A reinforcement learning-based Synergistic Hybrid Evolutionary Algorithm for multi-angle shipboard photovoltaic system MPPT under dynamic navigational shading

Shipboard photovoltaic systems, especially with multi-angle installations, face dynamic and complex partial shading conditions arising from navigation and structural obstructions, leading to multi-peak Power-Voltage curves. Existing maximum power point tracking algorithms, due to their fixed search strategies, often struggle to adapt to the continuously drifting global peak on these curves, causing sluggish tracking or local optima entrapment. To address this, this paper proposes the Synergistic Hybrid Evolutionary Algorithm. The proposed algorithm intelligently balances global exploration and local exploitation through a top-level adaptive operator scheduling strategy based on the reinforcement learning “multi-armed bandit” model. This strategy dynamically manages a quadratic interpolation-based particle swarm optimization operator, which fuses a physical model with swarm intelligence for high-precision local convergence, and a genetic algorithm operator for global exploration. Furthermore, its core internal parameters are driven by population diversity, reducing dependency on fixed empirical parameters. Simulations under standard, partial shading, and real-world operating conditions using measured shipboard data demonstrate that the proposed method outperforms the widely used Perturb and Observe method and other mainstream intelligent optimization algorithms. Under real-world dynamic conditions, it achieves a tracking efficiency as high as 99.61 % and improves the cumulative energy yield by 11.76 % compared to the Perturb and Observe method.