Study On Intelligent Methods Of Wheat Irrigation

With the rapid change of global climate and the increasing of global population,the contradiction between supply and demand of water resources is becoming more and more prominent.Agricultural irrigation water has become one of the primary problems in agricultural production and processing.The intellectualization and independence of agricultural irrigation methods have gradually become the inevitable trend of modern agricultural development.Based on this,this paper takes wheat field as an example to design a wheat irrigation method based on deep reinforcement learning.This paper mainly studied the prediction method of wheat evapotranspiration and the intelligent irrigation control strategy,and the main contents were as follows:1.A prediction model of wheat evapotranspiration based on deep extreme learning machine was designed.First,Principal Component Analysis(PCA)was used to analyze wheat evapotranspiration and screen out the key factors affecting wheat evapotranspiration.Then,combined with the advantages of Grey Wolf Optimization algorithm(GWO),aiming at the low prediction accuracy of Deep Extreme Learning Machine(DELM)model,A deep extreme learning machine model based on improved Grey Wolf algorithm was proposed to predict evapotranspiration of wheat.Finally,experimental results show that compared with the original prediction model,the algorithm proposed in this section has higher prediction accuracy.2.A wheat irrigation control strategy based on deep reinforcement learning was proposed.First,on the basis of Proximal Policy Optimization(PPO),combined with negative incentive clip function and dynamic cutting interval strategy,A deep reinforcement learning method based on Negative Motivation Dynamics PPO(NDPPO)algorithm is designed.Then,by designing reinforcement learning space and reward and punishment function,the improved deep reinforcement learning method was used to construct wheat irrigation control model.Finally,the results verify the effectiveness and feasibility of the proposed algorithm.