Deep reinforcement learning for photovoltaic performance ratio estimation and accurate degradation stage diagnosis based on precise RS-RP evaluation

• Advanced DRL algorithms achieve 100 % accuracy in detecting PV degradation conditions. • Precise estimation of R S-DEG , R P-DEG , and PR with MAE values below 0.0035 and R² of 99.9 %. • A framework for early degradation detection enables corrective and resilient PV system design. • DRL-based models significantly advance PV technology and maintain solar PV efficiency. Photovoltaic (PV) modules are susceptible to degradation over time, which primarily manifests as changes in series resistance (RS) and parallel resistance (RP). Such degradation leads to significant losses in energy output and efficiency. This paper analyzes various PV array I–V curves under degradation conditions by employing advanced deep reinforcement learning (DRL) algorithms, namely Deep Q-Network (DQN), and soft actor-critic (SAC), to (i) detect and distinguish degradation conditions using a binary classification model and (ii) estimate RS, RP, and the performance ratio (PR) through a multi-output regression model. The outcomes of the study are twofold: Firstly, it enhances the understanding of degradation patterns in PV modules, and it provides a robust framework for predicting and preempting performance declines. Secondly, accurate determination of RS and RP values not only facilitates immediate corrective measures but also informs the design of more resilient PV systems. As such, this paper contributes significantly to the advancement of PV technology, helping ensure that solar energy remains a reliable and efficient component of the renewable energy portfolio. The proposed DRL-based model has achieved remarkable results, detecting degradation conditions with an accuracy of 100 % and estimating the values of RS-DEG, RP-DEG, and PR with mean absolute error (MAE) values of 0.0024, 0.0034, and 0.0023, respectively and R-squared values of 99.9 % for all three variables, underscoring the precision and reliability of the approach in addressing PV module degradation.