Nonlinearity Compensation of Magneto-Optic Fiber Current Sensors Based on WOA-BP Neural Network

Compensating the temperature dependency and the slight nonlinearity of the magneto-optic response of a fiber optic current sensor (FOCS) are crucial to improve its accuracy. In this article, the nonlinearity originations of the FOCS with the static-biasing scheme for temperature compensation are analyzed and characterized. We found that the temperature dependency of the static bias and the varying magnetic domains of the Faraday crystal are the two main sources of the nonlinearity. A back-propagation neural network (BPNN) aided by the whale optimization algorithm (WOA) is developed to compensate for the temperature dependency and the nonlinear response of the fiber current sensor. Experimental results show that in the temperature range of −20 °C–60 °C and measurement range of 0 to 27 mT, the maximum relative error can be reduced to 0.38% with the proposed method, whereas without compensation the relative error is as large as 16.0%. The WOA-BPNN is also compared with the nonoptimized BPNN and the PSO-BPNN, the maximum relative errors of the last two methods are 4.04% and 0.99%, respectively, with the same set of experimental data. The WOA-BPNN enables the sensor to fulfill the accuracy requirement of the metering standard.