Research On Temperature Compensation Technology Of Optical Fiber Current Sensing System Based On FBG

With the continuous growth of voltage and transmission capacity and the development trend of digital power grid,the traditional electromagnetic current transformer has been unable to meet the development demand of large power grid,large capacity and digitalization.The optical current transformer has a series of unparalleled advantages of a series of traditional electromagnetic current transformers,which can meet the needs of the future power grid more.With the rapid development of optical fiber Prague grating(FBG)and the successful development of giant magnetostrictive materials,the optical fiber current sensors are outstanding in many optical current sensors because of their superior performance.However,the FBG current sensor is sensitive to temperature and strain,and it is necessary to remove the influence of temperature in the design.In this paper,the research of temperature compensation in FBG based current sensing system is focused on building a current sensing system and compensating the temperature.The main contents are as follows:(1)The relationship between the strain and temperature and the central wavelength of the FBG reflected wave is obtained by analyzing the sensing principle and sensing characteristics of FBG.The FBG current sensing system is set up.In the experiment,the 72 groups of experimental data were measured by two variables,which were controlled by current and ambient temperature.The paper analyses the data.(2)The BP neural network is used to compensate the temperature of the measured data,and the accuracy of the current measurement is obtained by simulation.The normalization and K fold cross validation is added to the neural network algorithm and the measured data is input into the trained BP network model.The average mean square error of the 9 fold cross validation is 0.005.(3)Using particle swarm optimization(PSO)algorithm and genetic algorithm(GA)optimize the weights and threshold of BP neural network.Then,the algorithms are simulated by MATLAB,the mean square error of cross validation PSO-BP algorithm is 0.0045,and the mean square error of cross validation GA-BP algorithm is 0.0038.The measurement errors of the two optimization algorithms are all reduced,and the GA-BP algorithm has the best optimization effect,the highest stability and the least prediction error.