Engineering Management Seminar Series: Yada Zhu
Friday, September 13, 2013 – 11:00 am
Location: Babbio 310
SSE Invited Talk
A Case Study of Maintenance Optimization for Electricity Distribution Utilities using Simulation Strengthened Analytics
Research Staff Member - Department of Business Analytics and Mathematical Sciences, IBM Thomas J. Watson Research Center
Friday - September 13 - 11:00 AM
ABSTRACT: Power Utilities are among the most capital-intensive industries in the world, where even modest improvements in asset maintenance and capital investment scheduling can potentially lead to very large annual cost savings in the order of millions. Such an improved scheduling model requires asset condition data, historical asset data, SCADA data, etc. However, lack of sufficient visibility into the asset condition and insufficient, disjointed or mismanaged historical data records are the major challenges for this improved asset management solution. The largest power distribution company in the Netherlands has collaborated with IBM to develop algorithms, which could optimize asset maintenance schedule and capital investment.
A holistic approach is developed to address the challenges for an improved asset management solution by combining the strength of both power network simulation and big data analytics, which is an emerging technology within the utilities. The essential new elements to this approach are:
(1) A data-mining technique to build time and space coherent data models from historical data records with a wide-range of time scales.
(2) An integrated power network simulation model to provide an additional data source for assets that partially or totally lack of instrumentation.
(3) A physical model based asset health condition assessment to quantify the risk associated of each asset based on its type, location, historical usage, and environment conditions.
(4) An effective algorithm that determines the optimal asset maintenance schedules (i.e., minimize the expected cost caused by asset failure) under constraints such as budget and personnel capacity.
Preliminary results have been achieved and tested on realistic power company data for optimizing secondary transformer maintenance in the Netherlands. The results have shown that this approach can provide near real-time visibility into asset conditions, conduct forward-looking what-if analysis for better grid planning, and reduce the operating costs including the cost associated with asset failure significantly. This can translate to multi-million dollar savings in operational costs under a fixed asset management budget.
BIO: Yada Zhu is a Research Staff Member in the Department of Business Analytics and Mathematical Sciences, IBM Thomas J. Watson Research Center, NY. She received her Ph.D. degree from the State University of New Jersey, Rutgers in 2010. Her research interests include survival data analysis, design of reliability testing plans, knowledge discovery with applications in monitoring and control of semiconductor manufacturing processes, and condition assessment of utility and energy systems. She has participated and lead a number of Smarter Water, Smarter Building and Smarter Energy projects. She is a senior member of IEEE, and member of ASQ, IIE and INFORMS.