Water reservoir operating rules are typically derived based on the assumption of streamflow stationarity, however, this assumption could be undermined by climate change. Adaptive reservoir operation is one of the most effective strategies to support water resources management under non-stationarity, yet until now, adaptive strategies considering non-stationarity across multiple time scales are rarely investigated. We propose an adaptive reservoir operation framework that incorporates streamflow non-stationarity across time scales simultaneously. Specifically, we first decompose the streamflow into four frequency categories to detect non-stationarity features through reservoir operation simulations. Next, we incorporate the non-stationarity information from each frequency category into adaptive reservoir operation by using Bayesian Model Averaging. We apply this framework to reservoir operation of the Grand Ethiopian Renaissance Dam on the Blue Nile River and evaluate its effectiveness with streamflow simulated from 21 general circulation models (GCMs) for two greenhouse gases emission scenarios. We find that streamflow non-stationarity from all GCMs varies by future period and frequency category. The proposed Bayesian adaptive reservoir operation framework can detect streamflow non-stationarity across all frequency categories and predominantly outperforms conventional adaptive strategies, especially in terms of firm power output. In general, firm output increases under the Bayesian framework as the power generation reliability increases. The proposed framework offers a robust approach to identify adaptive strategies for reservoir operation to address streamflow non-stationarity.

水库运行规则通常是基于水流平稳性的假设得出的，但是，这种假设可能会因气候变化而受到破坏。自适应水库调度是在非平稳状态下支持水资源管理的最有效策略之一，但是直到现在，很少研究跨多个时间尺度考虑非平稳性的自适应策略。我们提出了一个自适应的油藏运行框架，该框架同时包含跨时间尺度的水流非平稳性。具体而言，我们首先将流量分解为四个频率类别，以通过油藏运行模拟检测非平稳性特征。接下来，我们使用贝叶斯模型平均将来自每个频率类别的非平稳性信息合并到自适应油藏操作中。我们将此框架应用于Blue Nile河上的埃塞俄比亚大文艺复兴大坝的水库运营，并针对两种温室气体排放情景，利用21种通用循环模型（GCM）模拟的流量评估了其有效性。我们发现，来自所有GCM的流量非平稳性随未来时段和频率类别而变化。提出的贝叶斯自适应水库运行框架可以检测所有频率类别上的水流非平稳性，并且特别是在确定的功率输出方面，其性能优于传统的自适应策略。通常，在贝叶斯框架下，随着发电可靠性的提高，企业的产量也会增加。所提出的框架提供了一种鲁棒的方法来识别适应性策略，以解决储层运行问题，以解决水流不稳定的问题。