In the multi-angle dynamic light scattering for nanoparticle size analysis， the weighted Bayesian inversion algorithm is proved to have a good anti-noise capability. However， it suffers from initial value sensitivity and long time-consuming. This paper presents a method of non-negative least squares constrained weighted Bayesian inversion algorithm， in which the results of the non-negative least squares method are used as the prior value as well as the optimization range of the median diameter and peak width of the weighted Bayesian algorithm. The simulated and experimental results demonstrate that this method can improve significantly the convergence and the anti-noise performance of the unimodal particle system. When there is a big noise， the convergence speed is increased by more than 8 times and the distribution error is guaranteed to be within 0.070 9.