The group focuses on computational methods for the analysis and control of large scale (bulk) power systems, state-estimation, ancillary services for micro-grids and distribution networks interfaced with renewable sources, and the design of efficient power-electronic interfaces for renewable energy resources.
Current research themes:
Due to significant line resistances in a microgrid, active power variations produced by doubly fed induction generator (DFIG) based generation translate to corresponding power quality problems. We have proposed simple localized control schemes exercised through the rotor side converters for voltage regulation. Detailed dynamic simulations accounting for wind shear and tower shadow on a IEEE-13 bus distribution system conducted in EMTDC/PSCAD show that tight regulation (<1%) can be achieved with the proposed method compared to ~5% without control. We have also examined the effectiveness of energy-storage methods on frequency regulation in such systems. Our results show that 10% (with respect to the machine rating) storage is effective at restricting frequency deviations within 1%.
Computational methods for large to very-large scale power systems
Our goal is to improve the efficiency and scalability of algorithms used in the analysis, operation and control of electric power systems. In the future, static and dynamic analysis on large-scale (thousands of buses) to very-large scale (tens of thousands of buses) power systems may need to be conducted in (near) real-time. Our current focus is on two important problems: (a) reconfiguration in distribution networks and (b) identifying extreme contingencies in bulk power systems.
Extreme Contingency Screening
Determining contingencies that can potentially culminate in a blackout is a challenging computational task in large-scale power systems due to the combinatorial nature of the problem. Prediction tools to identify such cases generally restrict the search to scenarios with severe consequences that may arise from relatively few contingencies. One such approach is based on graph partitioning where the ratio of power imbalance to cut-size serves as a heuristic measure to identify critical line outages. We proposed an algorithm based on unbalanced-constrained partitioning and developed a public domain tool pro-PART that is: scalable, efficient and simple to implement. For example, the average CPU run-time is less than 0.1 s and 1.5 s for systems with 2831 buses (large-scale) and 43,501 buses (very large-scale) respectively. We are currently working on extensions of this approach to identify a sequence of contingencies that may culminate in a blackout.