On Wind Turbine Operation Monitoring and Maintenance – Part 1

By Ehsan Salehi

 

As well as being green, wind energy is one of the cheapest ways of generating electricity, and it is currently the most common generating technology for new installations in Canada. According to Canadian Wind Energy Association (CanWEA), more than 11,000 MW electricity is generated from installed wind turbines in Canada as of 2016 [1], and “In the last five years, Canada has seen more wind energy capacity installed than any other form of electricity generation, averaging 1275 MW of new build each year”. In order to make the generated wind energy more profitable comparing to fossil fuels, cost per kilowatt needs to be minimized. “Cost” in general refers to manufacturing and installation costs, and operation and maintenance costs. In terms of operation and maintenance cost control, there are three main strategies: Reactive, Preventive and Predictive maintenance [2].

A reactive strategy [3] means to run a component until it is damaged and causes the wind turbine (or a machine in general) to shut down. Subsequently, the repair is performed and operation resumes until the next incident. This strategy is not economical; the reason being, sudden component repairs and replacements could cost much more than planned maintenance, and also the breakdown of a component in the drive-train could damage other components, which itself is an additional cost. With reactive strategy, the operation budget is harder to control since the breakdown of a component is unpredictable, and spare parts or required contractors may not be available immediately.

A preventive maintenance [4] is a planned maintenance strategy (time-based) which is triggered and scheduled based on events. It relies heavily on operator experience, age of the machine, and manufacturer recommendations. The fundamental assumption is that an operating component has a certain life and is to be replaced or repaired at specific time frames. The main problem with preventive maintenance is that the intervals between inspections in most cases are too long to detect a defect at its early stage. The other issue with the preventive maintenance strategy is that the scheduled inspection intervals are based on the average operation conditions, whereas each wind turbine and wind farm has its own site and operating condition. The manufacturer recommended inspection schedule for a same component may not be suitable for a wind farm in different location, and most likely the 20-year operation life might not be met.

A predictive maintenance [5], which is also called condition-based strategy, is the cost-optimal strategy. It is performed by monitoring the status of the machine, based on several sets of data (such as vibration, oil, temperature, etc.). By analyzing the online data, the operator can potentially detect the issues as early as possible and schedule applicable economical remedies. For example, wind farm operators in Canada do not tend to schedule any maintenance in winter time due to the harsh weather and the cost. If they follow the reactive or preventive maintenance strategies, a sudden breakdown of a component might happen in winter and they do not have any other option other than shutting down the turbine and replace the part. However, if they detect the defect early enough by data analysis, they could effectively apply temporary remedies to delay the breakdown and have the part replaced in warmer seasons. Through this strategy, wind farm operators can also repair and replace a group of parts at the same time, as one of the major costs of wind turbine repair is the daily cost of crane rental. Instead of renting a crane for a few days to change only one part, they can replace a group of damaged components on several wind turbines in the farm.

Wind turbine failure is mostly associated with the abrupt changes of wind speed and direction. It causes severe stress and fatigue condition that can result in blades, rotor, coupling, gearbox, and bearings to fail more easily than in other mechanical systems. Drive-train failure is a common problem in mechanical systems. It is also one of the most studied problems in the field of mechanical component condition monitoring. Although there have been a large number of techniques developed for fault diagnosis, a reliable and easy-to-understand method that can deal with variable speed applications has not emerged. This represents the gap between research community and a broad range of industries in the field. From experience, older wind turbines have an annual maintenance cost of 3% (in average) of the original cost of the turbine. Condition monitoring and predictive maintenance can prevent fault propagation and significantly reduce maintenance cost. Canada as one of the leaders in renewable energy marketing and production, with the large number of wind turbines already installed, is focusing on maintenance and reliability issues to maximize turbine availability.

Several advanced signal processing schemes to improve reliability of machine diagnostics and prevent misdiagnosis have been developed. Yet, the industry typically relies on simple techniques. For industry, the problem is not lack of sophisticated diagnostic techniques, but rather scarcity of simple yet reliable approaches to support unskilled operators make important decisions without a specialist, or automated systems to allow a small group of engineers to run large wind farms.

Fault Diagnosis Techniques:

Fault diagnosis is the main part of the predictive maintenance approach, which is done before root causes analysis and prognosis. The wind power industry mainly uses three main techniques to detect drive-train faults:

  1. Oil Analysis

Oil analysis is used extensively in the wind energy industry as a useful method to monitor bearings and gearboxes [6]. Lubricant samples are collected in order to assess whether the lubricant is still healthy. Also, contaminants in the lubricant can indicate if any environmental debris/dirt or wear particles are present in the lubricant which significantly could reduce the service life by causing machine wear.

