Prof. Yair Weiss – Learning and Vision.
Dr. Amir Globerson – Learning.
Dr. Ami Wiesel – Statistical signal processing.
Mr. Ofir Bibi – PhD candidate.
- Probabilistic models.
Statistical signal processing:
- Detection theory.
- Large scale.
In general, the group focuses on load prediction and fault detection in the smart grid. Specifically, we exploit the special spatio-temporal statistical properties of the future grid and propose a robust and scalable solution appropriate for the grid’s dynamic characteristics, the available internal and external data, e.g. load readings, temperature and market pricing, the required performance metrics, e.g., peak consumption based accuracy criteria, and the practical constraints imposed by the current sensing and communication infrastructure.
- Spatio temporal statistical models for electricity load.
- Online and distributed message passing algorithms for load prediction.
- Simultaneous cross-meter fusion for improved performance.
- Robust methods which are not sensitive to outliers and failures.
- Huber regression.
- Robust cross-meter fusion.
- Time series, vector series and matrix series models.
- Autoregressive moving average vector models.
- Goal specific inference, e.g., max load estimation.
- External and latent variables effects
- Temperature – non-linear effects.
- Market and pricing.
- Collaboration with Powercom and Control Applications.
Fault detection and localization.
- Large scale models for nominal and faulty behavior in the grid.
- Gaussian graphical models.
- Error in variables models.
- Efficient, distributed methods for anomaly detection.
- Generalized likelihood ratio test.
- Distributed implementation via message passing.
- C. Wei, A. Wiesel and R. S. Blum, Change detection in smart grids using errors in variables models, in Proc. of IEEE SAM 2012, special session on smart grid, NJ, June 2012.
- C. Wei, A. Wiesel and R. S. Blum, Distributed change detection in Gaussian graphical models, in Proc. of IEEE CISS 2012, NJ 2012.
- A. Wiesel and A. Globerson, Time varying autoregressive moving average models for covariance estimation, Submitted to IEEE Trans. on Signal Processing, March 2012.