摘要
针对传统气温预测方法预测难度大、精度差及气象数据大而带来的计算量大等问题,提出一种基于主成分分析(PCA)和改进粒子群算法优化门控循环单元(GRU)的递归神经网络时间序列预测模型。首先,利用主成分分析算法对气象要素进行降维处理;然后,运用指数下降惯性权重和边界突变算子的改进粒子群算法(PSO)优化GRU神经网络。以南京地面观测站点的观测数据为样本数据,运用Python对模型进行训练,与传统的BP及LSTM神经网络预测模型对比,实验结果表明该模型具有更高的预测精度和稳定性。
In view of the fact that the traditional temperature prediction methods have difficult forecasting,poor accuracy,and heavy calculation burden caused by large meteorological data,a recursive neural network time series prediction model based on principal component analysis(PCA) and gated recurrent unit(GRU) optimized by improved particle swarm optimization(PSO) algorithm is proposed. The PCA algorithm is used to reduce the dimensions of the meteorological elements. The PSO improved by exponential decreasing inertia weight and boundary mutation operator is used to optimize the GRU neural network.The observation data from the ground observation site in Nanjing is taken as the sample data. The model is trained with Python.The experimental results show that the designed model has higher prediction accuracy and stability in comparison with the traditional BP(back propagation)and LSTM(long-and short-term memory)neural network prediction models.
作者
杨迎新
杜景林
武艳
YANG Yingxin;DU Jinglin;WU Yan(Nanjing University of Information Science&Technology,Nanjing 210044,China)
出处
《现代电子技术》
2022年第1期89-94,共6页
Modern Electronics Technique
基金
国家自然科学基金项目(41575155)。
