1 December 2017 in International Journal of Hydrogen Energy
1 August 2018 in International Journal of Hydrogen Energy
1 December 2020 in Energy Conversion and Management
25 August 2026 in Results in Engineering
20 July 2026 in SSRN Electronic Journal
In the macroscopic vision of building a clean, low-carbon, safe, and efficient modern energy system, the application scenarios of distributed energy systems accelerate their penetration from the traditional regional level to the household side. In residential hybrid power scenarios, frequent start-stop of household appliances produces impulsive loads that exceed the dynamic response limitation of mass transfer of fuel cells. This causes severe reactant supply deficiencies and exacerbated concentration polarization, ultimately leading to system output instability and a significant reduction in stack lifetime. This work develops a novel energy management strategy for residential fuel cell-lithium battery hybrid power systems. The proposed strategy deeply integrates the safety boundary constraints of Model Predictive Control (MPC) with the rapid optimization capabilities of the Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm. A multi-objective MPC optimization framework is constructed, with the comprehensive goals of minimizing global equivalent hydrogen consumption and maintaining a stable State of Charge (SOC) for the lithium battery. Traditional nonlinear MPC is limited by excessive online computational latency and poor adaptability to microsecond-level pulse loads. To tackle these issues, we adopt the TD3 deep reinforcement learning algorithm to realize offline approximation and ultra-fast online replacement of the optimal solution criteria for MPC. Results demonstrate that under typical practical residential variable load profiles, compared with the Equivalent Consumption Minimization Strategy (ECMS), fuzzy control strategy, and power-following strategy, the TD3-MPC strategy cuts equivalent hydrogen consumption by approximately 7.57%, 2.89%, and 8.74%, respectively.