Optimal maintenance of wind power plants
My name is Quanjiang Yu, and I am a PhD student in the optimization group at the Department of Mathematical Sciences, Chalmers University of Technology and University of Gothenburg. My research project ‘’Optimal maintenance of wind power plants’’ is financed by the Swedish Wind Power Technology Centre (SWPTC) and the Swedish Research Council. My supervisor is Michael Patriksson and my examiner is Ann-Brith Strömberg.
Climate change now affects every country on every continent. It disrupts national and personal economies, affects lives, communities and countries, today and even more tomorrow. Global warming has been attributed to increased greenhouse gas emission concentrations in the atmosphere by burning of fossil fuels. Wind power, as an alternative, is plenty, renewable, and produces almost no greenhouse gas during operation.
A big part of the wind turbine cost during operation is maintenance cost, especially for offshore wind farms. Based on a study done by the University of Strathclyde, for a 3.3MW wind turbine, in average, during operation 38% of the total cost is spent on maintenance. The area of wind turbine operation and maintenance represents a growing segment and business opportunity in the wind energy industry.
The main goal of this project is to develop an app, using a mathematical optimization model, which can generate a short-term maintenance schedule, take everyday data from the SCADA system into the model and update the schedule every week.
There are two ways to perform maintenance: first is corrective maintenance (CM), which is what people normally do, maintaining the components after they break down. But there are some consequences: when one component breaks down, there may be secondary damage; also the wind turbine owners need to order a maintenance team to maintain it, which takes time, and long downtime results in bigger production losses; and so on. So sometimes it is better to perform preventive maintenance (PM), which means that you maintain the components before they break down.
Figure 1: Comparison of PM and CM of gearbox. The right figure zooms in on the circled part of the left figure.
Above is a comparison between PM and CM of a gearbox. The x axis denotes how often we perform maintenance. The y axis denotes the cost per month, so we get the cost per month of PM/CM schedules, respectively. In the beginning the PM cost is really high, due to the fact that there is very little risk that the components will fail. Then the PM cost decreases. At around month 15, the PM cost is lower than the CM cost. After zooming in on the circled part, we get the right figure, where at around month 44 we find the best PM schedule, it costs around 1.7 (k$) per month. We can see that the CM schedule costs around 3 (k$) per month, which is almost twice as much as the best PM schedule. That is the reason why we need to maintain the component before they break down, i.e., perform PM.
This case only considers one component, but we have developed a mathematical optimization model that considers the whole wind farm, and generates a short-term maintenance schedule which indicates to the maintenance staff at what time they should maintain which components.