一种基于迁移学习的PM_(2.5)浓度预测混合模型  

A Transfer Learning-based Hybrid Model for PM_(2.5) Concentration Prediction

作　　者：卢新彪[1] 叶春林 陈艺森 吴文 陈钰丹 Lu Xinbiao;Ye Chunlin;Chen Yisen;Wu Wen;Chen Yudan(School of Artificial Intelligence and Automation,Hohai University,Nanjing 211100,China)

机构地区：[1]河海大学人工智能与自动化学院,江苏南京211100

出　　处：《系统仿真学报》2025年第4期882-894,共13页Journal of System Simulation

基　　金：国家自然科学基金(61573001)。

摘　　要：为解决PM_(2.5)浓度预测中因不相关特征导致的算力成本增加及数据分布随时间变化导致概率分布差异的预测精度下降问题,构建了基于迁移学习的混合深度学习模型TraTCN-LSTM-BiGRU。采用均值热力图算法,选择与PM_(2.5)浓度相关的气象因子作为模型输入特征;通过KL散度划分源域数据和目标域数据,并在模型中引入自适应层,实现领域间的分布适应性;设计TCN-LSTM-BiGRU模型,使用TCN提取多元变量中的高级空间特征,将提取的特征输入LSTM提取时间序列特征,通过残差连接融合特征并输入BiGRU进行预测。仿真结果表明:所提模型可以有效地预测PM_(2.5)未来变化趋势,并削弱数据分布差异所带来的影响。In order to solve the problems of increased computational cost due to irrelevant features and decreased prediction accuracy due to the difference in probability distribution caused by the change of data distribution over time in PM_(2.5) concentration prediction,this paper constructs a hybrid deep learning model TraTCN-LSTM-BiGRU based on migration learning.The meteorological factors related to PM_(2.5) concentration are selected as the model input using the mean-value heat map algorithm features;the source domain data and target domain data are divided by KL scatter and an adaptive layer is introduced into the model to achieve inter-domain distribution adaptation;the TCN-LSTM-BiGRU model is designed to extract high-level spatial features in multivariate variables using TCN,and the extracted features are fed into the LSTM for extracting time series features.The extracted features are fed into the LSTM to extract time-series features,and the features are fused by residual connection and fed into the BiGRU for prediction.Simulation results show that the model proposed in this paper can effectively predict the future trend of PM_(2.5),and effectively weaken the influence of the difference in data distribution.

关 键 词：迁移学习 PM_(2.5)浓度 均值热力图 概率分布差异 TraTCN-LSTM-BiGRU 

分 类 号：TP273.4[自动化与计算机技术—检测技术与自动化装置] TP391.9[自动化与计算机技术—控制科学与工程]