Research On Spatial Interpolation Of Grain Storage Temperature Field And Lost Data Interpolation Algorithm

Grain is the important strategic reserve material in our country,and the grain reserve is a key link in ensuring food security.In the process of grain storage,temperature and relative humidity are the most important environmental factors in food safety and quality.Therefore,it is necessary to analyze the temperature and humidity sensor data of grain bulk to ensure the food safety.With the popularization and application of information technology in my country’s grain depots,multi-dimensional grain situation analysis techniques is gradually being applied to the actual management of grain bulk.It is possible to obtain more accurate and detailed grain conditions based on temperature and humidity multi-dimensional data.In this case,the accuracy and completeness of the temperature and humidity data is the prerequisite for the multi-dimensional grain situation data analysis.However,the time series data of temperature and humidity sensor nodes are often missing due to the complex grain storage environment and sensor node failures in the grain depot.Incomplete temperature and humidity time series data will not be conducive to the reliability of grain monitoring.Therefore,this thesis has carried out the research on the prediction of grain bulk temperature,and the interpolation algorithm research on the missing time series data of temperature and humidity sensor equipment.The main research work completed in this paper is as follows:1.Through the time autocorrelation analysis method and the space Moran index analysis method to analyze the time and space correlation of the grain temperature time series,as the theoretical basis of the grain temperature prediction model,and find the appropriate time sliding window length.And dividing the time series data of each temperature sensor node of the actual grain bulk through the sliding the window method.2.In this paper,we proposed a grain temperature prediction model based on the historical temperatur sensor-nodes’ data from the grain bulk,considering the sensor-nodes’timeless and spatiality together.This model can predict the temperature value of each node in the future,and get the temperature values of the unknown nodes,so as to obtain the temperature nephogram of grain bulk in the future.3.The grain bulk temperature time-space prediction model can predict the temperature of nodes at unknown locations,and provide macro guidance for the grain situation monitoring of the grain depot.4.Aiming at the problem of temperature and humidity sensor data missing,in this paper we proposed a model based on the multi-modal data fusion autoencoder to interpolate the missing data of the temperature and humidity of the sensor node.Comparing with the auto-encoder data interpolation model based on single mode,this model has improved the accuracy in interpolating the missing data.And this paper adds association learning to improve the model.And the interpolation effect is also improved comparing with the modal fusion method that only uses feature splicing.In this paper,we propose two models to solve the problems of grain bulk temperature prediction and data loss interpolation.It enables the storage personnel of the grain depot to predict the occurrence of pests and mildew abnormalities in advance,and take preventive measures in time to reduce grain loss and ensure grain quality.And it provides effective data support for the analysis and research of grain condition data based on multi-field coupling,and helps to improve the completeness and multi-dimensionality of grain condition monitoring information.