The stochastic parallel gradient descent (SPGD) algorithm is one of the most efficient methods for enabling the precise phase control of laser coherent synthesis systems. However, the control bandwidth of the SPGD algorithm rapidly decreases with an increase in the number of laser paths involved in the synthesis. To meet the high bandwidth requirements of the phase control in large-scale laser coherent synthesis systems, this study proposes an SPGD optimization algorithm based on the adaptive random disturbance voltage,that is, Piecewise SPGD algorithm. The proposed algorithm uses a variable random disturbance voltage to segment the gain coefficient. The experimental simulations of coherent synthesis systems with different laser paths are conducted using the proposed algorithm. The results show that in experiments with seven laser paths, the convergence speed and stability of the optimized algorithm are improved by 36.4% and 70.6% compared with those of the traditional SPGD algorithm, respectively, and the performance is significantly improved. The proposed Piecewise SPGD algorithm can be extended to other laser coherent synthesis systems without parameter adjustment for strong universality and good performance of the phase control.