To improve the precision of gray modeling in forest fire and solve the problem of small date modeling, ER algorithm is proposed. Based on the senior introduced the robust estimation to gray modeling, this method interpolate the modeling date again. The method can achieve small date (3 dates) modeling. This research compared with three calculation methods: least squares method, least squares interpolation method and ER algorithm. According to the fitting precision, least squares method is 10.21%, least squares interpolation method is 1.08% and ER algorithm is 0.00%. That can be obtained by calculating ER algorithm has a good fitting effect.