基于EMD与ELM的输电线路山火蔓延速度组合预测模型

针对复杂环境下输电线路山火的影响因素，提出了基于经验模态分解（empirical mode decomposition，EMD）与极端学习机（extreme learning machine，ELM）的输电线路山火预测模型。首先利用小波去噪对采集的风速时间序列进行噪声分析，根据序列的不同进行归类重构，产生新的风速时间序列；然后利用经验模态分解将输电线路山火成因分解为一系列具有不同特征尺度的子序列；接着利用交叉验证法和重构相空间法确定学习机的各种参数和输入维数；再利用极端学习机输电线路山火进行建模预测分析。仿真结果表明基于经验模态分解与极端学习机的输电线路山火组合预测模型可以有效预测24 h之内的山火蔓延速度，为实现输电线路山火在线较高精度的预测提供了可能。

Mountain Fire Spread Speed Combined Forecasting Model for Transmission Line Based on EMD and ELM

According to the impact of mountain fire of transmission line in complex environment, the mountain fire prediction model was proposed based on the methods of empirical mode decomposition （EMD） and extreme learning machine （ELM）. Firstly, the noise of collected wind speed time series was analyzed by using wavelet transform, and the classification and reconstruction were carried out according to the different sequences, in order to reconstruct new wind speed time series. Secondly, the factors of the mountain fire were decomposed into a series of sub-sequences with different characteristics scales by using EMD. Thirdly, cross-validation method and phase space reconstruction method were used to determine various parameters and input dimensions of machine learning, and then the modeling and forecasting analysis was carried out for the mountain fire of transmission line by using ELM. The simulation results show that the combined forecasting model for the mountain fire of transmission lines based on EMD and ELM can effectively predict fire spread speed within 24 h, which can provide the possibility to realize the online prediction of the mountain fire in transmission line with high precision.