Improving the performance of adjustable speed pumped storage hydropower plants through hybridization with PV and with deep reinforcement learning-based governor to improve power network resiliency.

With the increasing penetration of renewable energy sources (RESs) into the power system, the reduction of the system’s total inertia has become a growing interest, along with the system’s intermittency and unpredictability and conserving the system’s resiliency under the reduced inertia condition. To tackle these concerns, add resiliency, and enhance the system’s readiness to meet load and fluctuating RESs, grid-scale energy storage systems (ESSs) such as pumped-storage hydropower (PSH) are being deployed. By far, PSH is the most mature and long-term economically viable grid-scale ESS. But to improve the various ancillary services it can provide, PSH technology configuration has evolved and has led to adjustable speed, ternary, and quaternary PSH plants. The adjustable speed pumped-storage hydropower (ASPSH) is the most widely used among these configurations because of its wide range of operation and its ease of construction from already existing conventional PSH units. First, to improve the ancillary service ASPSH can provide to the power network through hybridization, a hybrid system comprising an ASPSH and a PV unit together with their controllers was constructed and implemented in MATLAB/Simulink. To seamlessly coordinate the response of these individual units, a hybrid plant controller was formulated and a single-machine-infinite-bus (SMIB) integration and a 9-bus system integration tests were performed. The designed hybrid controller was able to successfully track power reference in both generation and pump modes, and demonstrated an improved frequency regulation in the event of a disturbance, while integrated into a 9-bus test system. The potential electricity cost savings by the designed hybrid ASPSH was evaluated and compared with that of a standalone ASPSH by employing mixed integer linear programming (MILP) to obtain their optimal operation schedule, and it was concluded that the hybrid ASPSH results in more savings. Finally, to tackle some drawbacks of the conventional proportional, integral, and derivate (PID) controller of an ASPSH governor, such as the need to retune its parameters to optimal values for different operation conditions, a deep reinforcement learning approach was used to design and implement an intelligent controller capable of observing the system conditions and track the optimal reference speed. An integration study in the 9-bus test system showed that the designed intelligent controller exhibits better primary frequency regulation compared to the PID controller.