Data for heuristic optimization of electric vehicles’ charging configuration based on loading parameters

Abstract

This dataset includes multiple files related to optimization of electric vehicles to minimize overloading in low voltage grids by varying the locations available to charge the EVs. The data include lognormally sampled hourly sorted scenarios across 11 charging locations for a stochastics-based Monte Carlo simulation. This simulation runs through 2 million scenarios based on actual probabilities to incorporate most possible situations. It also includes samples from normally distributed household electricity use scenarios based on agent-based modeling. The article includes the test grid parameters for simulation, which were used to create a benchmark grid in DigSilent Powerfactory software, as well as intermediate outputs defining worst case scenarios when electric vehicles were charged and results from three different optimization approaches involving a reduction in voltage drops, cable overloading and total line losses. The outputs from the benchmark grid were used to train a machine learning algorithm, the weights and codes for which are also attached. This trained network acted as the grid for subsequent iterative optimization procedures. Outputs are presented as a comparison between pre-optimization and post-optimization scenarios. The above dataset and procedure were repeated while varying the number of EVs between 0 and 100 in increments of 20, data for which are also attached. The data article supports a related submission titled “Minimization of Overloading Caused by Electric Vehicle (EV) Charging in Low Voltage Networks”.