The complementary operation of cascade hydropower (CHP) and photovoltaic (PV) can increase the integration of PV power and has become a trend in modern power systems. However, the randomness and variability of PV power output might cause challenges in precisely tracking the submitted generation plan of the CHP-PV complementary system. The safety and stability of the receiving-end power grid might deteriorate. In this paper, a multi-time scale dispatch framework with day-ahead and intra-day dispatch is established for the CHP-PV complementary system to address the PV uncertainties. The day-ahead dispatch is employed to provide reliable reference results and sufficient adjustment reservations for the intra-day dispatch. An Itô-theory-based stochastic optimization (ITB-SO) approach is proposed for the intra-day dispatch by modeling the PV prediction error with stochastic differential equations (SDE). The SDE model can be embedded into the ITB-SO approach without scenario generation, and thus computational burden can be significantly reduced compared with scenario-based methods. A complementary system in southwest China is chosen as a detailed case study. The simulation results demonstrate that real-time random PV fluctuations could be compensated by adjusting CHPs. Meanwhile, the target water level of cascade reservoirs can be tracked under the proposed approach.