【摘要】 针对降水量时间序列的多尺度非平稳性特点,提出了一种改进的集成经验模态分解(Modified Ensemble Empirical Mode Decomposition,MEEMD)-核极限学习机(Kernel Extreme Learning Machine,KELM)-果蝇优化算法(Fruit Fly Optimization Algorithm,FFOA)相结合的降水量预测模型。首先,利用MEEMD将非平稳的地降水量时间序列分解为一系列复杂度差异明显的降水量子序列;接着,针对每一个子序列分别建立KELM预测模型;为了进一步提高预测精度,将子模型的结果通过一组系数融合,并利用FFOA进行系数寻优,获得降水量的最终预测结果;最后,以重庆酉阳实测的年度降水量数据为例进行实验,并与BP神经网络、KELM以及MEEMD-KELM的三种预测模型进行比较。实验结果表明,MEEMD-KELM-FFOA模型的预测值能紧跟降水量的变化趋势,相比另外三种模型,体现出更好的预测效果。更多还原

【Abstract】 According to the multi-scale and non-stationary features of precipitation time series, a prediction model is proposed, based on Modified Ensemble Empirical Mode Decomposition(MEEMD)-Kernel Extreme Learning Machine(KELM)-Fruit Fly Optimization Algorithm(FFOA). Firstly, the precipitation time series was decomposed into a finite collection of intrinsic mode functions(IMFs) and a residue by MEEMD. Second, each IMF and the residue were utilized to establish prediction model using KELM. To further improve the accuracy of prediction, the result of each sub-series was multiplied by a coefficient and then summed up to obtain the final precipitation content. Here, all coefficients were optimized by FFOA. Finally, the simulation was performed by using the real precipitation data. The results show that the prediction value of the MEEMD-KELM-FFOA model can closely keep up with the trend of precipitation, and it is obviously better than the other three model.更多还原