The invention relates to an online hybrid forecasting method for the short-term wind speed of a wind power field, which comprises six models: a data preprocessing model, a wavelet decomposition model, a time-series forecasting model, a support vector machine forecasting model and a wavelet reconstruction model. After the wind speed data of the wind power field is acquired by a wind speed sensor, a formed historical wind speed series is preprocessed by the data preprocessing model; the wavelet decomposition model is applied to decompose the preprocessed wind speed series into a high-frequency part and a low-frequency part; a time-series forecasting method is utilized to forecast a high-frequency component obtained by decomposing; a support vector machine forecasting method is utilized to forecast a low-frequency component; and finally, a wavelet reconstruction method is utilized to reconstruct two forecasting results to obtain the forecasting result of the short-term wind speed of the wind power field of the next moment.