The Study On Dual Heated Altitude Humidity Sensor Data Modeling And Error Correction Technology

High altitude meteorological observation is important for weather forecast and the analysis of climate change,the conduct of national disaster prevention and mitigation,the meteorological research at home and abroad.The sensors are used to measure the various meteorological elements in high altitude meteorological observation.Due to measurement environment is complex and changeable,high altitude humidity measurement has always been the key and difficult point for meteorological observation in China.This article describes the structure and characteristics of high polymer humidity sensitive capacitor sensor which is suitable for the radiosonde.In the field of high altitude meteorological observation,sensor is susceptible to frost,ice and other pollution in the low temperature,this can lead to the decrease of detection precision.For this problem,the article discusses the humidity measure method that using the humicap with heating function,and specifically focuses on the principle of dual heated humidity measurement,using two humicap take turns to heating and measuring,which realized humidity measurement and dehumidification simultaneously.Temperature is the main factor lead to humidity measurement error.In order to reduce the impact of temperature on the humidity and improve the measurement accuracy of the humidity sensor,we must implement temperature compensation for the humidity sensor.This article puts forward the method that utilize support vector machine(SVM)for the humidity sensor temperature compensation.Aiming at the shortcomings of the parameters of SVM is difficult to select,improved the genetic algorithm(GA)and particle swarm optimization(PSO)algorithm,and combined them with SVM.This method can be used to build the model of temperature compensation for humicap,which can not only improve the measurement accuracy of humicap,but also provides an effective way to parameter selection of SVM.This article use Matlab language to realize humicap temperature compensation of SVM,compared the result of improved genetic SVM(GA-SVM),improved pso SVM(AMPSO-SVM)and standard pso SVM(PSO-SVM),BP neural network methods.The superiority of the proposed method was verified.Experimental results show that the GA-SVM effectively compensate the temperature influence,enhances the measurement accuracy of humicap,the approximation ability and generalization ability of the algorithm are superior to the BP neural network,but the convergence speed is slow.The AMPSO-SVM is effective to avoid the premature convergence of PSO-SVM and slow convergence speed of GA-SVM,can better realize the humidity sensor temperature compensation.In conclusion,SVM method can overcome the problem of the neural network that easy to fall into local minimum,more suitable for humicap temperature compensation;The AMPSO-SVM has the advantages of strong anti-precocious ability,high search precision and fast convergence rate compared to other methods.Therefore,it is the most effective method for humicap temperature compensation.The last,using graphical user interface functions of Matlab established SVM temperature compensation software of humidity sensor.Users do not need to open the background program,all algorithm and their parameters can be changed on the interface,and the compensation result can be observed in real time,which make it has very strong human-computer interaction.