Here, at the Power Lab we work on a wide array of topics:
Power Systems : In this lab, emphasis is on control strategies which take advantage of “Internet of things” to adapt the grid to intermittencies introduced by renewable energy sources. We engage in stimulating research on State Estimation, Power System Protection, Control and Dynamics with potential for real-life application. Cyber-Physical security is one of the other important issues we work on.
State Estimation & Cyber Physical Security: Our group works on developing algorithms to detect, identify and correct gross errors on measurements, parameters and topology of power systems. Using state estimation and machine learning techniques, we develop algorithms that will ameliorate the assessment of technical losses and accurately detect and identify non-technical losses on distribution systems and cyber security of smart grids.
Demand Response & Consumer Behaviour: Development of distributed methods to control flexible demand to help mitigate the variability in renewable generation, analysis of cases when consumers are strategic and design of parameters in such as way that efficiency loss is bounded, analysis of the case in which consumers act irrationally.
Renewable Energy Aggregation: Allocation of costs based on cost causation principle and development of co-operative game for the realized profit of an aggregation of renewable energy producers.
Ancillary Services: Growing renewable integration introduces more uncertainty in the power grid and requires resources with fast response to balance supply and demand. Our research concerns harnessing the flexibility of loads to provide ancillary service, without significant impact on quality of service (QoS) to consumers.
Reinforcement Learning: Approximate dynamic programming algorithms to estimate value functions, and applying them to solve stochastic control problems. One such application is speed scaling: a power management technique in computer system design. Here, we are required to control the processing speed of the computer so as to optimally balance energy and delay costs. Value function approximation can be used here as a part of the policy iteration algorithm to optimize the processing speed such that the average energy and delay costs are minimized.