Research

Spatiotemporal Graph Attention Network(STGAT)

The spatiotemporal Graph Attention Network (STGAT) model captures spatial correlations among different stations and temporal dependencies of time series. The STGAT model employs a single layer GAT network to extract spatial correlations among stations and temporal dependencies are captured by the GRU model with a acurate prediction. This model was used in Wind Power prediction and Multi-zone indoor temperature prediction.

Bayesian LSTM for Probabilistic prediction

This research focuses on the uncertainty set prediction of the aggregated generation of geographically distributed wind farms. A Spatio-temporal model is proposed to learn the dynamic features from partial observation in near-surface wind fields of neighboring wind farms. We use Bayesian LSTM, a probabilistic prediction model, to obtain the uncertainty set of the generation in individual wind farms. Then, spatial correlation between different wind farms is presented to correct the output results.

Application of time series model bagging in communication network

In this project, we use multiple time series bagging model to predict network flow in communication network to optimize network flow distribution within a multi-router system.