manual inspection is conducted to study erosion on the blades. Instead, burst photography and computer vision models can be leveraged to study defect location along the length of the blade. This data is then classified to identify severe wear, which will impact yield, and cosmetic wear and spotting, which has minimal to no impact on power output.

Bearings wear analysis with NNs and SVMs

Vibration due to interaction of wind-turbine subsystems is the leading cause of bearing wear. However, other causes include oil supply issues, overload, overheating, and dust. There are multiple techniques to detect bearing wear in wind turbines – for example, vibration analysis  with neural networks (NNs) or support vector machines (SVMs) can be leveraged to predict remaining useful life while accounting for wear due to vibration, or water content sensors can be used to detect humidity in oil and predict wear due to water and other factors. For generator bearings, anomaly detection techniques are applied on temperature data to predict generator failures.

The benefits of predictive maintenance of wind turbines

  • Reduced costs: By scheduling maintenance routines in a way that maximizes the usable life of each equipment, and optimizes replacement and repair actions accordingly.
  • Lower downtime: By alerting operators about the specific failure mode, predictive maintenance reduces the time it takes for a technician to perform corrective action.
  • Fewer technician trips: Predictive maintenance reduces the number of scheduled or emergency trips that technicians need to make, thereby cutting emissions and costs in offshore scenarios.
  • Improved RoI: Predictive maintenance minimizes downtime and increases the energy output per dollar of investment. 

What next?

Until now, wind turbine maintenance has been carried out using legacy methods like manual inspection, reactive repairs, and preventive strategies. However, with the availability of predictive maintenance solutions, wind farm operators can significantly cut down their maintenance costs. They also represent significant value for offshore wind farms, where unforeseen failures will result in a drop in yield or massive costs of repair.

 Modernize your wind turbine maintenance operations and lower the cost of operations with predictive maintenance. Find out what Uptime AI can do for you, by contacting now.