A Review of Photovoltaic Power Forecasting Integrating Meteorological Data

Abstract

: Photovoltaic (PV) power generation is affected by the coupled effects of solar irradiance, cloud cover, temperature, humidity, wind speed, precipitation, aerosols, terrain, module configuration, and power-station operating conditions. These factors make PV output random, intermittent, nonlinear, and strongly time-varying. With the expansion of distributed PV and large PV bases, short-term and ultra-short-term power forecasting has become essential for dispatch, reserve allocation, energy-storage operation, accommodation assessment, and plant operation and maintenance. This review synthesizes PV power forecasting studies from the perspective of meteorological-data integration. It clarifies the literature-search strategy, distinguishes this review from existing surveys, and organizes the field into a unified taxonomy covering data sources, meteorological mechanisms, forecasting horizons, algorithmic families, mechanism-data fusion, regional forecasting, and uncertainty-aware prediction. The review shows that solar irradiance remains the dominant driver of PV output, whereas temperature and wind speed modify conversion efficiency, and cloud cover, precipitation, aerosols, dust, and terrain amplify fluctuations and bias. Because reported accuracy values are often obtained from different datasets, horizons, capacities, and metrics, this paper emphasizes conditional comparison rather than direct ranking. Future work should focus on benchmark datasets, physically constrained learning, uncertainty calibration, foundation-model adaptation, cross-site generalization, computational cost