Sequential Reliability Forecasting for Wind Energy: Temperature Dependence and Probability Distributions

Sequential wind simulation models have been developed to forecast effective load-carrying capability (ELCC), but they produce data that follow a Gaussian distribution, which can be considerably different from real wind-speed distributions, and they do not explicitly model the influence of temperature on the evolution of wind speed. The latter issue is of significant importance considering the strength of correlation between electricity load and temperature and the influence electricity load shapes have on system reliability. This paper presents a new approach to reliability estimation of wind facilities that models wind-speed distributions nonparametrically and includes the effect of temperature on the evolution of wind speed. A relatively long tall tower anemometer dataset is then used to test whether or not the model output produces ELCC distributions that are statistically similar to observed distributions. Results indicate that, relative to temperature independent models, temperature-dependent time-series models are better at both short-term wind-speed forecasting and long-term reliability forecasting. Results also show that reliability forecasts are relatively unaffected by the shape of the wind-speed distribution. Finally, it is apparent that model performance is robust to variation in the average wind power in the years used for model parameterization.