In order to achieve short-term wind speed prediction with high accuracy, this paper proposes a Hybrid Population Particle Algorithm (HPPA) and multi-Quantile Robust Extreme Learning Machine (QR-ORELM) based short-term wind speed forecasting. In signal processing stage, it adopts a Time Adaptive Filter based Empirical Mode Decomposition (TVF-EMD) to decompose the original wind speed series into several Intrinsic Mode Functions (IMFs) to decrease the volatility. Then HPPA is used to extract the features of each IMF, and QR-ORELM is used to build their forecasting models respectively and the parameters are optimized by HPPA. Finally, the eventual result can be obtained through aggregating the prediction value of each IMF. Simulation results show that after adopting HPPA for feature selection and model parameter optimization, the proposed hybrid method has higher prediction accuracy. Meanwhile, TVF-EMD based Wind Speed Forecasting (WSF) methods could further improve the predicting accuracy. This work is supported by National Natural Science Foundation of China (No. 51707041) and Science and Technology Project of China Southern Power Grid (No. GDKJXM20162087).