arXiv subjects

Ning Lin

Publications and source records attributed to Ning Lin.

42 records · Page 3Linked to original sources

arXiv2020

Hurricane-blackout-heatwave Compound Hazard Risk and Resilience in a Changing Climate

Hurricanes have caused power outages and blackouts, affecting millions of customers and inducing severe social and economic impacts. The impacts of hurricane-caused blackouts may worsen due to increased heat extremes and possibly increased hurricanes under climate change. We apply hurricane and heatwave projections with power outage and recovery process analysis to investigate how the emerging hurricane-blackout-heatwave compound hazard may vary in a changing climate, for Harris County in Texas (including major part of Houston City) as an example. We find that, under the high-emissions scenario RCP8.5, the expected percent of customers experiencing at least one longer-than-5-day hurricane-induced power outage in a 20-year period would increase significantly from 14% at the end of the 20th century to 44% at the end of the 21st century in Harris County. The expected percent of customers who may experience at least one longer-than-5-day heatwave without power (to provide air conditioning) would increase alarmingly, from 0.8% to 15.5%. These increases of risk may be largely avoided if the climate is well controlled under the stringent mitigation scenario RCP2.6. We also reveal that a moderate enhancement of critical sectors of the distribution network can significantly improve the resilience of the entire power grid and mitigate the risk of the future compound hazard. Together these findings suggest that, in addition to climate mitigation, climate adaptation actions are urgently needed to improve the resilience of coastal power systems.

stat.AP↗

arXiv2018

Tetris: Re-architecting Convolutional Neural Network Computation for Machine Learning Accelerators

Inference efficiency is the predominant consideration in designing deep learning accelerators. Previous work mainly focuses on skipping zero values to deal with remarkable ineffectual computation, while zero bits in non-zero values, as another major source of ineffectual computation, is often ignored. The reason lies on the difficulty of extracting essential bits during operating multiply-and-accumulate (MAC) in the processing element. Based on the fact that zero bits occupy as high as 68.9% fraction in the overall weights of modern deep convolutional neural network models, this paper firstly proposes a weight kneading technique that could eliminate ineffectual computation caused by either zero value weights or zero bits in non-zero weights, simultaneously. Besides, a split-and-accumulate (SAC) computing pattern in replacement of conventional MAC, as well as the corresponding hardware accelerator design called Tetris are proposed to support weight kneading at the hardware level. Experimental results prove that Tetris could speed up inference up to 1.50x, and improve power efficiency up to 5.33x compared with the state-of-the-art baselines.

cs.LG↗

arXiv2018

Tropical Cyclone Intensity Evolution Modeled as a Dependent Hidden Markov Process

A hidden Markov model is developed to simulate tropical cyclone intensity evolution dependent on the surrounding large-scale environment. The model considers three unobserved (hidden) discrete states of intensification and associates each state with a probability distribution of intensity change. The storm's transit from one state to another is described as a Markov chain. Both the intensity change and state transit components of the model are dependent on environmental variables including potential intensity, vertical wind shear, relative humidity, and ocean feedback. This Markov environment-dependent hurricane intensity model (MeHiM) is used to simulate the evolution of storm intensity along the storm track over the ocean, and a simple decay model is added to estimate the intensity change when the storm moves over land. Data for the North Atlantic (NA) basin from 1979-2014 (555 storms) are used for model development and evaluation. Probability distributions of 6-h and 24-h intensity change, lifetime maximum intensity, and landfall intensity based on model simulations and observations compare well. Although the MeHiM is still limited in fully describing rapid intensification, it shows a significant improvement over previous statistical models (e.g., linear, nonlinear, and finite mixture models).

stat.AP↗

arXiv2018

A reconstruction of Florida Traffic Flow During Hurricane Irma (2017)

Recent Hurricane Irma (2017) created the most extensive scale of evacuation in Florida's history, involving about 6.5 million people on mandatory evacuation order and 4 million evacuation vehicles. To understand the hurricane evacuation process, the spatial and temporal evolution of the traffic flow is a critical piece of information, but it is usually not fully observed. Based on the game theory, this paper employs the available traffic observation on main highways (20 cameras; including parts of Route 1, 27 and I-75, 95, 4,10) during Irma to reconstruct the traffic flow in Florida during Irma. The model is validated with self-reported twitters. The traffic reconstruction estimates the traffic demand (about 4 million cars in total) and the temporal and spatial distribution of congestion during the evacuation. The results compare well with available information from news reports and Twitter records. The reconstructed data can be used to analyze hurricane evacuation decisions and travel behavior.

physics.soc-ph↗

arXiv2018

Rapid Assessment of Damaged Homes in the Florida Keys after Hurricane Irma

On September 10, 2017, Hurricane Irma made landfall in the Florida Keys and caused significant damage. Informed by hydrodynamic storm surge and wave modeling and post-storm satellite imagery, a rapid damage survey was soon conducted for 1600+ residential buildings in Big Pine Key and Marathon. Damage categorizations and statistical analysis reveal distinct factors governing damage at these two locations. The distance from the coast is significant for the damage in Big Pine Key, as severely damaged buildings were located near narrow waterways connected to the ocean. Building type and size are critical in Marathon, highlighted by the near-complete destruction of trailer communities there. These observations raise issues of affordability and equity that need consideration in damage recovery and rebuilding for resilience.

stat.AP↗

arXiv2012

Second-order continuous-time non-stationary Gaussian autoregression

The objective of the paper is to identify and investigate all possible types of asymptotic behavior for the maximum likelihood estimators of the unknown parameters in the second-order linear stochastic ordinary differential equation driven by Gaussian white noise. The emphasis is on the non-ergodic case, when the roots of the corresponding characteristic equation are not both in the left half-plane.

math.ST↗