Grain starch is a key index of quality for rice (Oryza sativa L.). This study developed a simulation model on grain starch formation in rice by analyzing the dynamic patterns of carbon assimilation and translocation in rice plant under varied environmental factors and genetic types based on a rice model RiceGrow. The model proposed that the rate of grain starch accumulation was determined by carbon availability restricted by source and carbon accumulation rate restricted by sink. Carbon accumulation rate restricted by sink was dependent on a potential starch accumulation rate and interaction of influencing factors as temperature, water, nitrogen conditions within plants and ability of carbon translation into starch. Carbon availability of grain restricted by source was sum of carbon assimilation from photosynthetic organs and remobilization from vegetative organs after anthesis. Photosynthetic product transported to grain directly after anthesis exhibited a logarithmic relationship to thermal time after post-anthesis. Post-anthesis carbon remobilization from vegetative organs included remobilization from leaves and stems. Testing of the model with independent datasets involving different years, eco-sites, varieties and nitrogen rates indicated that the values of RMSE were 11.03% and 13.82% for starch content and accumulation, respectively. The results showed that the established model could predict content and accumulation of grain starch in rice under different cultural conditions, which would provide a quantitative tool for quality prediction and regulation in rice.