Congratulations to Olivier Mulkin for his successful advancement

Committee: Mike Ludkovski (chair), Mengyang Gu, Tomoyuki Ichiba

Title: Distribution Grid Uncertainty Modelling using Bayesian Optimization

Abstract: 

On top of the uncertainty in electricity demand, operators of electrical distribution networks need to account for the risk related to the rapidly increasing adoption of Distributed Energy Resources (DERs) throughout the grid. In particular, rooftop solar photovoltaic (PV) panels are particularly challenging as they turn consumers into producers, and nodes of power consumption in the network may become sources and transfer their excess generated power to the grid. Current industry practices fail to properly incorporate this uncertainty into their grid planning and investing decisions and only rely on myopic worst-case scenarios by aggregated PV capacity which can lead to underestimation of PV-related grid stress. Inadequate planning for such risks can lead to consequences, such as damage to grid equipment and hazards, such as fire risk or brown-outs. Developing the first statistically-based methodology to uncover and quantify PV-adoption related risk to distribution grid operation and planning decisions is the purpose of the current project. By modeling grid stress with Gaussian Process (GP) surrogates, and using a multi-objective Bayesian Optimization algorithm to tackle the computationally prohibitive evaluation of grid stress, which we treat as a black-box, our algorithm successfully identifies critical scenarios of PV adoption. Furthermore, a methodology based on GP simulations is leveraged to quantify the uncertainty associated with those risk-dominating scenarios. This is a joint work with Lawrence-Berkeley National Laboratory.