IT Seminar: Souhaib Ben Taieb

calendar icon

Event Details

Date:
10 May 2017 at 2:00 pm – 10 May 2017 at 3:00 pm
Venue:
Clayton CL_26_G12A, VC to 7.84 Caulfield

Description

Title: A probabilistic forecasting algorithm for hierarchical electricity smart meter data

Speaker: Souhaib Ben Taieb

Souhaib Ben Taieb

Abstract: Forecasting electricity demand is critical for electric utilities in order to undertake appropriate planning of generation and distribution. Recently, the large-scale deployment of smart electricity meters has made available a large amount of time series data representing household electricity demand at intervals from 1 minute to one hour. These time series can often be represented in a hierarchical or grouped structure comprising a set of time series along with aggregates of subsets of these series. For example, the electricity demand for a whole country can be disaggregated by states, cities, regions and households. Although forecasts can be produced independently for each series in the hierarchy, typically this does not lead to coherent forecasts - the property that forecasts add up appropriately across the hierarchy. State-of-the-art hierarchical forecasting methods usually reconcil these independently generated forecasts to satisfy the aggregation constraints. A fundamental limitation of prior research is that it has looked only at the problem of forecasting the mean of each time series, while probabilistic forecasts are often required to quantify the uncertainty of future demand. We consider the situation where probabilistic forecasts are needed for each series in the hierarchy. We define forecast coherency in this setting, and propose an algorithm to compute predictive distributions for each series in the hierarchy. Our algorithm has the advantage of synthesizing information from different levels in the hierarchy through a sparse forecast combination and a probabilistic hierarchical aggregation. We evaluate the accuracy of our forecasting algorithm on hierarchical electricity smart meter data. The results show consistent performance gains compared to state-of-the art methods.

Short bio: Dr. Souhaib Ben Taieb is a Lecturer in Statistical Machine Learning in the Department of Econometrics and Business Statistics at Monash University. His research interests include statistial machine learning, time series analysis, probabilistic forecasting, big data processing and smart  grid analytics. He received an M.Sc and a Ph.D. in Computer Science with a specialization in Machine learning from the Free University of Brussels in Belgium. He was a postdoctoral research fellow in the Spatio-Temporal and Data Science Group at KAUST. He was a visiting scholar at several international institutions, including the School of Earth, Energy and Environmental Sciences, Stanford University, and the Said Business School, University of Oxford. Dr. Ben Taieb has received multiple awards including an Early-Career Faculty Research grant from Monash Business School, the Solvay Awards 2015 for his PhD thesis, an IEEE Power & Energy Society Award for the Global Energy Forecasting Competition 2012, and a Doctoral research fellow grant from the Belgian National Fund for Scientific Research.


Event Contact

Name
Geoff Webb
E-Mail
Geoff.Webb@monash.edu