The procedure includes oil sampling, analytical tests and data analysis [7], which provides information on form, quantity, and size of the derbies. If there are wear particles in the oil samples, as a result of a defect in the component, the defect is potentially severe and immediate action needs to be taken. While this technique has proven to be useful, oil analysis cannot be used to detect the location of the defect in the component, as they are usually manufactured from a same material [8].

  1. Temperature Analysis

Thermocouples or similar devices (e.g. Resistance Temperature Detector- RTD) are attached to the component to collect temperatures to analyze the gradient [9]. However, while this method is useful, thermal analysis is also not a robust analysis to detect the location and size of the faults, specially in bearings [7]. Non-destructive infrared thermography method, on the other hand, is capable of detecting faults at their early stages providing their locations, yet this method is not currently cost efficient and easy to implement for wind turbines [10].

  1. Vibration Analysis

Vibration analysis is perhaps the most efficient type of drive-train defect detection method [11]. For example, an undamaged bearing generates a steady state vibration, but a fault in any elements of it can change the condition and produce noticeable vibration impulses. In other words, fault(s) on bearing element amplify the vibration. Therefore vibration analysis is a great tool to detect these types of changes. Vibration analysis (including time domain, frequency domain and combination of time and frequency domains) of mechanical components has been used for a long time in both academia and industry, and has been significantly improved during almost the last two decades. In terms of wind turbine application, the very old turbines did not benefit from online vibration monitoring, but today’s installed turbines are typically fully equipped with vibration sensors on different parts of the drivetrain including main bearing, gearbox, gearbox bearings, and generator bearings, and an operation centre monitors the status of the drive-train [12]. For a typical 1.5 MW wind turbine, eight to eleven vibration sensors are installed on the drive-train [13]. For these wind turbines there is one sensor on the main bearing, six on the gearcase, and four on the generator bearings (two on drive-end and two on non-drive-end bearings). Vibrations from the wind turbine drive-train, unlike oil samples, can be monitored remotely from the diagnosis center. There are several communication configurations, but typically a group of wind turbines are connected locally to a small server, which itself is connected to wind farm server via wireless connection. The wind farm server is then connected to the diagnosis server via a local-area network (LAN) and can be controlled and monitored remotely.

 

References:

[1] http://canwea.ca/wind-energy/installed-capacity

[2] C. A. Walford, Wind turbine reliability: understanding and minimizing wind turbine operation and maintenance costs. United States. Department of Energy, 2006.

[3] L. Swanson, “Linking maintenance strategies to performance,” International journal of production economics, vol. 70, no. 3, pp. 237–244, 2001.

[4] J. Nilsson and L. Bertling, “Maintenance management of wind power systems using condition monitoring systems- life cycle cost analysis for two case studies,” IEEE Transactions on energy conversion, vol. 22, no. 1, pp. 223–229, 2007.

[5] A. Kusiak and W. Li, “The prediction and diagnosis of wind turbine faults,” Renewable

Energy, vol. 36, no. 1, pp. 16–23, 2011.

[6] W. Musial, S. Butterfield, and B. McNiff, “Improving wind turbine gearbox reliability,” in European Wind Energy Conference, Milan, Italy, pp. 7–10, 2007.

[7] A. Rezaei, Fault Detection and Diagnosis on the rolling element bearing. PhD thesis, Carleton University Ottawa, 2007.

[8] T. Akagaki, M. Nakamura, T. Monzen, and M. Kawabata, “Analysis of the behaviour of rolling bearings in contaminated oil using some condition monitoring techniques,” Proceedings of the Institution of Mechanical Engineers, Part J: Journal of Engineering Tribology, vol. 220, no. 5, pp. 447–453, 2006.

[9] H. Maxwell, “How to install maintainable bearing temperature sensors,”

[10] W. Kim, J. Seo, and D. Hong, “Infrared thermographic inspection of ball bearing; condition monitoring for defects under dynamic loading stages,” in 18lh World Conference on Nondestructive Testing, Durban, no. 256, Citeseer, 2012.

[11] W.Wang and O. A. Jianu, “A smart sensing unit for vibration measurement and monitoring,” IEEE/ASME Transactions on Mechatronics, vol. 15, no. 1, pp. 70–78, 2010.

[12] R. Hyers, J. McGowan, K. Sullivan, J. Manwell, and B. Syrett, “Condition monitoring and prognosis of utility scale wind turbines,” Energy Materials, 2013.

[13] S. Sheng, H. Link, W. LaCava, J. Van Dam, B. McNiff, P. Veers, J. Keller, S. Butterfield, and F. Oyague, “Wind turbine drivetrain condition monitoring during grc phase 1 and phase 2 testing,” Contract, vol. 303, pp. 275–3000, 2011.