Search for: All records

Climate change is expected to intensify the effects of extreme weather events on power systems and increase the frequency of severe power outages. The large-scale integration of environment-dependent renewables during energy decarbonization could induce increased uncertainty in the supply–demand balance and climate vulnerability of power grids. This Perspective discusses the superimposed risks of climate change, extreme weather events and renewable energy integration, which collectively affect power system resilience. Insights drawn from large-scale spatiotemporal data on historical US power outages induced by tropical cyclones illustrate the vital role of grid inertia and system flexibility in maintaining the balance between supply and demand, thereby preventing catastrophic cascading failures. Alarmingly, the future projections under diverse emission pathways signal that climate hazards — especially tropical cyclones and heatwaves — are intensifying and can cause even greater impacts on the power grids. High-penetration renewable power systems under climate change may face escalating challenges, including more severe infrastructure damage, lower grid inertia and flexibility, and longer post-event recovery. Towards a net-zero future, this Perspective then explores approaches for harnessing the inherent potential of distributed renewables for climate resilience through forming microgrids, aligned with holistic technical solutions such as grid-forming inverters, distributed energy storage, cross-sector interoperability, distributed optimization and climate–energy integrated modelling. 

Placing and orienting a camera to compose aesthetically meaningful shots of a scene is not only a key objective in real-world photography and cinematography but also for virtual content creation. The framing of a camera often significantly contributes to the story telling in movies, games, and mixed reality applications. Generating single camera poses or even contiguous trajectories either requires a significant amount of manual labor or requires solving highdimensional optimization problems, which can be computationally demanding and error-prone. In this paper, we introduce GAIT, a Deep Reinforcement Learning (DRL) agent, that learns to automatically control a camera to generate a sequence of aesthetically meaningful views for synthetic 3D indoor scenes. To generate sequences of frames with high aesthetic value, GAIT relies on a neural aesthetics estimator, which is trained on a crowed-sourced dataset. Additionally, we introduce regularization techniques for diversity and smoothness to generate visually interesting trajectories for a 3D environment, and to constrain agent acceleration in the reward function to generate a smooth sequence of camera frames. We validated our method by comparing it to baseline algorithms, based on a perceptual user study, and through ablation studies. The source code of our method will be released with the final version of our paper. 

Abstract Connecting the solar wind observed throughout the heliosphere to its origins in the solar corona is one of the central aims of heliophysics. The variability in the magnetic field, bulk plasma, and heavy ion composition properties of the slow wind are thought to result from magnetic reconnection processes in the solar corona. We identify regions of enhanced variability and composition in the solar wind from 2003 April 15 to May 13 (Carrington Rotation 2002), observed by the Wind and Advanced Composition Explorer spacecraft, and demonstrate their relationship to the separatrix–web (hereafter, S-Web) structures describing the corona’s large-scale magnetic topology. There are four pseudostreamer (PS) wind intervals and two helmet streamer (HS) heliospheric current sheet/plasma sheet crossings (and an interplanetary coronal mass ejection), which all exhibit enhanced alpha-to-proton ratios and/or elevated ionic charge states of carbon, oxygen, and iron. We apply the magnetic helicity–partial variance of increments ( H m –PVI) procedure to identify coherent magnetic structures and quantify their properties during each interval. The mean duration of these structures are ∼1 hr in both the HS and PS wind. We find a modest enhancement above the power-law fit to the PVI waiting-time distribution in the HS-associated wind at the 1.5–2 hr timescales that is absent from the PS intervals. We discuss our results in the context of previous observations of the ∼90 minutes periodic density structures in the slow solar wind, further development of the dynamic S-Web model, and future Parker Solar Probe and Solar Orbiter joint observational campaigns. 

This paper aims to develop distributed algorithms for nonconvex optimization problems with complicated constraints associated with a network. The network can be a physical one, such as an electric power network, where the constraints are nonlinear power flow equations, or an abstract one that represents constraint couplings between decision variables of different agents. Despite the recent development of distributed algorithms for nonconvex programs, highly complicated constraints still pose a significant challenge in theory and practice. We first identify some difficulties with the existing algorithms based on the alternating direction method of multipliers (ADMM) for dealing with such problems. We then propose a reformulation that enables us to design a two-level algorithm, which embeds a specially structured three-block ADMM at the inner level in an augmented Lagrangian method framework. Furthermore, we prove the global and local convergence as well as iteration complexity of this new scheme for general nonconvex constrained programs, and show that our analysis can be extended to handle more complicated multi-block inner-level problems. Finally, we demonstrate with computation that the new scheme provides convergent and parallelizable algorithms for various nonconvex applications, and is able to complement the performance of the state-of-the-art distributed algorithms in practice by achieving either faster convergence in optimality gap or in feasibility or both.

 

Supervised training of optical flow predictors generally yields better accuracy than unsupervised training. However, the improved performance comes at an often high annotation cost. Semi-supervised training trades off accuracy against annotation cost. We use a simple yet effective semi-supervised training method to show that even a small fraction of labels can improve flow accuracy by a significant margin over unsupervised training. In addition, we propose active learning methods based on simple heuristics to further reduce the number of labels required to achieve the same target accuracy. Our experiments on both synthetic and real optical flow datasets show that our semi-supervised networks generally need around 50% of the labels to achieve close to full-label accuracy, and only around 20% with active learning on Sintel. We also analyze and show insights on the factors that may influence active learning performance. Code is available at https://github.com/duke-vision/ optical-flow-active-learning-release. 

Understanding the magnetic structure of filament channels is difficult but essential for identifying the mechanism (s) responsible for solar eruptions. In this paper we characterize the magnetic field in a well-observed filament channel with two independent methods, prominence seismology and magnetohydrodynamics flux-rope modeling, and compare the results. In 2014 May and June, active region 12076 exhibited a complex of filaments undergoing repeated oscillations over the course of 12 days. We measure the oscillation periods in the region with both Global Oscillation Network Group Hαand Solar Dynamics Observatory (SDO) Advanced Imaging Assembly EUV images, and then utilize the pendulum model of large-amplitude longitudinal oscillations to calculate the radius of curvature of the fields supporting the oscillating plasma from the derived periods. We also employ the regularized Biot–Savart laws formalism to construct a flux-rope model of the field of the central filament in the region based on an SDO Helioseismic and Magnetic Imager magnetogram. We compare the estimated radius of curvature, location, and angle of the magnetic field in the plane of the sky derived from the observed oscillations with the corresponding magnetic-field properties extracted from the flux-rope model. We find that the two models are broadly consistent, but detailed comparisons of the model and specific oscillations often differ. Model observation comparisons such as these are important for advancing our understanding of the structure of filament channels.