Wind energy is rapidly emerging as an appealing energy option because it is both abundant and environmentally friendly. Because of the stochastic nature and intrinsic complexity of wind speed, precise and reliable wind-speed prediction is vital for wind-farm planning and the operational planning of power grids. To improve wind-speed forecasting accuracy or stability, many forecasting approaches have been proposed. However, these models usually only consider one criterion (accuracy or stability) and have limitations associated with using individual models. In this paper, an ensemble method optimized by a novel multi-objective optimization algorithm is introduced. With respect to ensemble weight coefficients, a bias-variance framework, which is formulated by a multi-objective optimization problem, is used to assess model accuracy and stability. A novel hybrid flower pollination with bat search algorithm is proposed to search for the optimal weight coefficients based on the previous step, while Pareto optimality theory provides the necessary conditions to identify an optimal solution. In addition, data decomposition and de-noising are also incorporated into the data pre-processing stage. To evaluate the forecasting ability of the proposed model, a case study of 12 wind-speed datasets from two wind farms in the eastern coastal areas of China was completed. The experimental results of this study indicate that the developed ensemble model is superior to other comparison models in terms of the high precision and stability of wind-speed prediction.