Automatic management and control of stationary energy storage systems (ESS) are a prerequisite before the value of connecting ESS into a grid or a microgrid can be fully realized. Various types of batteries can be used for energy storage to bring benefits such as grid stabilization, peak demand shaving, or a market participation (combined with an associated renewable intermittent energy source). In all cases, information available to any control system is subject to uncertainties, including due to uncertain and variable user demand forecast, environmental forecasts (wind and solar activity), grid demand. This intrinsic uncertain nature of the environment where the ESS operate means that any control strategy for ESS needs to be able to handle uncertainties and noise efficiently for closed-Â¬loop operation. To be able to truly benefit from renewable energy sources and ESSs integrated into the energy network, a suitable control strategy needs to exist, able to handle uncertainties present in real¬-life application. The control algorithm needs to be combined with an estimation algorithm ("digital twin") that would track the changes in the system due to degradation or unknown changes in the operating conditions. However, when considering uncertainties within optimal control settings, there are oftentimes very high computational requirements associated with running the controllers. Given the nonlinear characteristics of many physical systems, a nonconvex optimization problem ensues, for which a costly global solver is needed, adding up to the computational burden. The use of local solvers has recently been proposed to deal with similar settings. That work, though, does not derive theoretical feasibility guarantees and while a local solver is used, the resulting optimization problem is still very large.
As part of this project, the student will be expected to
Undertake a literature review of energy storage system (ESS) modelling and an investigation into the nature of the relevant uncertainties (e.g., environmental or demand) and controller-¬oriented uncertainty characterisation (e.g., a probability distribution, scenarios).
Develop robust optimal control design methods for nonlinear dynamical systems.
Derive open-¬loop and/or closed-¬loop guarantees.
Develop methods that exploit the structure in the optimal control problem, given the nature of available information about the uncertainties (e.g., probability distribution) and/or the type of system nonlinearity based on ESS models in the literature.
Undertake a simulation¬-based study of the proposed robust controllers to a microgrid or a grid-¬connected energy system with renewables, ESS, users, etc., subject to uncertainties, using relevant models from the literature.
Applicants should have a first-class Master's degree (or equivalent) in Electrical/Electronic Engineering, Computing, Mathematics, or related areas. Suitable backgrounds for these PhD positions include, but are not limited to, control engineering, mathematical optimization and power engineering. They should be highly motivated individuals with a keen interest in conducting interdisciplinary research. Students must also meet the eligibility requirements for Post-Graduate Studies at Imperial College London.
Applications are invited for a PhD studentship, to be undertaken at Imperial College London (Control and Power Group, Department of Electrical and Electronic Engineering). This studentship is funded by an EPSRC iCASE award and industrial partner SLB. As part of the project, the student will work closely with staff at the SLB Cambridge Research Centre, have access to their experimental and computational facilities, and benefit from mentoring and supervision of SLB staff.
The project will be supervised by Prof. Eric Kerrigan (Professor of Control and Optimization, Imperial College London) and Dr Anna Sadowska (Senior Research Scientist, SLB).