Research On BP Neural Network Interpolation Algorithm In Meteorological Sensor Network

Meteorological sensor network is a distributed network which is composed of a large number of operations has function and communication function of sensor nodes,it is used to acquire, collect, and transmit meteorological data for monitoring targets in a node deployment area. As the meteorological sensor network is mostly deployed in the harsh environment of the monitoring area, the nodes are mostly randomly distributed, resulting in overlapping monitoring areas or missing monitoring areas, so it is necessary to adopt interpolation method to compensate, complete the monitoring data. The back-propagation neural network can be used to interpolate meteorological sensor network data because of its infinite approximation principle. This paper combines the characteristics of meteorological sensor network with the related routing protocols, proposes two data interpolation methods based on improved BP neural network. Specific research results are as follows:(1) BP neural network interpolation algorithm based on improved genetic algorithm. Methods The LEACH routing protocol and the improved BP neural network interpolation algorithm are used to predict the data collected by the meteorological sensor network. The improved BP neural network algorithm is carried out by optimizing the genetic algorithm, which mainly includes improving the network model of the BP neural network, determining the optimal hidden layer node number, and determining the optimal weight and threshold parameters of the BP neural network to make it more accurate interpolation results. Experiments show that the data interpolation algorithm based on improved BP neural network can achieve more accurate interpolation results than the commonly used data interpolation method.(2) Based on momentum - adaptive BP neural network interpolation optimization algorithm. It is necessary to control the data traffic and increase the service life of the nodes when the nodes collected in the meteorological network may contain large amounts of redundant or invalid data. In this chapter, TEEN protocol is used as the routing protocol of meteorological sensor network, and its idea is integrated into the BP neural network model with momentum and adaptive improvement to generate efficient fusion data interpolation model. The experimental results show that the proposed method not only improves the efficiency of the interpolation model, but also improves the interpolation accuracy.