At the origin of WindTurBars is the wind energy industry in Europe, with a need to reduce the rate of power system failures and the corresponding default downtime that will contribute significantly to the overall reduction of the operating and maintenance cost associated with current and future wind turbines.
The European Windturbars project has achieved its objective of developing and building a sensor and monitoring system for predictive maintenance of the electrical system for off-shore wind turbines.
Stopping the activity of a wind farm to carry out maintenance work or due to the occurrence of a fault is a situation that must be avoided at all costs. But when the wind farm is installed offshore, the maintenance of the wind turbines, known as off-shore, is particularly costly and difficult to plan. In many cases, helicopter access is necessary and weather conditions are particularly harsh.
The interest of the wind energy market in incorporating these predictive maintenance technologies is substantial, given that electrical failures in wind turbines have a great impact on their performance. Until the development of the project, predictive maintenance was generally oriented towards mechanical failures, but in the Windturbars project we have given it a new focus, diagnosing the different electrical components of the wind turbine.
Due to adverse weather conditions, the electrical system components of an off-shore generator are subject to more wear and tear than in other environments, so it is necessary to pay more attention to maintenance in order to minimize malfunctions and, thus, reduce the number of shutdowns in power generation. The great advantage of this methodology is that we anticipate catastrophic failure. Once the anomalous degradation of a component is detected, maintenance tasks are activated to prevent the failure from occurring. Since degradation is progressive, maintenance tasks can be fitted into other planned periodic maintenance.
The major technological challenges faced have been mainly in two areas.
The first one, focused on the field of power electronics, where the great challenge of having an operational power train to perform system validation has been addressed. This has required a combination of simulation, programming, signal processing and, finally, hardware implementation techniques, all within a very tight time frame.
In the second area, “we have worked on migrating prediction techniques applied in logistics areas and taking them to an application as different as the field of electrical fault diagnosis.
In its operation, we detect patterns that appear in the electrical signals of the main elements that make up the wind turbine as they progressively degrade. Basically, we analyze in the time and frequency domain processed electrical signals. One of the novelties of the algorithm developed for data analysis is the ability of the system to perform a self-learning process in order to be able to base decision making on the presence or absence of a failure pattern based on historical data. This allows the detection system to be optimally and individually adapted to each wind turbine.
