A real-coded genetic algorithm (RGA) for fiber Bragg grating (FBG) distributed sensing is presented. The distributed strain fields along the fiber Bragg grating sensor (FBGS) are real coded into genes, and the concept of elitism and simulated annealing are also included in this algorithm. Compared with the binary coded genetic algorithm, this method is more simple and efficient. Only with the reflect spectrum of distributed FBGS, the strain fields distribution can be exactly demodulated even in the regions with significant strain gradients. The algorithm is a promising method for demodulating the distributed FBGS, which can be used for structural failure analysis and structural damage identification.