Error Analysis And Compensation Of Heated Twin Radiosonde Humidity Sensor

The heated twin radiosonde humidity sensor,which is one of the hot spots in field of the high altitude meteorological observation,effectively solves the condensation and freeze problems encountered in low temperature upper atmosphere by means of integrating controllable heating chip on the capacitive humidity sensor.However,the employment of the heating function inevitably brings a new disturbance to the measurement and calculation accuracy of the relative humidity.Consequently,it is necessary to make a compensation and correction to the data measured by the heated twin radiosonde capacitive humidity sensor and to make effective control of the heater.In the first part,a method to calculate the actual relative humidity is constructed based on the exploration of the relationship among relative humidity,temperature,dew point and saturation vapor pressure.Three most frequently used computational formulas of the saturation vapor pressure have been used to make a comparison,which demonstrates the Magnus formula has a better result,when it is being used in the relative humidity computational model to calculate the actual relative humidity value after heating,than that of other two formulas of Goff-Gratch and Wexler.In the second part,a modified pi-sigma fuzzy neural network error calibration model base on T-S fuzzy model is established to make the temperature compensation and nonlinear correction to the experimental data of relative humidity.Depending on the K-means clustering algorithm,weight direct determination method and momentum BP learning algorithm,the network parameters are set,which improves the network training speed as well as the calculate accuracy.In the last part,the fuzzy RBF neural network PID controller is optimized.The Jacobian information identification module is implemented in the current controller,which makes the parameters adjusting process relying on the changing rate of the controlled object's output to the controller's input more accurate.MATLAB simulation results illustrate the modified fuzzy RBF neural network PID controller is equipped with the ability to fix the heating model temperature on a given value quickly and steadily,which meets the requirements to the heater control of the heated twin radiosonde humidity measurement.