Abstract:Wind speed forecasting provides the upcoming wind information and is important to the safe operation of High-Speed Railway (HSR). However, it remains a challenge due to the stochastic and highly varying characteristics of wind. In this paper, we propose a novel Transformer-based method for short-term wind speed forecasting, named WindTrans. Two major cruxes are addressed. First, the task is performed on fine-grained wind speed gathered from multiple sensors. These data present dynamic intra-series and inter-series correlations, which are hard for previous methods to recover. We advance a Transformer-based deep learning model, which has two distinctive characteristics: (1) a graph encoder, which captures the dynamic spatial correlation among wind speeds at different locations, and (2) a temporal decoder to model long sequence wind speed time series, which is resistant to noise in time series. Second, wind speed patterns gradually evolve in long-term periods, thus deactivating prediction models trained on historical data. To tackle this bottleneck, we put forward an experience replay-based scheme to renew the model regularly. To ensure that the renewed model still dominates historical wind patterns, we store and replay only a small portion of historical data named episodic memory. A simple but efficient strategy is designed to constitute episodic memory and thus relieve the computation burden. Experiments conducted on two real-world datasets demonstrate the superiority of our method over existing approaches. Particularly, WindTrans surpasses state-of-the-art methods by up to 36.7%, 29.3% and 13.3% improvement in MAPE measure for 1 hour ahead prediction on 10-minute, 5-minute, and 1-minute-based tasks, respectively. Furthermore, via our continual learning scheme, the model retains competitive performance with only 6.9% datum stored and retrained on.

Visualwind: a Novel Video Dataset for Cameras to Sense the Wind

The Visual-Inertial- Dynamical Multirotor Dataset

The Visual-Inertial-Dynamical Multirotor Dataset

Wind Measurement by Computer Vision on Unmanned Sailboat

Optical Flow Dataset and Benchmark for Visual Crowd Analysis

Seeing the Wind: Visual Wind Speed Prediction with a Coupled Convolutional and Recurrent Neural Network

WindSeer: Real-time volumetric wind prediction over complex terrain aboard a small UAV

Visual anemometry: physics-informed inference of wind for renewable energy, urban sustainability, and environmental science

WindSeer: real-time volumetric wind prediction over complex terrain aboard a small uncrewed aerial vehicle

WindTrans: Transformer-Based Wind Speed Forecasting Method for High-Speed Railway

Machine-learned flow estimation with sparse data -- exemplified for the rooftop of a UAV vertiport

Large-scale Wind Field Reconstruction Based on Satellite Image Data

Cloth in the Wind: A Case Study of Physical Measurement through Simulation

Estimation of Wind Speed Based on Schlieren Machine Vision System Inspired by Greenhouse Top Vent

Learning Optical Flow from Event Camera with Rendered Dataset

Understanding User Behavior in Volumetric Video Watching: Dataset, Analysis and Prediction

OmniFlow: Human Omnidirectional Optical Flow

Wind Speed Measurement via Visual Recognition of Wind-InducedWaving Light Stick Target

RealFlow: EM-Based Realistic Optical Flow Dataset Generation from Videos

Wind estimation using quadcopter motion: A machine learning approach

Physics-inspired and data-driven two-stage deep learning approach for wind field reconstruction with experimental validation