<p>Incorporating photovoltaic (PV) systems into power grids is increasingly critical, especially in sun-rich Arab countries. This study introduces a novel approach using spiking neural networks (SNNs) for forecasting solar radiation. We utilize the Lava framework to develop and compare different neural architectures in both univariate and multivariate scenarios. Our findings reveal that SNNs achieve performance comparable to traditional artificial neural networks (ANNs) in modeling and forecasting. Specifically, the error rates of SNNs, when using a convolutional neural network (CNN) architecture on multivariate time series data, are similar to those of their ANN counterparts. However, SNNs present a significant advantage in power efficiency, being approximately 9 times more efficient when estimated theoretically on both the Loihi neuromorphic chip and traditional GPUs using the direct training approach with the SLAYER algorithm. Furthermore, the efficiency improvement is about 7 to 8 times on both Loihi and GPUs using the conversion training approach with a bootstrap algorithm. This research contributes to optimizing PV operations in Arab regions by introducing SNNs as a power-efficient alternative to traditional ANNs for solar radiation forecasting. It aligns with global sustainability goals by offering a robust model for efficiently predicting solar energy outputs while minimizing the computational energy footprint. Although the datasets in this study are drawn from sources in the Arab region, the methodologies and findings are applicable to other areas due to their universal relevance. The practical implications of our findings support the expansion of sustainable energy infrastructures, underscoring the strategic importance of innovative forecasting models in enhancing the reliability and efficiency of renewable energy sources.</p>