Mahnoosh Alizadeh

Demand Response and Renewable Energy Integration
Power system reliability requirements dictate that supply and demand have to be balanced at all times. The intermittency of renewable energy supplies (e.g., solar and wind) threatens the ability of grid operators to ensure this balance. That is why enabling technologies that harness the intrinsic flexibility of end-use electricity demand is an inevitable step for efficient trading of high levels of non-dispatchable renewable energy resources. However, electricity demand has historically been left completely out of the control system design. In order to make demand more elastic, we need to re-imagine the end-use experience of electricity delivery services and how we operate electricity markets. Engaging demand in the loop poses a highly complex control, learning, and market design problem. Electricity demand is comprised of a large number of heterogeneous subcomponents that interact through a complex, coupled physical environment operating over many spatial and temporal scales. These subcomponents are also serving the needs of customers with heterogeneous preferences. In our research, we aim to design scalable and decentralized protocols that dictate how a large population of flexible appliances owned by different customers can engage with the grid, with the goal of achieving network-wide near-optimal performance and providing the highest possible quality of service to customers.

Electric Transportation Systems
The mobility scene is going to change rapidly in the coming years as electric vehicle (EV) adoption rates increase, ride sharing continues to grow, and autonomous vehicles proliferate. The fact that these changes coincide with the smart grid revolution is both a great opportunity and a challenge. Without infrastructure interoperability, it would be challenging to manage the effect of a growing number of EVs on the power grid. As a consequence, EV charging patterns could create many issues for power transmission and distribution systems, and reduce the environmental benefits of electrification. Another related issue is that we currently lack adequate EV charging stations in less populated areas and practical control mechanisms to allocate charging spots to EVs, leading to range anxiety in drivers on some routes as well as possibly long wait times to find a spot at popular locations.

In the past few years, our group has been working on the design and testing of real-time optimization and network control algorithms for mobility-aware smart charging that allow power and transportation networks to cooperatively minimize the carbon footprint of EVs. We have considered various mobility scenarios, including EV fast charging stations, workplace charging facilities, electric vehicle fleets, and autonomous mobility on demand systems.

Safe Learning in Cyber-Physical Systems
Learning and optimization algorithms have found many applications in systems that repeatedly deal with unknown stochastic environments and seek to optimize a long-term reward by simultaneously learning and exploiting the unknown environment. They are also naturally relevant for many cyber-physical systems with humans in the loop (e.g., pricing end-use demand in societal-scale infrastructure systems such as power grids or transportation networks to minimize system costs given the limited number of user interactions possible). However, existing learning and optimization algorithms might not be directly applicable in these latter cases. One critical reason is the existence of safety guarantees that have to be met at every single round when interacting with the environment. For example, when managing demand to minimize costs in a power system, it is required that the operational constraints of the power grid are not violated in response to our actions. Thus, for such systems, it becomes important to develop new learning and optimization algorithms that account for critical safety requirements.

Resiliency in Multi-agent Networks
Networked cyber-human-physical systems such as the smart grid or intelligent transportation systems are increasingly relying on distributed protocols based on optimization and game theory. The distributed and networked nature of these protocols makes these systems susceptible to external influences which, if left untreated, can arbitrary lower system efficiency. In our work, we design new protocols or strategies to reduce the vulnerability of distributed multi-agent networks to external manipulation. We have specifically considered this challenge in the context of graphical coordination games as well as distributed optimization algorithms.