High-penetration photovoltaic (PV) integration into distribution grids presents significant challenges, including diverse equipment characteristics, voltage instability induced by PV fluctuations, and the trade-off between economic and safety objectives. To address these issues, this paper proposes a two-stage, multi-timescale collaborative optimization framework. The day-ahead stage optimizes the discrete states of on-load tap changers and capacitor banks to minimize daily power loss, voltage deviation, and equipment operating costs. The intraday stage leverages PV inverters, static var compensators, and energy storage systems to mitigate real-time PV fluctuations and maintain voltage stability. Furthermore, an Improved Crocodile Ambush Optimization Algorithm (ICAOA) is developed, incorporating a Sobol sequence for population initialization, adaptive parameter adjustment, and a hybrid perturbation strategy to enhance its search capability and convergence performance. Validation results from CEC2022 benchmark functions and IEEE 33 system simulations demonstrate the effectiveness of the proposed approach: the day-ahead optimization achieves a daily loss of 1.02 MWh and an average voltage deviation of 0.014 p.u.; the intraday correction reduces the rolling-horizon loss to 0.032 MWh with zero voltage violations. The framework and algorithm collectively improve the economic and stable operation of high-penetration PV distribution networks.