Predicting solar yield is critical for optimizing electricity production and supporting the efficient design of photovoltaic (PV) systems. As the deployment of utility-scale PV plants expands, the need for accurate solar yield forecasting becomes more pressing, particularly to enhance grid performance under variable conditions. In this work, we evaluate the effectiveness of Pearson correlation as a feature selection method in the context of solar yield prediction using Artificial Neural Networks (ANNs). Our study investigates how well Pearson correlation identifies key input variables and compares these results to ANN-based feature importance rankings. The findings show that, while Pearson correlation provides insights into linear relationships, it may fail to capture the complexity of non-linear interactions within the ANN model. In particular, excluding the variable ‘precipitation’ led to the highest performance, while relying solely on the variable ‘DNI’ resulted in the poorest accuracy. These results suggest that Pearson correlation can be a useful but limited tool for feature selection in ANN-based solar yield predictions, especially when considering computational efficiency.