Data-Driven Performance Optimization of Wind Farms

Grant # 07-01
Principal Investigator: Andrew Kusiak
Organization: The University of Iowa
Technical Area: Renewable Energy

The generation of wind energy on an industrial scale is relatively new. Therefore, the performance of wind power farms has not been thoroughly studied. One of the weakest points in wind power generation is the low predictive accuracy of the energy output. Like industrial corporations managed by enterprise-wide systems, a software solution for prediction of wind farm performance (including the amount of energy produced) is needed. The envisioned wind farm performance prediction software should be able to predict the amount of energy to be produced on different time scales, e.g., 15 min, 2 hour, …, a day, and so on. Such software would transform a wind farm into to a wind power plant.

The wind energy industry is growing and the goal is to use wind power to generate as much as 20% of the total national energy needs. Research and education are needed to accelerate the planned growth of industry transforming wind (often called free fuel) into electricity. The State of Iowa is the largest producer of wind energy per capita, and is the third largest (after California and Texas) producer of wind energy in the nation. The proposed research is needed to attract new electric utilities, wind energy equipment manufacturers, and software developers to the state, thus further strengthening Iowa’s position in the wind energy enterprise. This research program enhanced by the newly established Wind Power Management Program at the University of Iowa will accelerate the growth of the wind energy generation in Iowa and lead to the creation of new, high quality jobs.


The proposed research offers a novel approach to modeling the performance of individual wind turbines as well as their collection (a wind farm). Besides the utility scale applications, the proposed solution can be scaled down to optimize residential wind turbines (KW range) dispersed over large areas. The lower sensory capability of household turbines will be mitigated by the wide availably of wireless communication and Internet connectivity.


The basic methodology used is this research is data mining. It is a new science of model discovery from large data sets. Data mining is ideally suited for research in wind power, and this new industry generates multiple volumes of data that have not been adequately explored.

The data-mining algorithms to be used in the proposed research offer numerous advantages:
- Generation of explicit knowledge and models that are understandable to a user
- High prediction accuracy of the generated models
- Evolution of models. The wind power research calls for the development of models that adjust to the changing wind and operating conditions.

Anticipated Outcomes
The basic deliverables of this research are:
- Data mining algorithms for autonomous creation of wind power performance models
- Predictive performance models of turbines and the wind park
- Predictive maintenance models based on monitoring data as a natural outgrowth of the research
- Prototype software for optimization of wind farm performance
- Demonstration of the software at the operating wind energy sites
- Graduate training of students entering the wind power profession
- Increased public awareness of wind power
- Improvement of the prediction accuracy of wind power output by least 9 percent