基于二层分解技术和改进极限学习机模型的PM_(2.5)浓度预测研究

准确的PM_(2.5)浓度预测对于保护公众健康和提高空气质量有重要意义,然而,由于PM_(2.5)浓度序列的随机性、非线性以及非平稳性等特征增加了对其准确预测的难度.本文提出了一种基于二层分解技术和改进极限学习机(ELM)模型的PM_(2.5)浓度预测方法,该方法融合了快速集成经验模态分解(FEEMD)和变分模态分解(VMD)两种分解技术以及经过差分演化(DE)算法优化的ELM模型.为了验证所提出预测方法的有效性,本文使用该方法对北京市和石家庄市的PM2.5浓度序列进行了预测研究.结果表明:1)相比于单层分解技术,本文提出的二层分解技术可以更加有效地降低PM2.5浓度序列的非线性及非平稳性特征;2)基于二层分解技术的DE-ELM预测模型可以显著提高PM_(2.5)浓度的预测精度.

PM2.5 concentration forecasting based on two-layer decomposition technique and improved extreme learning machine

Accurate PM_2.5 concentration forecasting is crucial for protecting public health and improving air quality. However, the randomness, non-linearity and non-stationarity of PM_2.5 concentration series increase the difficulty in PM_2.5 concentration forecasting. In order to improve the accuracy of PM_2.5 concentration forecasting, this paper proposes a novel hybrid model based on two-layer decomposition technique integrated fast ensemble empirical mode decomposition (FEEMD), variational mode decomposition (VMD) and extreme learning machine (ELM) model optimized by differential evolution (DE) algorithm. To testify the validity of the proposed model, the PM_2.5 concentration series of Beijing and Shijiazhuang are taken as the test cases to conduct empirical study. Based on the experiment results, the following two conclusions can be obtained: 1) compared with single decomposition technique, the proposed two-layer decomposition technique can efficiently decrease the characteristics of non-linearity and non-stationarity of PM_2.5 concentration series; 2) the proposed FEEMD-VMD-DE-ELM model can precisely forecast the PM_2.5 concentration.