Machine learning models offer dynamic systems that can surpass the limits of quantitative models for estimating solar photovoltaic (PV) output, particularly under changing seasonal weather conditions. This study evaluates data-driven models, such as linear regression (LR), decision trees (DTs), and artificial neural networks (ANNs), to determine solar energy production. Historical metadata from May 2024 to December 2025 utilized as the input data pertained to climatic factors such as sun irradiance, temperature, humidity, wind speed, cloud cover, and precipitation. The ML systems were trained and validated in MATLAB, with integrated hyperparameter adjustment to enhance the model performance. Quantitative indicators comprising the mean squared error (MAE), root mean square error (RMSE), and coefficient of determination (R²) were applied to validate and compare the generalizability of the resultant models. The ANN model surpassed LR and the DT in identifying nonlinear relationships, with an average performance of an MSE = 0.7512, an RMSE = 0.8667, and R² = 0.9725. This was achieved on an optimized ANN architecture using the Bayesian Regularization training methodology, one hidden layer with eight neurons, a sigmoid activation function, and a learning rate of 0.001. These results demonstrate that the ANN provides a more efficient strategy for reliable PV output estimation, vital to improving energy planning and facilitating the larger integration of solar technology into weather-dependent power networks.