Rice Growth Prediction Based On Improved Elman Neural Network

Rice growth prediction is a key part of agricultural precision management.If a prediction model can be established,the corresponding rice growth trend can be predicted according to the input environmental parameters before actual production,and then the final yield can be estimated,which will have a positive significance in enhancing the potential of rice field production and guiding fanning.isRice growth and development a complex process in which a variety and environmental factors work together,so the establishment of its prediction model is also a nonlinear eomplex problem.At present,there are two different ideas in the field of rice growth prediction:one is a crop growth model established by simulating physiological processes such as assimilation of rice and dry matter distribution,which can give the rice yield under the limited conditions of growth factors.The other is based on data analysis of rice growth prediction modeling,mining the hidden relationship between rice yield and tenqjerature,light,water and other environmental factors,and then predict its output,the mainstream methods are statistical models and neural network.By understanding the growth characteristics of rice during each growth period and considering the complex relationship between environmental factors and growth in each growth period,the improved Elman neural network is used to establish a prediction model.The model can predict the growth of different periods in stages and predict the final yield.Based on the study of the growth and development characteristics of rice and the historical data of the recent growth,the paper gives the definition of rice growth in consideration of the growth cycle and key growth indicators in each period,in order to characterize the growth of rice in each growth period,and use Elman neural network to determine each The relationship between environmental factors and growth during the growth period.At the same time,in order to avoid the algorithm easily falling into local optimum,the paper proposes a solution to optimize the initial weight and threshold of Elman neural network with improved genetic algorithm.The six growth cycles of the regreening stage,the tillering stage,the jointing and booting stage,the heading and flowering stage,the filling stage and the maturity stage were modeled separately,and the relationship between the growth amount and the yield of the sixth stage was independently modeled.The training samples were composed of various environmental parameters.And the composition of rice physiological indexes in each stage,the experiment uses several years of historical samples for network training,and the weight of each layer of the model is obtained,and the accuracy of the improved model is improved.Based on the Internet of Things technology,the project team senses environmental data changes through sensors such as temperature,illumination,and water level.The ZigBee module integrates environmental data and transmits it to the aggregation gateway.The GPRS module uploads it to the host computer,and the data is saved to the cloud service data terminal.The Elman neural network establishes a predictive model and completes online training and production forecasting of the network.At present,the system has been handed over to the actual measurement,which can dynamically predict the growth of rice in each growth period,and provide decision support for assisted precision irrigation.