Monitoring-based prediction and electric vehicle charging in smart grid cities

Abstract

For the efficient coordination of diverse power entities in smart grid cities, this paper first develops prediction models for groundwater and wind speed, and then designs a battery management scheme for wind energy integration in electric vehicle charging. To estimate the power consumption in water management system, the massive data of day-by-day groundwater level readings, which have been accumulated for about 10 years in Jeju City, are fed to 3-layer neural networks for modeling the change pattern. The trained neural network yields an accurate groundwater level prediction with its maximum error bounded by 4.76%. In addition, for hourly wind speed predictions, 53% of forecast errors fall to the range of 0 to 0.05 in normalized size, in spite of occasional error spikes resulting from time lag. Finally, the battery operation according to the current and the next hour wind speed, not only improves the renewable energy gain by up to 9.1% but also obtains a stable gain curve for the given battery capacity range, compared with the deep cycle scheme.