Probabilistic Electric Vehicle Load Management in Distribution Grids

Increasing the number of electric vehicles (EVs) in the transportation sector can change the load pattern in power grids and affect its service quality and operational reliability. As EVs are mostly parked during the day, an EV load management (EVLM) system has been shown effective in mitigating those effects on power grids. The uncertainties in EV load and the computational burden of its coordination in the large-scale EV penetration, however, make EVLM challenging. In this paper, we address the former using the non-parametric diffusion kernel density estimation (DKDE) to estimate the expected charging energy demand and EV availability. To deal with the latter, we formulate EVLM considering the power flow constraints as a consensus problem which is solved by the alternating direction methods of multipliers (ADMM) in a distributed manner. Using real data, we evaluate the performance of DKDE and compare its results with Gaussian kernel density (GKDE). Owing to the smoothing properties of linear diffusion process and optimal bandwidth selection, DKDE is more adaptive to the training dataset and results in more accurate load estimation. Applying the prediction results to IEEE-13 bus system, the effectiveness of the proposed EVLM in nodal voltage improvement and feeder congestion mitigation is validated.