The present study aims to estimate the power production (PG) from 2-axis tracking On-grid solar systems located at the coastal cities in the Mediterranean region using four empirical models. To this aim, latitude (Lat), longitude (long), altitude (alt), solar irradiation (SR), ambient temperature (AT), and clearance index (CI) are considered as input variables for the models. The outcomes demonstrate that the ENN model outperforms both MLPNN and ARIMA. Additionally, the results indicate that the NLNE approach could improve the performance of a single model by at least 20%. Therefore, the NLNE model enhances predictive accuracy by predicting the PG across different coastal areas more consistently and accurately. This suggests that the durability of the solar power prediction models might be significantly increased by employing ensemble learning techniques, which would make them more suitable for practical applications in energy management and grid integration.