IOT-DRIVEN IMPLEMENTATION OF AI PREDICTIVE MODELS FOR REAL-TIME PERFORMANCE ENHANCEMENT OF PEROVSKITE AND TANDEM PHOTOVOLTAIC SYSTEMS

Abstract

This systematic review synthesizes how Internet of Things infrastructures and artificial intelligence predictive models enhance real-time operation of perovskite and tandem photovoltaic systems. Following a prospectively registered protocol and the PRISMA framework, we searched major scholarly databases, screened records with two independent reviewers, and extracted commensurate metrics for quantitative aggregation alongside structured narrative synthesis. In total, 115 articles met the eligibility criteria and were included in the final synthesis. Findings highlight four operational layers. For forecasting and nowcasting, multimodal pipelines that fuse plant telemetry with all sky imagery achieved error reductions relative to persistence baselines, and attention or graph based temporal models improved skill on multi hour horizons; practical latency was reported with gateway inference suitable for supervisory control. For fault and anomaly diagnostics, deep classifiers and segmentation models operating on infrared, electroluminescence, photoluminescence, and SCADA streams delivered high discriminative performance and supported explainable overlays for technician workflows. Degradation and remaining useful life estimation benefited from physics informed or Bayesian models that combine electrical and thermal or optical channels, improving early warning and calibration over purely data driven regressors. Finally, controller guidance for maximum power point tracking and thermal regulation increasingly leverages edge aware architectures while secure data fabrics align with IEC 61850 and FAIR principles. Across these layers, perovskite and tandem aware features reduce bias under heat and spectral variability and help close the gap between laboratory devices and fielded assets. The review also offers a taxonomy and decision matrix linking sensing, models, and deployment choices to operational objectives.