Optimal Power Flow and Smart EV Charging

Optimal power flow problems are fundamental because they underly numerous power system operations.  OPF is nonconvex and NP-hard.  It is usually solved using local algorithms such as Newton-Raphson or convex relaxation, but neither guarantees globally optimal solutions.  Even though OPF is hard in theory, it seems ``easy’’ in practice in the sense that, empirically, both methods often yield global solutions.  In the first half of the talk, we present necessary or sufficient conditions for an OPF problem to both have exact relaxation and no spurious local optimal. 

 

In the second half, I’d describe a smart electric vehicle (EV) charging infrastructure and an open-source testbed at Caltech. We formulate the problem of optimally coordinating a network of EV chargers as a Quadratic Program, present some data from the deployment of our adaptive charging network. Finally, we describe the open-source ACN (Adaptive Charging Network) Research Portal that consists of large-scale EV charging data (ACN-Data), a realistic simulator (ACN-Sim), and the potential development of a real testbed (ACN-Live).  

 

(Joint work with Fengyu Zhao (Caltech), Zach Lee (Caltech/PowerFlex), Sunash Sharma (Caltech/Berkeley))