Prior knowledge of solar radiation potential is a necessity for the decision making and the establishment of solar projects. Current prediction methods are site-dependent and their performance outside the area of application is questionable. Therefore, a solar radiation zoning method using a combination of supervised and unsupervised machine learning techniques and satellite gridded data is proposed. The method uses Agglomerative Hierarchical Clustering (AHC) following the ward’s method based on the monthly mean solar insolation on a horizontal surface (H), and Support Vector Machine Classifier (SVM-C) to assist AHC to effectively classify all the areas based on solar radiation data plus temperature, humidity, wind speed and precipitation. Four solar radiation zones have been established. The zones are homogeneous in space and have distinctive solar and meteorological characteristics. The t-distributed Stochastic Neighbor Embedding (t-SNE) is applied to the obtained zones to distinguish trends within solar radiation zones based on temperature, humidity and wind speed, and a total of 8 sub-zones with distinctive meteorological characteristics have been identified. The resulting solar radiation zones and sub-zones cover all Morocco can help for preliminary assessment and decision making, especially for areas with no solar radiation records.

太阳能辐射潜力的先验知识对于太阳能项目的决策和建立是必要的。当前的预测方法依赖于站点，它们在应用领域之外的性能值得怀疑。因此，提出了一种结合有监督和无监督机器学习技术以及卫星网格数据的太阳辐射分区方法。该方法采用基于水平表面（H）上月平均太阳日照的病房方法的凝聚层次聚类（AHC）和支持向量机分类器（SVM-C）辅助AHC有效地对所有基于太阳的区域进行分类。辐射数据加上温度、湿度、风速和降水。已经建立了四个太阳辐射区。这些区域在空间上是同质的，具有独特的太阳和气象特征。将t分布随机邻域嵌入（t-SNE）应用于获得的区域，根据温度、湿度和风速区分太阳辐射区域内的趋势，共识别出8个具有鲜明气象特征的子区域。由此产生的太阳辐射区和子区覆盖了摩洛哥的所有地区，可以帮助进行初步评估和决策，特别是对于没有太阳辐射记录的地区。共确定8个具有鲜明气象特征的分区。由此产生的太阳辐射区和子区覆盖了摩洛哥的所有地区，可以帮助进行初步评估和决策，特别是对于没有太阳辐射记录的地区。共确定8个具有鲜明气象特征的分区。由此产生的太阳辐射区和子区覆盖了摩洛哥的所有地区，可以帮助进行初步评估和决策，特别是对于没有太阳辐射记录的地区。