The identification of the chaotic optical system with BP feedforward neural network supported by a modified BP algorithm is suggested. The feasibility of this suggestion was demonstrated with computer simulation of the identification of the Bragg acoustooptic bistable & chaotic system. The computer simulation result shows that, the three layers foreward neural network, if trained with the chaos speedup BP algorithm, is indeed a fine identifier with less training iterations than usual. Thus it could be used to reconstruct the dynamics of the chaotic optical system with its output time series very efficiently.