Article’s

This paper introduces an AI-driven MPPT framework designed to address real-time power optimization challenges caused by PSCs. The system consists of five interconnected modules: contextual hierarchical transfer graph embedding (CHTGE) is used for transfer learning across various environmental conditions through policy graphs based on shading history and weather context. The spatio-temporal feature attention-based indexing (STFAI) module aids in detecting transient phenomena by using attention maps that are temporally synchronized and derived from real-time multimodal sensor data. In the third module, differential contextual residual optimization (DCRO) corrects inaccuracies and achieves rapid stabilization by applying residual corrections in highly variable environments. Outputs from conventional MPPT methods are enhanced with multi-agent decision fusion using quantum-inspired adaptive logic (MADF-QAL). Evolution-based causal disentanglement networks (ECDN) offer fault localization and explainability through latent representation. The proposed framework offers an interpretable, resilient, and intelligent MPPT control suitable for real-world operating conditions.