【摘要】 为实现风速的超前多步高精度预测,提出一种基于经验模式分解与自适应神经模糊推理的混合方法。该方法利用经验模式分解法对铁路风速进行多层分解计算以降低风速的强随机性,对分解后的各层风速数据分别建立自适应神经模糊推理预测模型并完成预测计算,最终加权各层预测值获得原实测数据的对应步数的预测结果。运用所提出的方法对青藏铁路某监控点的风速进行预测。研究结果表明:所提出的混合方法有效融合了经验模式分解法的信号细分性能和自适应神经模糊推理法的非线性追踪能力,混合模型的超前1步、2步、3步预测的平均相对误差分别为6.24%,11.11%和14.30%,体现出良好的非平稳信号预测性能。更多还原

【Abstract】 To get high-precision forecasting results, a hybrid method was proposed by adopting the empirical mode decomposition and the adaptive neural fuzzy inference system. The procedures of the proposed method are as follows. Firstly, use the empirical mode decomposition to decompose the non-stationary wind speed series into a group of sub wind speed layers. Secondly, utilize the adaptive neural fuzzy inference system to build multi-step forecasting models for all the decomposed wind speed layers. Thirdly, sum up the multi-step forecasting results of the decomposed layers to get the final predictions for the original wind speed signals. Experiment was made using the wind speed data sampled from a monitoring wind station along the Qinghai—Tibet railway. The results show that the proposed method combines the decomposing performance of the empirical mode decomposition and the nonlinear performance of the adaptive neural fuzzy inference system effectively. The mean percentage errors of the one-three step ahead forecasting results are 6.24%, 11.11% and 14.30%,respectively.Those errors indicate that the proposed method has satisfactory forecasting performance.更多还原