摘要
青椒生长期内需水量与气温、气压、相对湿度等因子之间存在复杂的非线性关系,需水量变化呈现出时序性和周期性的规律,为提高青椒生长期日均需水量的预测精度,提出一种PSO-GRU (粒子群算法-门控循环单元)青椒生长期日均需水预测模型。以2014—2018年实验所得的青椒需水和气象环境等数据为数据源,将日均气温、气压、风速等六维数据作为特征集,需水量作为标签,GRU神经网络作为需水预测的训练模型,并针对GRU超参数容易陷入局部最优的问题,利用PSO优化GRU模型的超参数,通过仿真实验对青椒生长期日均需水量进行预测,并与RNN, LSTM和GRU等模型进行对比,验证PSO-GRU模型的优越性。仿真实验结果表明:PSO-GRU模型的预测精度和拟合效果显著提高,RMSE为0.505, MAE为0.388, MAPE为7.73,R2为0.888。PSO-GRU模型可为制定灌溉计划提供依据,有利于节水灌溉,推动农业种植水利信息化。
There is a complex nonlinear relationship between water demand and other factors such as temperature,air pressure and relative humidity during the growth period of green pepper. The water demand change shows the regularity of timing and periodicity. In order to improve the prediction accuracy of daily average water demand in green pepper growth period,a PSO-GRU(Particle Swarm Optimization-Gated Recurrent Unit) water demand prediction model is proposed. The water demand data of green pepper and meteorological data from the experiments in 2014—2018 are used as data sources,the six-dimensional data such as daily average temperature,air pressure and wind speed are used as feature sets,while the water demand is used as labels. The GRU neural network is applied as the training model for water demand prediction. Due to the problem that GRU hyper-parameters tend to fall into local optimum,PSO is applied to optimize the hyper-parameters of GRU model. Through simulation experiments,the daily average water demand in green pepper growth period is predicted,which is compared with RNN,LSTM and GRU models to validate the superiority of the PSO-GRU model. The simulation results show that prediction accuracy and fitting ef fect of the PSO-GRU model are significantly improved,with the RMSE 0.505,the MAE 0.388,the MAPE 7.73 and the R2 0.888. The PSO-GRU model can provide reference for irrigation plans. It benefits water-saving irrigation and promotes water conservancy informatization in agricultural planting.
作者
连晓晗
马永强
刘真
刘心
LIAN Xiaohan;MA Yongqiang;LIU Zhen;LIU Xin(School of Information&Electrical Engineering,Hebei University of Engineering,Handan 056038,China)
出处
《水利信息化》
2023年第1期33-39,共7页
Water Resources Informatization
基金
河北省高等学校科学技术研究青年基金项目(QN2021034)。
关键词
PSO
GRU
需水预测
神经网络
青椒生长期
节水灌溉
PSO
GRU
water demand prediction
neural network
green pepper growth period
water-saving irrigation
