The Analysis And Research On The Data Model Of Micro-current Sensor

In the power equipment,online insulation monitoring technology is an important means to improve the safely and reliability of the power supply. And the leakage current test is a major content in the insulation montoring.To achieve the online monitoring of the leakage current,the design of the sensor is one of the crucial links.Thus,it has the significant theoretical meaning and practical value to devolop a micro-current sensor which has the wide bandwidth,good sensitivity,high accuracy and good linearity.In the first place,the paper introduced the amplitude-frequency characteristics of the micro-current sensor.By selecting a high permeability core material and adding the operational amplifier circuit,not only the bandwidth of the sensor is broadened,but also the sensitivity of the sensor is improved.Then the error of the micro-current sensor had been studied in depth.The article analyzed the soure of the current error-excitation current and respectively discussed the relationship between the primary input current,the secondary coil turns and the excitation current in detail.At the same time,because the selection of the secondary coil turns is blind and there is no theoretical basis to fix the selection,which needs the designers to experiments many times,the text proposed a algorithm to determine the optimal secondary coil turns according to the minimum excitation current principle and proved its correntness,which provided a new way for the design of the current sensor.Secondly,in order to improve the measurement accuracy and linearity of the micro-current sensor in further,the author described two common types of error compensation methods,that is the passive compensation and the active compensation.It also compared their advantages and disadvantages.By contrast,it's known that the latter method is an effective method to obtain the high accuracy.And then,the structural model, learning process together with the standard algorithm of the forward multilayer neural network (BP neural network) was presented,which is widely used nowadays.To improve the slow convergence rate and avoid to falling into the local minimum,the paper used the improved BP algorithm that combined the adaptive learning rate and momentum factor.In addition,use the improved algorithm to compensate the current error,which enhanced the linearity of the sensor.