The Tibetan Plateau, a globally sensitive region to climate change, exhibits significant implications of its high-altitude lake surface water temperature (LSWT) evolution for regional ecological security. When exploring the factors influencing lake water temperature changes, various influencing factors are involved, such as meteorological conditions and topography. However, conventional methodologies demonstrate limited capacity in quantitatively resolving nonlinear interactive effects among multiple drivers. This study investigated 106 large-medium lakes across the Tibetan Plateau, developing a deep learning model based on Long Short-Term Memory (LSTM) networks integrated with SHapley Additive explanation (SHAP) interpretability analysis. We quantitatively differentiated the individual and interactive contributions of seven drivers (air temperature, precipitation, downward longwave radiation, downward shortwave radiation, air pressure, specific humidity, and wind speed) to LSWT variations at both regional and individual lake scales, systematically elucidating driving mechanisms and synergistic patterns. Key findings reveal: (1) Longwave and shortwave radiation emerged as dominant drivers, collectively contributing more than 80.0% of global SHAP values across scales, exhibiting significant positive correlations with LSWT. Air temperature and specific humidity demonstrated secondary influences, while precipitation and wind speed showed minimal impacts. (2) Ubiquitous interactive effects identified four predominant synergistic modes: linear pattern (e.g., downward longwave radiation-air temperature, 67.92% lakes), inverted U-shape pattern (e.g., specific humidity-air temperature, 51.89% lakes), effect cross-driven pattern (e.g., wind speed-specific humidity, 70.75% lakes), and threshold-constrained pattern (e.g., precipitation-air pressure, 100% lakes). (3) SHAP methodology effectively quantified nonlinear synergistic characteristics, revealing plateau lakes" heightened sensitivity to radiative factors attributable to high solar radiation permeability under thin atmospheric conditions. This study innovatively integrates deep learning with interpretability analysis, establishing a quantitative framework for disentangling complex driving mechanisms of high-altitude LSWT. The findings provide critical insights for predicting thermal responses under climate change and formulating differentiated regulatory strategies, bearing substantial practical scientific significance.