The rapid expansion of utility-scale photovoltaic (PV) is critical for the advancement of the global energy transition and climate mitigation. However, the ecological feedback mechanisms of PV arrays driving vegetation dynamics through microclimate and soil modification in degraded ecosystems remain unresolved. This study applied the interpretable machine learning model (XGBoost-SHAP) framework with a high-resolution field monitoring network at the Laijiahe (LJH) PV station in China's Qinling Mountains to reveal how PV arrays influence vegetation greenness and phenology dynamics through microclimate and soil modification. Two vegetation indices and 11 microclimate and soil variables within the PV station and Non-PV (NPV) control sites were acquired in this study from July 2023 to June 2024. The results show that PV arrays reshaped the local microclimate and soil conditions through forming a warmer and drier effect in air conditions and cooler and wetter in the soil environment. GCC and NDVI in the PV area increase by an average of 15% and 0.2 compared with the NPV area, indicating PV facilities significantly enhance vegetation greening. Advance phenology shifts in 3-15 days earlier for the start (SOS), peak (POS), and end (EOS) of the growing season are detected, exceeding global warming impacts and accompanied by a 10 times increase in growth rates. The XGBoost-SHAP analysis reveals that greening variations under PV arrays are mainly driven by soil hydrothermal coupling related factors with 63.37% contributions, although near-surface heat balance dominance in natural ecosystems. The adaptability of the PV-regulated local ecosystem was enhanced by a multiple and intensifying interaction network among microclimate and soil factors, with interactions contributing to 51.8%. The warming effect beneath PV arrays is a key driver of the earlier phenology phases, which is more significant in the autumn season. Such vegetation dynamic shifts demonstrate that PV-induced buffer effects in shading, rain collection, and warming form synergistic microclimates and soil conditions to enhance ecological resilience by creating more favorable conditions for vegetation activities in wasteland regions. The findings offer important support for integrating renewable energy infrastructure with ecosystem management, highlighting the great potential of solar power generation in promoting vegetation resilience in semi-arid landscapes.