Fast and Robust Wind Speed Prediction Under Impulsive Noise via Adaptive Graph-Sign Diffusion

Online estimation of time-varying wind speed across various locations is a crucial task for applications such as renewable energy generation, weather prediction, and environmental science. In this paper, we propose an adaptive Graph-Sign Diffusion (GSD) algorithm to predict the time-varying wind speed in real time. Leveraging the expressiveness power of Graph Signal Processing, our proposed GSD algorithm is formulated on a combination of adaptive graph filtering, graph diffusion, and l(1)-norm optimization. The GSD algorithm outputs a fast and robust prediction of time-varying graph signals under impulsive noise in an online manner. Experimenting with real-world data shows that the GSD algorithm accurately predicts the time-varying wind speed at multiple sensor locations.

Automated summary

In this paper, we propose an adaptive Graph-Sign Diffusion (GSD) algorithm to predict the time-varying wind speed in real time, which is crucial for applications like renewable energy generation and weather prediction. The GSD algorithm, formulated on a combination of adaptive graph filtering, graph diffusion, and l(1)-norm optimization, outputs a fast and robust prediction of time-varying graph signals under impulsive noise in an online manner. Experimental results demonstrate the accurate predictions of the GSD algorithm for wind speed at multiple sensor locations.