To study the physical stability of insulin in drug delivery particles, we developed a coarse-grained (CG) model for insulin based on dissipative particle dynamics (DPD). Three insulin modelling schemes were considered: each amino acid as a bead (IM1), each amino acid being separated into one to three beads (IM2), and adding secondary structural information of insulin to IM2 (IM3). The best possible bead-bead interaction parameters were obtained from Hildebrand and Hansen solubility parameters by performing the constant-temperature DPD simulation with insulin models in 20% acetic acid solution. IM3 showed good results in terms of RMSF, RMSD and A1B30 distance compared to those of all-atom models from the literature. Then, the IM3 model was considered in an oil-filled poly (isobutyl cyanoacrylate) (PIBCA) nanocapsule. Two crucial factors were found that mainly influence the stability of insulin in oil: the PIBCA shell thickness and the amount of ethanol in the oil droplet. An appropriate PIBCA shell thickness is necessary to block the interaction between insulin and water outside, and ethanol could stabilise insulin with its good affinity for both insulin and oil.