In the context of green port construction, this study proposes an intelligent dispatch method for onshore power supply systems (OPS) that accounts for source-load uncertainties to enhance economic efficiency and improve the accommoda-tion of distributed photovoltaic (PV) generation. To address the interference of complex port environments on PV output, a physical correction model is established by considering factors such as dynamic shading from gantry cranes, salt-fog attenuation, and sea-surface reflection. On this basis, infor-mation-gap decision theory (IGDT) is introduced, and a time-varying uncertainty radius is employed to characterize the uncertainty envelope of PV forecasting deviations. Meanwhile, the OPS load is decomposed by integrating prior information from vessel berthing schedules, and a time-varying disturbance intensity is adopted to describe stochastic fluctuations on the demand side. Subsequently, an energy management model aimed at minimizing total operat-ing cost is formulated, and the deep deterministic policy gradient (DDPG) algorithm is utilized to achieve online opti-mization of continuous decision variables, including grid power purchase, PV accommodation, and charg-ing/discharging of energy storage. Simulation results indicate that the proposed method can effectively improve the economy and robustness of the port OPS while ensuring reliable operation.