The sustainability of a clean energy transition for electric vehicle transportation is clearly affected by increased energy consumption cost, which is associated with large-scale electric vehicles (EVs) charging on a fossil-fuel dependent electricity grid. This places a potential threat on the safe operations and stability of the grid and increases the emissions of greenhouse gases (GHGs) from the power stations that generate the electricity. Furthermore, the uncontrolled large-scale integration of EVs charging on the grid will increase exponentially in the coming years. Because of this, new peaks on the grid will be generated due to the EV charging load variance, and a significant impact on the transformer limit and substation capacity violation will occur. To mitigate the significant impact of the high cost of energy consumption by large-scale EVs charging on the grid, and to reduce the emissions of GHGs, there is a need to provide a multi-level optimization approach that is robust and dynamic to solve the uncontrolled charging problem of large-scale integration of EVs to the grid. This paper investigates the grid energy consumption by EVs and reviews recent applications of EV charging controls and optimization approaches used for the energy management of large-scale EVs charging on the grid. Energy management in this context is not trivial. It implies that the objectives such as load shifting, peak shaving, and minimizing the high cost of electricity consumption with a stable grid operation can be achieved. In the context of this study, EVs charging on the grid includes both battery electric vehicles (BEVs), which have larger battery banks with a longer charging duration and higher energy consumption capacity, and plug-in hybrid electric vehicles (PHEVs) which have smaller battery capacities.

The increased demand for electricity is inevitable due to transport sector electrification. A major part of this demand is from electric vehicle (EV) charging on a large scale, which is now a growing concern for the grid power distribution system. The lack of insight into grid energy demand by EVs makes it difficult to manage these consumptions on a large scale. For any grid load management application to be effective in minimizing the impact of uncontrolled charging, there is a need to gain insight into EV energy demand. To address this issue, this study presents data-driven modeling of EV charging sessions based on machine learning (ML) techniques. The purpose of using ML as an approach is to provide insight for estimating future energy demand and minimizing the impact of EV charging on the grid. To achieve the aim of this study, firstly, we investigated the impact of large-scale charging of EVs on the grid. Based on this, we formulated an objective function, expressed as a sum of utility functions when EVs charge on the grid with constraints imposed on voltage levels and charging power. Secondly, we employed a graphical modeling approach to study the temporal distribution of EV energy consumption based on real-world datasets from EV charging sessions. Thirdly, using ML regression models, we predicted EV energy consumption using four different models of fine tree, linear regression, linear SVM (support vector machine), and neural network. We used 5-fold cross-validation to protect against overfitting and evaluated the performances of these models using regression analysis metrics. The results from our predictions showed better accuracy when compared with the results from the work of other authors.

Electric vehicle (EV) charging from the grid presents significant challenges for the grid due to increased power demand added to the existing base load. This scenario leads to grid instability arising from voltage drops and new peaks generated due to uncontrolled charging. This study proposes an artificial intelligence (AI)-based optimization approach to mitigate the impact of large-scale, uncontrolled charging of EVs. Using real-world datasets of EV charging sessions, we first quantify the impact of EV charging under a baseline “uncontrolled” charging scenario using the IEEE 33-bus test system. We formulate the charging problem as an optimization task with the objective of minimizing energy consumption and peak power demand. A genetic algorithm (GA) and model-predictive control (MPC), combined with solar PV, are employed as charging strategies to solve the optimization problem. Simulation results show that the AI-driven strategy significantly reduced peak power demand by 38.35% and energy consumption by 12.09% with solar PV utilization. A sensitivity analysis of peak power demand reveals that grid stability will be violated with higher charging rates.