Several climatologists and experts in the renewable energy field agree that GHI and DNI calculation models must be revised because of the increasingly unpredictable and powerful climatic disturbances. The construction of analytical mathematical models for the prediction of these disturbances is almost impossible because the physical phenomena relating to the climate are often complex. We raise the question over the current and future PV system’s sustainable energy production and whether climate disturbances will be affecting this sustainability and resulting in supply decline. In this paper, we tried to use deep learning as a tool to predict the evolution of the future production of any geographic site. This approach can allow for improvements in decision-making concerning the implantation of solar PV or CSP plants. To reach this aim, we have deployed the databases of NASA and the Tunisian National Institute of Meteorology relating to the climatic parameters of the case study region of El Akarit, Gabes, Tunisia. In spite of the colossal amount of processed data that dates back to 1985, the use of deep learning algorithms allowed for the validation of the previously made estimates of the energy potential in the studied region. The calculation results suggested an increase in production as it was confirmed by the 2019 measures. The findings obtained from the case study region were reliable and seemed to be very promising. The results obtained using deep learning algorithms were similar to those produced by conventional calculation methods. However, while conventional approaches based on measurements obtained using hardware solutions (ground sensors) are expensive and very difficult to implement, the suggested new approach is cheaper and more convenient. In the existence of a protracted controversy over the hypothetical effects of climate change, making advances in artificial intelligence and using new deep learning algorithms are critical procedures to strengthening conventional assessment tools of the production sites of photovoltaic energy and CSP plants.

中文翻译：

---

基于深度学习的气候变化下光伏电站潜力估算：突尼斯El Akarit案例研究

可再生能源领域的一些气候学家和专家都认为，由于不可预测和强大的气候干扰，必须修改GHI和DNI计算模型。用于预测这些干扰的解析数学模型几乎是不可能的，因为与气候有关的物理现象通常很复杂。我们对当前和未来的光伏系统的可持续能源生产以及气候干扰是否会影响这种可持续性并导致供应下降提出疑问。在本文中，我们尝试将深度学习作为一种工具来预测任何地理站点未来产品的发展。这种方法可以改进有关太阳能PV或CSP装置植入的决策。为了达到这个目的，我们已经部署了NASA和突尼斯国家气象研究所的数据库，这些数据库与突尼斯加贝斯El Akarit案例研究区域的气候参数有关。尽管可追溯到1985年的处理数据量巨大，但深度学习算法的使用允许验证先前对研究区域中的能源潜力进行的估算。计算结果表明产量增加，这一点已得到2019年措施的确认。从案例研究区域获得的发现是可靠的，并且似乎很有希望。使用深度学习算法获得的结果与传统计算方法产生的结果相似。然而，尽管基于使用硬件解决方案（地面传感器）获得的测量值的常规方法昂贵且难以实施，但建议的新方法更便宜，更方便。在关于气候变化的假说影响存在长期争议的情况下，人工智能的进步和使用新的深度学习算法是加强光伏能源和CSP工厂生产场所常规评估工具的关键程序。