This research proposes to use Artificial Bee Colony algorithm to realize the inversion of the multimodal distribution of the particle size of small angle forward scattering method based on Mie scattering theory. The inversions of uniform spherical particle systems which obeying normal distribution， Rosin-Rammler distribution or Johnson's S_B distribution function were performed， and the particle group issimulated as unimodal， bimodal and trimodal distribution respectively. The particle size can be well inverted by the Artificial Bee Colony algorithm. In the case of unimodal distribution， the relative root mean square error of the particle weight frequency distribution curve can be as low as 3.53×10^-8. The relative root mean square error of the particle weight frequency distribution curve of the wide bimodal distribution decreased from 3.38% and 2.70% to 1.53% when it's compared with the independent mode algorithm Philip-Twomey-NNLS and Chahine. The accuracy of the Artificial Bee Colony algorithm is higher， and as the number of peaks increases， the width of the distribution curve becomes narrower and the noise increases， the errors of the Philip-Twomey-NNLS algorithm and the Chahine algorithm increase to 44.99% and 24.36%， respectively， while the error of the Artificial Bee Colony algorithm is 18.22%. A particle measurement system based on small angle forward scattering method was constructed， and the scattering images of 35 μm particle group and 30μm and 51μm mixed particle group of national standard particles were collected for experimental study. The relative error of the characteristic particle diameter obtained by Artificial Bee Colony algorithm is within 5%， which is about 50% lower than that of Philip-Twomey-NNLS algorithm.