Turbulent flows can exhibit extreme events, which are characterized by sudden bursts in the system observables. These events pose significant challenges for prediction and control owing to their intermittent, high-dimensional, and strongly nonlinear nature. In this work we present a predominantly data-driven control framework for the suppression of extreme events in turbulent flows, leveraging reduced-order modeling for nonlinear compression of high-dimensional flow fields and a data-driven clustering algorithm for the identification of precursors to extreme events. A control law is defined in the low-dimensional latent space and is optimized to efficiently mitigate extreme events through actuation values associated with the preidentified clusters. To address issues in the latent space structure that arise from invariance transformations present in many nonlinear systems, leading to an inflated latent space, we employ symmetry-aware autoencoders to establish a more structured and compact latent space. By analyzing high-fidelity simulations of a canonical chaotic flow (Kolmogorov flow), we demonstrate how the resulting closed-loop dynamics exhibit a substantial reduction (up to ca. 99.4%) in the frequency and intensity of extreme events. The framework exhibits scalability to a more chaotic flow regime characterized by higher Reynolds number without encountering numerical instabilities, achieving a 96.9% reduction in the occurrence of extreme events. The ability to incorporate controller limitations, including actuator latency relevant to practical applications of the framework, has also been demonstrated. This study highlights the efficacy of data-driven methods that require little to no prior knowledge of the underlying system dynamics to achieve effective flow control, thereby providing a pathway for general real-time suppression of extreme events in turbulent flows. ... Read more

Accurate estimates of surface solar irradiance (SSI) are essential for solar resource assessments and solar energy forecasts in grid integration and building control applications. SSI estimates for spatially extended regions can be retrieved from geostationary satellites such as Meteosat. Traditional SSI satellite retrievals like Heliosat rely on physical radiative transfer modelling. We introduce a machine-learning-based satellite retrieval for instantaneous SSI and demonstrate its capability to provide accurate and generalizable SSI estimates across Europe. Our deep learning retrieval provides near real-time SSI estimates based on data-driven emulation of Heliosat and fine-tuning on pyranometer networks. By including SSI from ground stations, our SSI retrieval model can outperform Heliosat accuracy and generalize well to regions with other climates and surface albedos in cloudy conditions (clear-sky index < 0.8). Our results indicate that the generalizability of a data-driven SSI retrieval model is not only related to the model training data or training method, but also depends on the amount of cloudiness present in the location at which SSI is retrieved with the data-driven model. We found that, in cloudy conditions, a model trained only on ground stations can estimate SSI accurately even in locations with different surface albedos, far away from the training test domain. We also show that the SSI retrieved from Heliosat exhibits large biases in mountain regions, and that training and fine-tuning our retrieval models on SSI data from ground stations strongly reduces these biases, outperforming Heliosat. Furthermore, we quantify the relative importance of the Meteosat channels and other predictor variables like solar zenith angle for the accuracy of our deep learning SSI retrieval model in different cloud conditions. We find that in cloudy conditions multiple near-infrared and infrared channels enhance the performance. Our results can facilitate the development of more accurate satellite retrieval models of surface solar irradiance. ... Read more