Coastal Wind Speed's Multi-Parameter Forecasting Based on Quantitative Measurement of Its Predictability

Abstract

Wan, J.; Yao, K.; Ren, G.R.; Cao, Y.; Wang, W.; Zhao, X.X., and Yu, J.L., 2020. Coastal wind speed's multi-parameter forecasting based on quantitative measurement of its predictability. In: Bai, X. and Zhou, H. (eds.), Advances in Water Resources, Environmental Protection, and Sustainable Development. Journal of Coastal Research, Special Issue No. 115, pp. 662-665. Coconut Creek (Florida), ISSN 0749-0208.Wind speed prediction has important significance and value to coastal wind power. And the statistical model based on historical data is one of the most common prediction methods, which has made great progress for hourly mean wind speed forecasting with almost no consideration of predictability. In this paper, the correlation analysis was introduced to analyze the optimal wind speed predictable length based on the non-stationarity analysis of wind speed time series. The experimental results based on a large number of measured wind speed data shown that there exists a specific correlation length for the nonlinear wind speed time series. With the increasing of the correlation length, the prediction error gradually decreased and then increased. Thus, the correlation length corresponding to the turning point could be recognized as the optimal predictable length. Therefore, correlation analysis could be used to measure the predictability of wind speed. Based on the results, the quantitative measurement and prediction strategy with multiple characterization parameters of wind speed uncertainty is also presented. And the SVR and DBN algorithms are selected for several tests to verify that this multi-parameter forecasting method is feasible in practice.