In this paper, a control-oriented soot model was developed for the use of soot real-time prediction and combustion condition optimization. As a promising combustion concept, Partially Premixed Combustion (PPC) achieves high engine efficiency, and reduces soot and NOx simultaneously. However, soot emission was found to be significantly increased with high load, high EGR and split injection. In order to investigate factors that influence soot emissions on a multi-cylinder heavy duty gasoline PPC engine, sensitivity analysis upon EGR rate, injection profile, intake pressure and intake temperature were mainly studied. An empirical model was modified based on the original Hiroyasu model according to the sensitivity results. By introducing pilot ratio as a compensation factor, This model can be used to predict soot emission under double injection. 7 model parameters were identified using experiment data under a few representative operating points. In addition to focus on the engine-out soot emission value, this model is also able to describe the whole trajectory of soot formation and oxidation. This model was validated under load and speed transient operating conditions and agreed with the trend of measurement result, which is sufficient enough for soot prediction as a constraint on further engine efficiency control.