Research Of Temperature Data Correction Algorithm In Sensor Networks For Meteorological Observation

Wireless sensor networks have the characteristics of popularity,ease of use,low price,and are increasingly used in many industries and scientific research.At present,various weather meteorological elements are mainly monitored through automatic weather stations in China.However,high-density deployment is often not possible because the construction cost is too high.The use of wireless sensor networks to monitor the surface air temperature can greatly reduce the monitoring cost,and it has the characteristics of simple deployment and strong mobility.However,in this scheme,the inexpensive sensor and shell material used are not professional enough.The sensor node is affected by the weather such as solar radiation,precipitation,precipitation,etc.while sensing the temperature,resulting in the sensed value deviating from the actual data.And some abnormal values will be generated during the operation of the node.Therefore,this paper proposes an error correction scheme based on BP(Back Propagation)neural network and an outlier correction scheme based on sliding window.The research contents and innovation results of this paper are as follows:(1)The monitoring data of a single node in a meteorological wireless sensor network are preprocessed and analyzed.It is found that the most powerful factor affecting the air temperature data quality is the solar radiation intensity by comparing with the standard temperature.Then a BP neural network model is designed,which takes solar radiation intensity as input parameter and the error between node temperature and standard temperature as output parameter.In the correction phase,the values of solar radiation intensity in the test data are used as input to get the values that need to be corrected at each time node.Then the correction value is subtracted from the actual sensing value,and the modified data is obtained.The experimental results show that the outliers can be greatly corrected by this method.(2)On the basis of BP neural network correction,in order to detect outliers in temperature time series data and improve data quality and decision quality,a new outlier detection algorithm based on sliding window prediction is proposed.In this method,the time series are segmented by sliding window,and then the prediction model is established according to the historical data,which is used to predict the future value.It is assumed that the observed values are considered to be abnormal and need to be modified beyond the given predictive confidence intervals,where the confidence intervals can be calculated by the predicted values and the confidence coefficients.This paper discusses the sliding window size and parameter setting of the algorithm,and verifies the algorithm with actual data.The experimental results show that this algorithm can not only effectively detect outliers in time series of meteorological data,but also improve the efficiency of correction.