当前位置>主页 > 期刊在线 > 计算机技术 >

计算机技术2020年4期

​基于改进 PSO-LSTM 神经网络的气温预测
杨孟达
(成都信息工程大学 软件工程学院,四川 成都 610225)

摘  要:气象数据的数据量通常较大,传统长短时记忆(LSTM)神经网络针对气象数据人为调参十分困难,为了解决这个问题,提出了一种改进 PSO-LSTM 模型。其通过使用非线性变化惯性权重和学习因子的粒子群算法(PSO)对 LSTM 神经网络的相关参数进行优化,去除人为调参因素。实验使用两个不同气象站点的气象数据集,结果表明,与竞争预测模型相比,改进PSO-LSTM 模型具有更高的预测精度。


关键词:长短时记忆神经网络;粒子群算法;气温预测



中图分类号:TP183         文献标识码:A         文章编号:2096-4706(2020)04-0110-03

Temperature Prediction Based on Improved PSO-LSTM Neural Network

YANG Mengda

(School of Software Engineering,Chengdu University of Information Technology,Chengdu 610225,China)

Abstract:The amount of meteorological data is usually large. It is very difficult to adjust meteorological data manually by traditional long short memory (LSTM) neural network. In order to solve this problem,an improved PSO-LSTM model is proposed. The PSO algorithm,which uses the nonlinear inertia weight and learning factor,optimizes the parameters of the LSTM neural network and removes the human parameters. The experimental results show that the improved PSO-LSTM model has higher prediction accuracy than the competitive model.

Keywords:long and short-term memory neural network;particle swarm optimization algorithm;temperature prediction


参考文献:

[1] 施晓芬,陈翔,曹永勇,等 . 基于回归 BP 神经网络误差分级迭代法的逐时气温预测方法研究 [J]. 数学的实践与认识,2019,49(1):145-151.

[2] FUENTES M,CAMPOS C,GARCIA-LOYOLA S.Application of artificial neural networks to frost detection in central Chile using the next day minimum air temperature forecast [J].Chilean Journal of Agricultural Research,2018,78(3):327-338. 

[3] LIANG Y,QIU L,ZHU J,et al.A Digester Temperature Prediction Model Based on the Elman Neural Network [J].Applied Engineering in Agriculture,2017,33(2):143-148.

[4] 林希,张政.基于Elman神经网络的气温时间序列预测 [J].科技咨询导报,2007(25):4-5.

[5] 陶晔,杜景林 . 基于随机森林的长短期记忆网络气温预测 [J]. 计算机工程与设计,2019,40(3):737-743.

[6] XU C L,CHEN H X,WANG J Y,et al.Improving prediction performance for indoor temperature in public buildings based on a novel deep learning method [J].Building and Environment,2019(148):128-135. 

[7] 倪铮,梁萍 . 基于 LSTM 深度神经网络的精细化气温预报初探 [J]. 计算机应用与软件,2018,35(11):233-236+271.


作者简介:杨孟达(1995.06-),男,汉族,四川广元人,在读研究生,研究方向:计算机技术。