Short-term peak shaving optimization of hydro-wind-solar-storage hybrid system based on pumped storage cluster using soft actor-critic and sequential allocation strategies

To address the increasing penetration of renewable energy and growing peak-valley difference, multiple pumped storage power stations (PSPSs) are often constructed within regional power grid to jointly improve flexibility, forming a pumped storage cluster (PSC) characterized by complex hydraulic and electrical couplings. This coupling relationship considerably increases the complexity of peak shaving, requiring hierarchical optimization from the overall PSC to individual PSPSs and further to units, which leads to a highly time-consuming solution process. Moreover, the uncertainty of renewable energy further intensifies this challenge. To address, this study proposes a short-term joint peak shaving optimization framework for hydro-wind-solar-PSPSs based on the PSC (SPOFHWSP). First, a peak shaving scheme generation strategy at PSC-level using the SAC algorithm is formulated, in which the state space, action space, and reward functions are carefully designed to rapidly generate feasible schemes. Second, a peak shaving sequence allocation strategy at PSPSs-level considering initial conditions is introduced to preferentially allocate larger loads to PSPSs with greater available electricity, thereby reducing solution dimensionality. Finally, a water head loss optimization strategy at units-level is proposed by coordinating the number of operating units and flow distribution in the tunnels to stabilize calculated water head and reduce water head loss. A case study of multi-energy joint peak shaving in regional grid with a large-scale PSC (including six PSPSs) indicates that, compared with the baseline model, the proposed SPOFHWSP reduces the solution time from 5412 s to 794 s, while simultaneously decreasing water head loss in the tunnel by 9.63 m.