To both enhance the flexibility of the power system and absorption capacity of renewable energy connected to a grid and achieve the efficient use and cooperation of multi-side complex type energy storage resources on the source, grid, and load sides, a collaborative optimal scheduling system architecture of source-grid-load-storage (SGLS) considering multiple energy storage types was constructed. Latin hypercube sampling and sample reduction based on k-medoids clustering were adopted to generate the SGLS scenarios, considering the emergency circumstances of the power system. The flexibility of the multi-scenario power system was evaluated, and the battery energy storage stations, pumped storage, and electric vehicles with sufficient capacity were configured on the source, grid, and load sides, respectively, to participate in the scheduling. A multi-scenario SGLS cooperative optimisation scheduling model that considers multiple energy storage capacity configuration types was constructed for economic and environmental protection. Based on data-driven, a multi-objective optimisation algorithm was proposed by using the Gaussian process regression algorithm and non-dominated sorting genetic algorithm II, combined with the manifold interpolation batch evolution mechanism. Finally, using actual regional power grid data for verification, the proposed strategy effectively reduces system operating cost, enhances energy storage battery life, and improves the renewable energy consumption capacity.