基于二次分解和改进沙猫群优化算法的空气质量预测
Air quality predication based on two-layer decomposition and improved sand cat swarm optimization
朱菊香 1张诗云 2张涛 2孙君峰 2张赵良3
作者信息
- 1. 无锡学院轨道交通学院 无锡 214105
- 2. 南京信息工程大学自动化学院 南京 210000
- 3. 无锡学院轨道交通学院 无锡 214105;江苏省工业环境危害要素监测与评估工程研究中心 无锡 214105
- 折叠
摘要
准确预测空气质量对人们的日常生活具有重要意义,提出了一种二次分解和改进沙猫群算法(improved sand cat swarm optimization,ISCSO)优化长短期记忆网络(long short-term memory,LSTM)相结合的预测模型.首先,利用完全自适应噪声集合经验模态分解(complete ensemble empirical mode decomposition with adaptive noise,CEEMDAN)算法将PM2.5数据分解为多个子序列,对预测效果不满意的重构序列使用变分模态分解(variational mode decomposition,VMD)方法进行二次分解;其次,引入Cubic混沌、螺旋搜索策略和麻雀警戒机制改进沙猫群算法,有效提高了算法的全局搜索性能和收敛速度;最后,采用改进的沙猫群算法对LSTM模型参数进行优化,将各个子序列导入ISCSO-LSTM模型预测并叠加得到最终预测结果.实验结果表明,CEEMDAN-VMD-ISCSO-LSTM组合模型具有较低的预测误差,相比CEEMDAN-VMD-LSTM和CEEMDAN-VMD-SCSO-LSTM模型,该模型在均方根误差方面分别降低了2.21和1.04 μg/m3,在拟合度方面分别提高了4.9%和2.1%.
Abstract
Accurate prediction of air quality is of great significance to people's daily life,therefore,a predictive model based on quadratic decomposition and improved sand cat swarm optimization(ISCSO)to optimize the long short-term memory(LSTM)network was proposed.First of all,The PM2.5 data was decomposed into multiple subsequences using complete ensemble empirical mode decomposition with adaptivenoise(CEEMDAN)algorithm,and the reconstructed sequence that are not satisfied with the prediction effect was quadratically decomposed by variational mode decomposition(VMD)method.Secondly,the sand cat swarm optimization was improved by introducing Cubic chaotic,spiral search strategy and sparrow alert mechanism to improve the global search performance and convergence speed of the algorithm.Finally,a improved sand cat swarm algorithm was used to optimize the LSTM model parameters,the individual subsequences were input into the ISCSO-LSTM model for prediction and superimposed to obtain the final prediction results.The experimental results show that the CEEMDAN-VMD-ISCSO-LSTM combination model exhibits lower prediction errors,compared to the CEEMDAN-VMD-LSTM and CEEMDAN-VMD-SCSO-LSTM,the model proposed in this article has a 2.21 and 1.04 μg/m3 reduction respectively in root mean square error,and has a 4.9%and 2.1%higher respectively in term of fit.
关键词
空气质量预测/二次分解/改进沙猫群算法/长短期记忆网络Key words
air quality predication/two-layer decomposition/improved sand cat swarm optimization/long short-term memory network引用本文复制引用
基金项目
"太湖之光"科技攻关计划(k20221050)
出版年
2024
NETL
NSTL
万方数据