Partial shading conditions (PSC) significantly reduce the efficiency of photovoltaic (PV) systems by causing uneven irradiation and mismatched power losses. To address this

this study proposes a novel momentum vectorized adaptive deep deterministic policy gradient(MVA-ADDPG) algorithm for hybrid PV-thermoelectric generation (PV-TEG) systems. The PV-TEG system integrates thermoelectric generators with PV modules to capture waste heat and uses intelligent energy storage coordination to reduce temperature sensitivity and improve system stability. Unlike conventional PV-energy storage systems

which suffer from high energy losses and maintenance costs

the proposed MVA-ADDPG-driven PV-TEG system employs a triple-action heuristic exploration strategy. It combines momentumaccelerated policy gradients with dynamic exploration-exploitation balance. At each step

three candidate actions are evaluated

generated through both heuristic and gradient-based approaches.

This enables fine-grained optimization of battery distribution and system performance.Experimental validation on 6 4 to 6 6 PV-TEG arrays show an average power increase of 26.5%and a mismatch loss reduction of 45.2%. The method achieves fast convergence and maintains reliable performance under varying shading conditions. By recovering waste heat and optimizing cell compensation

the proposed approach extends system lifespan

and enhances economic viability. It offers a robust solution for efficient energy management in complex PV environments.