In this study, we employed two distinct neural network models, specifically, a backpropagation neural network (BPNN) and a radial basis function network (RBFN) to forecast daily solar radiation levels in diverse cities across Libya. The training process involved the utilization of geographical and meteorological data amassed over a comprehensive six-year timeframe from 25 cities within Libya. The dataset encompassed a spectrum of parameters, including Latitude, Longitude, and month. To fine-tune the models for optimal predictive performance, a variety of learning parameters were incorporated during the training phase. Subsequently, in the experimental stage, both neural networks underwent testing using data that had not been part of the initial training set, revealing distinct performance differentials. Experimentally, the radial basis function network emerged as the superior performer, surpassing the backpropagation neural network with accuracies of 93.20% and 93.00%, respectively.