基于PSO-GRU神经网络的青椒生长期需水预测

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，采用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，利用粒子群算法(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。

Prediction of water demand in green pepper growth period based on PSO-GRU neural network

There is a complex nonlinear relationship between water demand and factors such as air temperature, air pressure and relative humidity during the growth period of green pepper, and the change of water demand shows a time series and periodic law. In order to improve the prediction accuracy of the average daily water demand of green peppers during the growing period, a PSO-GRU water demand prediction model is proposed. The water demand data of green pepper and meteorological environment data obtained from 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, the water demand is used as labels, and the Gru neural network is used as the training model for water demand prediction. Aiming at the problem that GRU hyperparameters tend to fall into local optimality, particle swarm optimization (PSO) is used to optimize the hyperparameters of GRU model. Through the simulation experiment, the average daily water demand of green pepper in the growth period is predicted, then compared with RNN, LSTM and GRU to verify the superiority of the PSO-GRU model. The simulation results show that the prediction accuracy and fitting effect of the PSO-GRU model are significantly improved, RMSE is 0.505, MAE is 0.388, MAPE is 7.73 and R2 is 0.888. The PSO-GRU model can provide a basis for making irrigation plan and is conducive to water saving irrigation, and promote agricultural planting and water conservancy informatization.