ALBERT

Description: As a promising renewable energy source, wind energy has increasingly gained worldwide attention. Providing high accuracy wind energy forecasting allows us to improve the economic and social benefits of wind power management, which reduces the generation costs and improves the security of the wind power system. In this paper, a novel hybrid forecasting model called E-SA-BP, which combines ensemble empirical mode decomposition, a simulated annealing (SA) algorithm, and a back-propagation neural network (BPNN), is developed to perform wind speed forecasting. First, ensemble empirical mode decomposition is used to decompose the original wind speed data series aiming to de-noise and then reconstruct the data series. Next, BPNN is applied to perform short-term wind speed forecasting, because BPNN can implement any complex nonlinear mapping function (as proven by mathematical theory) and approximate an arbitrary nonlinear function with satisfactory accuracy. However, due to the instability of the structure of the BPNN, SA is utilized to optimize the weight and threshold values of the BPNN through simulating the annealing process of metal objects after heating. Last, the data of six wind speed observation sites in Jiaodong Peninsula of China are chosen to test the performance of the forecasting models. The results show an effective decrease in the forecasting errors of E-SA-BP when it is compared with the Moving Average(1), Exponential Smoothing (ES)(1), ES(2), Autoregressive Moving Average Model, Autoregressive Integrated Moving Average, BP, SA-BP, and E-BP models.