Abstract: Multi-energy hybrid is an important way to promote clean energy grid-connected consumption and improve power regulation. The scale of multi-source access exacerbates the coexistence of decision and risk. The paper aims to optimize the searching dimension and coordinate the source-load interaction to cope with the energy consumption and peak load regulation, ensuring the reliability of the power plant and unit operation. Therefore, a hierarchical collaborative searching strategy of hydro-PV-load module is proposed. In the outer layer of the strategy, the robust feature of the PV-load prediction deviation is proposed. The feature parameters retrieval is carried out according to the scene sets, and the confidence intervals are obtained accounting for the multiple uncertainties in the PV-load module. In the middle layer, the reconstruction of the hydro-PV module is realized in combination with the source-load regulation ability to support the peak load regulation goal of the system. In the inner layer, the power quantity and fluctuation features in the middle-layer power reconstruction is introduced to identify the operating risk intervals of the unit operation. The risks of the unit crossing the vibration zone are reduced by gradually optimizing the unit arrangement and the load transfer. In this way, the dynamic cooperative searching strategy for the inter-layer target tracking and the intra-layer risk control is established to form the power generation decision space under the PV-load uncertainty. Finally, the mixed power generation system in the Tibet section of the Lantsang River is taken as the research objective. The results show that compared with the traditional method, this strategy effectively reduces the operation risks of the system under the alternating fluctuation of the PV and load responding to the peak shaving target.