将一种改进的变增益系数自适应随机并行梯度下降(SPGD)控制算法应用到大阵列光纤激光相干合成中,计算不同增益系数对算法收敛速度的影响程度,分析判断该方法的控制带宽、控制时间与合成光束质量、合成路数的关系以及其应用于大规模阵列相干合成的可行性。计算结果显示,在7束光纤激光相干合成中,该方法由于采用了变增益系数的控制策略,相比于传统的固定增益系数SPGD算法,具有收敛速度快、控制带宽高、适用于多组束光纤激光相干合成等优点。将该方法应用在37束、91束和100束光纤激光阵列锁相中,也得到了快速的收敛效果,采用自适应SPGD算法分别将收敛速率提高了37.8%、63.8%和75.0%,说明该方法在合成路数较大时优势更加明显,进一步表明其具备向大阵列光束相干合成扩展的潜力。 An improved adaptive parallel gradient descent (SPGD) algorithm with variable gain coefficient was applied to large array fiber laser coherent synthesis to calculate the influence of different gain coefficients on the convergence rate of the algorithm. The control bandwidth of the method was analyzed and the control Time and synthetic beam quality, the relationship between the number of synthetic and its application to the feasibility of large-scale array coherent synthesis. The calculation results show that the proposed method has the advantages of fast convergence rate and high control bandwidth compared with the traditional fixed gain coefficient SPGD algorithm, because of adopting the variable gain coefficient control strategy in seven fiber laser coherent synthesis. Fiber laser coherent synthesis and so on. The method was also applied to the phase locking of 37, 91 and 100 fiber laser arrays. The fast convergence was also achieved. The convergence rate was improved by 37.8%, 63.8% and 75.0% respectively using adaptive SPGD algorithm, The method is more advantageous when the synthesis path is larger, further indicating that it has the potential to expand to coherent synthesis with a large array of beams.