基于多尺度组合模型的区域农业旱灾损失率预测

Prediction of Regional Agricultural Drought Loss Rate Based on Multi-scale Combined Model

针对旱灾系统的非平稳和非线性特征,提出了基于经验模态分解和最小二乘支持向量机的多尺度组合预测模型。为了避免分解过程中旱灾序列取值域发生改变的问题,首先采用逆Logistic变换对原始序列的取值域进行扩展;然后使用经验模态分解将旱灾序列进行平稳化,提取出旱灾序列中不同时间尺度的子序列,根据子序列的波动特征选择适用的方法进行预测;最后将预测子序列进行重组和还原。以河南省农业旱灾综合损失率为例进行3步仿真预测。预测结果表明,多尺度组合模型的预测效果和精度均好于最小二乘支持向量机模型的,说明在旱灾序列处理中应优先选用多尺度分析法。

For the non-stationary and non-linear characteristics of drought system, a multi-scale combined forecasting model based on empirical mode decomposition and least squares support vector machine was proposed. In order to avoid changing the range of drought sequence in the decomposition process, the inverse Logistic transform was used to extend the range of original sequence. Then, the empirical mode decomposition was used to smooth the drought sequence and extract the subsequences of different time scales in the drought sequence. According to the fluctuation characteristics of the subsequences, an appropriate method was selected to predict the drought sequence. Finally, the prediction subsequences were reconstructed and restored. The three-step simulation prediction was carried out based on the comprehensive loss rate of agricultural drought in Henan. The result shows that the multi-scale combined forecasting model is better than the least squares support vector machine model in forecasting effect and accuracy, which indicates that the multi-scale analysis method should be given priority in drought sequence processing.