Research On Grain Pile Temperature Prediction Model Based On Improved BP Neural Network

Food security is very important to maintain social stability,and scientific grain storage is the necessary means to ensure food security.Because the grain in the granary is stored directly in the granary without any processing and packaging,if the temperature of the granary is too high,it will directly lead to the occurrence of mildew,germination,breeding microorganisms and so on.Therefore,the key to scientific grain storage is that the warehouse managers need to grasp the temperature change of the grain stack in time,because the temperature change of the grain stack is an important sign of the change of grain quality,and then take ventilation and antimildew measures and other safety management measures according to the variety of temperature changes,in order to ensure better storage of grain in the grain depot,so establish and improve the grain depot.It is necessary to establish a grain heap temperature prediction system.In order to realize the accurate prediction of grain heap temperature,the following research work has been done in this paper.After analyzing the ecological environment of grain heap temperature change,and according to the complexity and uncertainty of temperature data change,this paper uses the combination algorithm model based on BP neural network algorithm to achieve short-term accurate prediction of grain heap temperature.The first combination method combines particle swarm optimization algorithm with BP neural network algorithm.Because the initial weights of the neural network are obtained by initialization,and then optimized by iteration in the training process,but we don't know that the weights and biases of the neural network obtained after that iteration are optimal,so we propose the first combination method,using the first method.Particle swarm optimization(PSO)optimizes the weight and bias of BP neural network algorithm,and then predicts the grain heap temperature.The second combination method combines the component analysis method with BP neural network algorithm.Because the original data has some noise,and some independent variables have little or no influence on dependent variables,the principal component analysis method is used to denoise the data first,and then the neural network is used to analyze and predict.Experiments show that the combination method of particle swarm optimization and BP neural network algorithm has higher accuracy and stronger generalization ability than single prediction model.At the same time,the combination method of principal component analysis and BP neural network algorithm has a higher accuracy in predicting results.Compared with the results of the above combination algorithm,the accuracy of this combination algorithm is higher than that of the above combination algorithm,and the mean square error of the prediction results is reduced by three times.These methods can realize grain heap temperature prediction.The measurement provides a new technical scheme.