An Intelligent Data-Centric Architecture for High-Accuracy Solar PV Output Prediction Using Advanced Predictive Analytics Models

Abstract

The increasing integration of photovoltaic (PV) systems into modern power grids has intensified the need for accurate, scalable, and adaptive predictive models for solar energy forecasting. This study proposes an intelligent data-centric architecture designed to enhance the accuracy of PV output prediction by integrating structured environmental datasets, historical power generation records, and advanced machine learning-based predictive analytics. Unlike conventional model-centric approaches, the proposed architecture prioritizes data quality, feature relevance, and multi-source integration, enabling more robust forecasting under highly variable meteorological conditions. The system leverages insights from established predictive methodologies, including linear regression, artificial neural networks, decision tree ensembles, and random forest-based models, to construct a hybrid analytical pipeline.
The framework is evaluated conceptually through comparative synthesis of existing literature and methodological benchmarking. Prior studies demonstrate that machine learning models significantly outperform traditional statistical approaches in PV forecasting due to their ability to capture nonlinear relationships between environmental variables and solar output (Scott et al., 2023). Additionally, region-specific modeling strategies incorporating temperature, irradiance, and humidity have been shown to improve predictive stability (Khandakar et al., 2019). The proposed architecture extends these insights by embedding data preprocessing, feature engineering, and model orchestration into a unified pipeline.
The findings suggest that a data-centric approach improves generalization performance, reduces overfitting risks, and enhances adaptability across diverse climatic conditions. The study contributes a structured framework for next-generation solar forecasting systems, offering practical implications for smart grid optimization, renewable integration, and energy planning.