In order to overcome the influence of self-absorption on quantitative analysis, the optimizing process of very fast simulated annealing algorithm was studied. According to basic theory of plasma emission spectrum, a new algorithm for self-absorption correction based on multi-particles spectra was proposed, and the algorithm flowchart was given. With the self-absorption correction algorithm mentioned above, the spectra of refining slag and blast furnace slag were corrected. The effect of self-absorption correction on the quantitative analysis results was analyzed based on calibration free method. Comparison of Boltzmann plots before and after self-absorption correction indicated that the plasma temperatures calculated with spectra after self-absorption correction tended to be uniform, and remained stable around 11 600 K. The Boltzmann plots constructed with plasma spectra of the same particle after self-absorption correction indicated that the intercepts were almost the same except for one group data. With calibration free method and spectra after self-absorption correction, the contents of components in slag were analyzed. For refining slag, quantitative analysis precision of MgO was low. If ignoring the existence of MgO, the relative errors of quantitative analysis results of CaO, Al2O3 and SiO2 were much smaller. For blast furnace slag in which the content of MgO was 8.49%, the relative error of quantitative analysis result of Al2O3 was 2.38%, which was the smallest. And the relative error of quantitative analysis result of MgO was 28.27%, which was still the biggest.