The ever‐growing secondary market of photovoltaic (PV) systems (i.e., the transaction of solar plants ownership) calls for reliable and high‐quality long‐term PV degradation forecasts to mitigate the financial risks. However, when long‐term PV performance degradation forecasts are required after a short time with limited degradation history, the existing physical and data‐driven methods often provide unrealistic degradation scenarios. Therefore, we present a new data‐driven method to forecast PV lifetime after a small performance degradation of only 3%. To achieve an accurate and reliable forecast, the developed method addresses the fundamental challenges that usually affect long‐term degradation evaluation such as data treatment, choosing a good degradation model, and understanding the different degradation patterns. In the paper, we propose and describe an algorithm for degradation trend evaluation, a new concept of multiple “time‐ and degradation pattern‐dependent” degradation factors. The proposed method has been calibrated and validated using different PV modules and systems data of 5 to 35 years of field exposure. The model has been benchmarked against existing statistical models evaluating 11 experimental PV systems with different technologies. The key advantage of our model over statistical ones is the ability to perform more reliable forecasts with limited degradation history. With an average relative uncertainty of 7.0%, our model is outstanding in consistency for different forecasting time horizons. Moreover, the model is applicable to all PV technologies. The proposed method will aid in making reliable financial decisions and also in adequately planning operation and maintenance activities.

中文翻译：

---

基于退化模式的光伏寿命预测模型

不断增长的光伏（PV）系统二级市场（即太阳能工厂所有权交易）要求提供可靠且高质量的长期PV退化预测，以减轻财务风险。但是，当在短时间内具有有限的降级历史记录后需要长期的PV性能降级预测时，现有的物理和数据驱动方法通常会提供不切实际的降级方案。因此，我们提出了一种新的数据驱动方法，以预测性能仅下降3％之后的PV寿命。为了获得准确而可靠的预测，已开发的方法解决了通常会影响长期退化评估的基本挑战，例如数据处理，选择良好的退化模型以及了解不同的退化模式。在纸上 我们提出并描述了一种用于退化趋势评估的算法，一种新的概念，该概念涉及多个“时间和退化模式相关”的退化因子。所提出的方法已使用不同的光伏组件和5至35年的现场暴露系统数据进行了校准和验证。该模型已针对评估11种采用不同技术的光伏系统的现有统计模型进行了基准测试。与统计模型相比，我们的模型的主要优势在于能够以有限的降级历史执行更可靠的预测。我们的模型的平均相对不确定度为7.0％，在不同的预测时间范围内均具有出色的一致性。此外，该模型适用于所有光伏技术。