This Small Business Technology Transfer Research (STTR) Phase I project aims to develop quantitative models of wind variability to aid the design of reliable, low-carbon electric grid systems with high wind penetration. All abundant renewable resources are naturally variable, creating a challenge for their integra¬tion onto an ?always on? electric grid. While this variability challenge is now beginning to rival the importance of further reductions in the costs of generation hardware, existing variability models suffer numerous shortcomings that limit their accuracy and flexibility. This STTR project aims to develop and validate new models with a high degree of generality and universal applicability based on atmospheric sciences and the physics of turbulent flows. The proposed models will provide new conceptual insights into the temporal and geographic aspects of wind variability along with superior performance for risk-assessment and reliability estimation based on rare and infrequent events including wind droughts. These models will enable design (without overdesign) of a reliable grid, including not only changes to its electrical system, but also changes to the policy, regulatory, and financial systems that are essential parts of the electricity delivery system.
The broader impact/commercial potential of this project will be realized across multiple sectors. The project will provide enabling tools to broad engineering and financial constituencies involved in transforming the electricity system?s infrastructure and organization as it moves towards a sustainable and renewable future. Application of these tools will lower the costs of wind integration, thereby speeding grid transformation and hastening the reduction of electric-sector greenhouse gas emissions. Grid planners and operators will benefit from easier-to-use engineering and simulation modules in commercial grid software packages. In the financial sectors, the model tools will enable accurate estimation of wind energy risks, allowing new risk-management market instruments and contract structures. The fundamental conceptual understanding of variability along with actionable quantitative metrics made possible by the proposed models will enable the formation of new ventures and enterprises capitalizing on emerging opportunities in the smart-grid space.
Better models of wind variability could be helpful in the integration of wind power into our "always-on" electric grid. The work reported here focused on transcending the empirical approaches typically used, and developing instead robust statistical variability models allowing results from one place or one time to be extended with known confidence to new locations and situations. This lays the foundation for comprehensive and flexible models for applications in the electric-utility and financial-risk industries. Combining outputs from wind plants separated enough to be uncorrelated reduces power variability. However, the way that correlation on different variability timescales falls off with increasing distance between plants is not fully understood. We investigated two large wind-speed data sets, one from Canada and the other from the Bonneville Power Authority (BPA), and found they exhibited very similar behavior. Timescale has essentially no effect on correlation fall-off for wind-speed variations that change slower than about once per four days. At timescales of twice per day correlation falls off ten times faster than for slow variations. Correlations in wind-speed variations that occur on a 3-hour timescale fall with distance 100 times faster than do variations slower than once per four days. The degree of wind-power variability reduction depends on the relevant metric. The odds that combined power of N wind plants falls below a chosen threshold is important to electric utilities, who use this value to determine the amount of generation they should hold in reserve for reliable 24/7 operation. Using five years of hourly wind-speed records from nine tall U.S. towers, we found that as more wind plants are combined, the odds of sub-threshold output fall exponentially as exp(–QN). This fall-off with N is much faster than the square-root of N reduction for conventional variability metrics like the standard deviation. We found that adding just three more plants to a combination could reduce the odds of generation falling below 1% by a factor of 20. Another variability metric of crucial importance to grid operators and planners is the likelihood of wind-power ramps of different magnitudes. Our literature research uncovered work in turbulence theory that offered "a parsimonious and universal description of turbulent velocity increments." We found that A single parametric probability density function fit BPA wind power ramps over 0.08 h to 32 h lag times, capturing the change of probability tail shape from exponential to Gaussian for both single stations and for combined outputs of multiple stations. This Normal Inverse Gaussian distribution’s three independent parameters followed simple, "universal" trajectories with changing lag. Finally, we developed a quantitative definition of wind drought, and applied it to the Canadian wind-speed data set, where we found that wind power exhibits the "Hurst Phenomenon" whereby the total battery or financial-capital "reservoir" needed to cover wind droughts over a given time period scales as the three-quarter power of the length of the time period. Wind speeds exhibit long-memory, indicating droughts, and their corresponding financial risks, will be longer and more severe than predicted by today’s short-memory ARMA and Markov models. These results demonstrate the value of using physical and statistical theories to understand wind power variability. This approach uses well-understood phenomena and techniques to make the most of scarce data, and allows results from one place or one time to be extended with known confidence to new locations and situations.
