Based on historical operational loss data in Chinese commercial banks from year 1994 to 2012，a dynamical model containing different-time correlations is established to describe the mechanism of unexpected loss occurrence，transmission and evolution． The model takes into account the interactions among different event types，the spontaneous generation of losses and the economical capital reservation set by banks to cover expected losses． Through simulations of loss evolution model，unexpected operational risk loss scenarios ( low frequency high severity) and calculation of one-year VaＲ of total unexpected aggregate losses can be achieved． The empirical results demonstrate that the dynamical model is good for description of different-time dependence structures among event types and can pass the robustness test． At a high confidence level，additive sum of VaＲ systematically overestimates total risk and the degree of overestimation will go up with rising of confidence level． Internal fraud has become a main type of operational risk in Chinese commercial banks．