The low velocity impacts (LVIs) monitoring based on optical fiber Bragg grating (FBG) sensors have attracted more attention in recent years. The center wavelength migrations of FBG sensors were determined by strain and residual strain during and after LVI on composite material structure. We presented a method to discriminate the energy characters of LVI response signals related to LVI locations. By analyzing the wavelet packet energy spectra of LVI response signals monitored by FBG sensors, the sixth node's energy was found to be sensitive to LVI location. Thus, the sixth node's energies as LVI feature values, were used to predict the LVI locations by the method of support vector regression (SVR). By optimization of the SVR models' free parameters, predicting accuracy was 4.62% in the work.