Control Systems Group: Research
We are a research group focusing on control systems, including both theories and applications. Our main research activities in control applications in recent years involve renewable energy, dynamics testing and battery power management systems. Our research objective is to develop novel efficient control strategies and use them to solve practical problems. The ongoing researches are as follows, and we are in the mean time looking for other promising directions.
Constrained optimal control
Many problems in social, biological and engineering systems can be mathematically formulated as constrained optimizations, that is, opimizing some outputs of the systems subject to the constraints from dynamics and other concerns such as safety issues. Constrained optimal control aims to solve the constrained optimization problem to provide a control input profile that can drive the system to behave optimally as desired. If the system model is linear time invariant and only subject to dynamic constraints, then an analytical solution can be derived using the classical linear optimal control theory. However, if the constraints on the system's input, state and output are present, the control problem would become complicated and in many cases analytic solutions cannot be derived. I am focusing on two constrained optimal control strategies:
- Model Predictive Control (MPC). MPC is the most extensively used advanced control strategy in industries. MPC resolves a constrained optimization problem online at each sampling instant. This feature enables MPC to achieve the best performance for multivariable systems when the constraints are present. We are interested in two vital issues that help promote MPC applications. One is to study the robustness of the MPC system when uncertainties are present; the other is to expedite the computational speed for large-scale systems.
- Anti-windup (AW) compensation. Actuators play an important role in real-time control systems. The AW technique helps to fully utilize the power of the actuators within their safety limits to improve the control performance. My research aims to propose novel AW compensation strategies and improve their robustness and performances.
Absolute stability analysis
This is one of the most fundamental learning branches in control theory since its birth. It is used to check the stability of the system when the nonlinearities are restricted to a specific region. Many celebrated control theorems (e.g. Popov Criterion, Kalman-Yakubovich-Popov lemma, etc) originated from it. We are interested in using advanced theoretical tools, such as integral quadratic constraints and linear matrix inequalities, to generalize the applicability of the existing methods (e.g. extend the Popov criterion with indefinite multiplier from SISO to coupled MIMO case), and propose new methods which are more computationally reliable but less conservative than the existing ones.
We are interested in applying control techniques to a wide range of sustainable energies. We are currently working on marine energy, especially wave energy.
Ocean waves provide an enormous, persistent and spatially concentrated energy compared with other renewable energy resources (e.g. solar and wind energies) . Despite enormous previous efforts on device design, wave energy generation is still a relatively immature technology for commercial purposes and is far from being competitive with traditional fossil fuel or nuclear sources of electricity.This is mainly caused by the following two problems:
- Inefficient energy extraction. The energy conversion abilities of most WECs are still not efficiently utilized.
- Risk of device damages. In order to prevent the WECs from being damaged during severe storms, the WECs have to be shut down. Especially during winter storms, such periods of inactivity can last for days.
Considering these main problems affecting the energy extraction, the objectives of WEC control designs are therefore: 1) to maximize the extracted power per annum and 2) to guarantee the safety of the WECs. The essential problem centred around this research is to reduce the cost of hardware and increase the energy output.
Dynamically Substructured System Testing Technique
The dynamically substructured system (DSS) is a hybrid testing method. The principle of dynamically substructured system (DSS) testing is receiving worldwide interest in various fields in recent years such as civil, aerospace, biological, automotive engineering. A major feature of the DSS method is that it allows physical components of a system to be tested, whilst the remaining parts are numerically simulated in real-time. This feature makes DSS testing superior than conventional pure physical and pure numerical testings. The success of a DSS test relies on a high-fidelity controller used to synchronize the numerical substructure and physical substructure of a DSS. When Dr Li was working in Bristol University, a hydraulically-actuated quasi-motorcycle mechanical system was built as a proof-of-concept test rig. To cope with different assumed testing scenarios for the test rig, a variety of control strategies (e.g. MPC, Anti-windup compensation, LQG, adaptive, neural network control) have been designed and successfully implemented. Moreover, we are also interested in the theoretical development for this research. For example, recently I have proposed a novel DSS framework on strictly separation of physical and numerical components. This novel DSS framework unifies almost all of the existing DSS application examples available, and significantly improved the DSS formulation and configuration. I am also working on the first ever DSS testing validation method, which would provide an efficient numerical tool to validate the performance of DSS testing.
Battery Management System
Batteries has been widely used as storage devices in many applications, such as aircrafts, laptops, hybrid vehicles, and renewable energy generation devices. Efficient control strategies are indispensable for improving the performance of battery management system (BMS) while reducing its hardware cost. However, designs and implementations of control algorithms are complicated by the nonlinearities and uncertainties inevitably introduced into the dynamic models due to the complex electrochemical dynamics. The constraints imposed on BMS, which are normally related to safety issues such as limits on current, voltage and temperature, increase the complexity of the resulting control problems. For example, one important problem for the batteries used in electricle or hybrid vehicles is optimal charging, which requies the shortest charging time withing the constraints of safety and aging process. The objective of this research is to develop efficient control strategies for BMS.