基于EEMD与极限学习机的短期风速组合预测模型

Short-term Wind Speed Forecasting Based on Combination of EEMD and Extreme Learning Machine

针对风速的波动性和随机性大导致风速难以准确预测的问题,提出了基于经验模态分解方法(EEMD)和引力搜索算法(GSA)优化极限学习机(ELM)组合的短期风速预测模型.在风速预测建模过程中,为了降低风速的非线性和波动性,利用EEMD算法将原始风速数据分解成不同频率的相对稳定子集;然后,应用ELM对各个子集序列分别构建风速预测模型;使用全局搜索能力强的GSA算法优化ELM的输入权值和隐含层阈值,避免ELM陷入局部最优,提高组合模型的风速预测性能;最后,将各ELM模型的预测结果进行叠加,得到组合模型的最终风速预测结果.通过安徽某风电场的历史风速数据验证该预测模型,并与ELM、EEMD-ELM、EMD-GSA-ELM、WT-GA-SSVM及Persistence等模型比较,仿真结果表明,所提出的基于经验模态分解方法和引力搜索算法优化极限学习机的组合风速预测模型获得较好短期风速预测结果.

In order to solve the problem that wind speed is difficult to be predicted accurately due to its fluctuation and randomness,a short term wind speed prediction model based on the combination of em pirical mode decomposition （EEMD） and gravity search algorithm （GSA） optimizing extreme learning machine （ELM） is proposed.To reduce the nonlinearity and instability of wind speed,EEMD algorithm is applied to decompose the wind speed data into relatively stable subsets with different frequencies. Then,ELM is adopted to construct wind speed prediction models for each subset sequence respectively. To improve the prediction effect of ELM,GSA algorithm with strong global search ability is employed to optimize the input weight and hidden layer threshold of ELM.Finally,the prediction results of the ELM models are accumulated to obtain the final wind speed prediction results of the proposed model.The his torical wind speed data of a wind farm in Anhui province are used to evaluate the EEMD GSA ELM model.The simulated results suggest that the proposed model is an effective short term wind speed fore casting method.