The output characteristic of fiber optic voltage sensor is analyzed. It is found that the response of the voltage sensor is nonlinear and the sensor suffers from the influence of temperature over a wide measuring range. A method for temperature tracking compensation of fiber optic voltage sensor based on artificial neural network is presented in this paper. In order to realize temperature tracking compensation over a full range, the neural network is trained to properly represent the nonlinear mapping between the sensor reading and their represented output accurately at different environmental temperature. It is revealed from the computer simulation result that not only the influence of temperature fluctuation can be eliminated effectively, but also a desired linear relationship between the sensor input and the neural network output can be obtained.