Optimized Dispatch for CSP-PV Hybrid Plants in Sandy-Gobi-Desert Regions Considering Frequency-Inertia Coordinated Constraints

[Objective] To address the high penetration and low-inertia characteristics arising from the grid integration of large-scale renewable energy bases in “sandy-gobi-desert” regions， as well as the limitations of traditional dispatch methods that primarily focus on steady-state power balance and economic operation with insufficient attention to frequency security risks， this paper proposes an optimal dispatch method for concentrated solar power-photovoltaic hybrid power plants considering generalized inertia and rate of change of frequency hard constraints. [Methods] First， the optimal dispatch of concentrated solar power-photovoltaic hybrid power plants is modeled as a markov decision process. This model incorporates dynamic security constraints， such as the rate of change of frequency and generalized inertia， to achieve strict control over grid frequency security boundaries. Second， to efficiently solve this highly nonlinear scheduling model and overcome the limitations of the conventional twin delayed deep deterministic policy gradient （TD3） algorithm—specifically addressing the low utilization efficiency of historical information and the insufficient learning of critical disturbance samples—an improved TD3 algorithm integrating long short-term memory （LSTM） networks and a prioritized experience replay （PER） mechanism is proposed. [Results] Multi-scenario simulation results based on modified IEEE-30 and IEEE-57 bus systems demonstrate that， under the complete frequency-inertia constrained scenario， the proposed LSTM-PER-TD3 algorithm achieves zero frequency-limit violations and zero reserve shortage， with a total operating cost only 2.94% above the mixed integer linear programming （MILP） theoretical optimum. [Conclusions] Concentrating solar power plants possess dual advantages of thermal energy storage for time-shifting regulation and physical inertia support from synchronous units， which can firmly guarantee frequency security for high-penetration new energy bases. The developed LSTM-PER-TD3 optimization algorithm realizes efficient unit scheduling while balancing operational safety and economy. It provides theoretical foundations and technical support for active frequency support and intelligent economic scheduling of power systems with high-penetration new energy integration.