Robust Distributed Wind Power Engineering

Abstract:

Executive Summary

To detect wind turbine blade defects, an embedded platform must be mounted on the wind turbine rotor. The platform, proposed by Purdue, supports all sensor channels needed to detect cracks and indicate how to mitigate damaging loads. The platform will run a Java virtual machine, thus our fault detection algorithms will be ported to Java. With the embedded platform installed on the wind turbines, crack detection will be performed in real-time at the Vanderbilt Wind Turbine Test Facility. Results from the experiments will help designing a wind farm optimization plan which combines real-time fault detection with the wind turbine sensitivity matrix to provide control decisions that minimize damage propagation while optimizing the farm's power output. Macroprogramming and an active control feedback will validate this methodology on the wind farm scale. A final report and a journal paper will document our work.

Background

The wind power industry has grown rapidly in recent decades. As many wind turbines have been built and deployed, it turns out that the maintenance costs consume a significant portion of the total energy costs of wind farms [3]. To minimize unscheduled maintenance costs (which are typically much higher than scheduled maintenance costs), it is important to detect defects on the turbines before they lead to a catastrophic failure of the system. Wind turbine blades require robust structural health monitoring technologies not only because of their higher costs of replacement but also because the failure of a wind turbine blade leads to failures of other subsystems in the turbine such as the tower and drive train. 50% of turbines fail once per year for ten years leading to 130 hours of maintenance per turbine. Blade damage is the most expensive damage to repair and also has the longest repair time. Blades account for 15-20% of the total turbine cost. Blade failure leads to secondary component damage in the entire system. Predictive maintenance of wind turbines could provide an estimated 81% reduction in cost when compared to reactive maintenance, the current industry standard. The life of a damaged blade could be further extended by coupling blade damage detection algorithms that can provide estimates of blade life to the wind turbine control system. The control system could then change the blade pitch and turbine yaw to reduce the load on a turbine with a damaged blade to optimize energy output and allow for planned repair and maintenance of the blade which would reduce operating and maintenance cost. This combination of blade damage prognostics and wind turbine control could be applied to an entire wind farm to optimize the power generation and operating and maintenance costs.