The optimization of solar panel tilt angle plays a vital role in enhancing the energy yield of fixed photovoltaic (PV) systems, particularly across regions with diverse climatic variability like India. This study presents a data-driven Transformer-based framework for predicting the optimum tilt angle (OTA) using key solar parameters global solar radiation, clearness index, and extraterrestrial radiation collected from multiple locations representing different climatic zones across India. Four different model configurations were evaluated: a baseline Transformer model (with and without data augmentation), a Moving Average (MA) combined Transformer model, and an Empirical Mode Decomposition (EMD) integrated Transformer model. The results indicate that the EMD + Transformer model outperformed all others, achieving the lowest MAE (10.027), RMSE (13.779), and the highest R2 score (0.639). The study emphasizes that traditional data augmentation techniques offer limited benefits for such prediction tasks, whereas signal decomposition significantly enhances learning by filtering noise and extracting relevant patterns from non-linear solar inputs. The proposed model demonstrates high reliability and generalizability in predicting OTA across India’s diverse solar climatic conditions, providing a valuable tool for PV deployment planning and optimization.