I-Corps: Global solar irradiance forecasting with flexible prediction time horizons

这个I-Corps项目更广泛的影响/商业潜力是开发一种可推广的方法，根据免费提供的卫星云图准确和可扩展地预测太阳光（辐照度）。可再生能源的高渗透率对于电网脱碳和减缓人为气候变化至关重要。然而，可再生能源的不稳定性目前是其成本效益整合的一个障碍，这只能通过改进能源储存来部分解决。提高太阳辐照度预测的准确性是有效利用和整合太阳能的重要一步，因为它使公用事业和电力市场运营商能够在批发电力市场中做出明智的决策，以安排备用容量和设计有效的投标策略。此外，它允许在各种规模上评估太阳能部署的战略价值。该项目利用大数据的力量和深度学习的最新发展，开发了一个可推广的卷积全球水平辐照度预测模型，该模型可以帮助电力公司、能源市场运营商和政策制定者在监管、规划和运营电网时做出明智的决策。 具体来说，该模型利用卷积神经网络和云图像来准确预测太阳辐照度，并已在美国多个地点得到验证。 已经进行了严格的后验分析，以研究模型的变量技能作为预测时间范围和云量的函数，并展示了将中尺度大气动力学用于增强预测性能的好处。该技术解决了最先进的太阳辐照度预测技术的关键根本差距，因为它结合了基于物理的太阳辐照度预测模型（无需现场测量）的优势和统计模型的准确性和可扩展性。该奖项反映了NSF的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

The broader impact/commercial potential of this I-Corps project is the development of a generalizable approach for accurate and scalable prediction of solar light (irradiance) based on freely available satellite cloud images. High penetration of renewable energy is essential for decarbonizing the grid and decelerating anthropogenic climate change. The intermittency of renewable energy, however, is currently a hurdle in its cost-effective integration, which is only partially solved by improved energy storage. Improving the accuracy of solar irradiance forecasts is a vital step towards effective utilization and integration of solar energy since it allows utilities and electricity market operators to make informed decisions for scheduling reserve capacity and designing efficient bidding strategies in the wholesale power markets. Moreover, it allows the strategic value of solar deployment to be assessed at various scales. The proposed technology can benefit utilities, energy market operators, and policymakers by allowing them to make informed decisions in regulating, planning, and operating the grid.This I-Corps project harnesses the power of big data and the latest developments in deep learning to develop a generalizable Convolutional Global Horizontal Irradiance prediction model. Specifically, the model leverages convolutional neural networks and cloud imagery to accurately predict solar irradiance and has been validated across several U.S. sites. Rigorous posterior analyses have been conducted to study the model's variable skill as a function of prediction time horizon and cloud cover and to demonstrate the benefits of incorporating mesoscale atmospheric dynamics for enhanced prediction performance. The technology addresses key fundamental gaps in state-of-the-art solar irradiance forecasting technologies as it combines the benefits of physics-based solar irradiance prediction models (with no need for in-situ measurement) with the accuracy and scalability of statistical models.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.