Fiber bragg grating sensor has been widely used in the strain measuring at present，but because the cross influence of strain and temperature，the sensor measurement accuracy is limited certainly, and it can’t meet the practical requirement of industrial. In this paper, temperature compensation of FBG sensor experimental data by BP neural network method，in order to reduce influence by temperature. in the whole temperature range，the relative maximum error of the measured can be reduced by using BP neural network method. Error is controlled in range . it has important meaning to the useful of sensor.