Beyond the Charging Point
PEV Integration and Planning in Power Distribution Systems
Time: Mon 2026-12-14 10.00
Location: Kollegiesalen, Brinellvägen 8, Stockholm
Language: English
Subject area: Energy Technology
Doctoral student: Priscila Costa Nascimento , Kraft- och värmeteknologi, Distributed Energy Resources and Smart Energy Networks
Opponent: Professor Filip Johnsson, Chalmers tekniska högskola
Supervisor: Professor Björn Laumert, Kraft- och värmeteknologi; Dr. Silvia Trevisan, Kraft- och värmeteknologi
Abstract
The electrification of road transport is introducing substantial new demand into power distribution systems. The challenge is not only the additional demand but also its uncertainty: when vehicles charge, where they connect and how much power they draw vary across users and locations. As plug-in electric vehicle (PEV) adoption increases, these uncertainties affect both network operation and the assumptions used to plan and dimension distribution infrastructure. This thesis represents those uncertainties with a locally adaptable probabilistic model of PEV charging and follows them through to grid impacts, mitigation and charging-control strategies, and peak-demand estimation and transformer-sizing decisions.
A vehicle-level Monte Carlo charging model represents the spatial and temporal variability of PEV charging using locally representative mobility, vehicle, charging and electricity-demand data. The model is combined with quasi-static time-series power-flow analysis and applied to detailed models of real power distribution systems in Brazil and Sweden, ranging from low-voltage networks to a large urban medium- and low-voltage system in Stockholm County. Evaluation against one year of measured home charging shows that, with locally derived inputs and revised charging rules, the model reproduces annual energy and peak demand within approximately 3% and 5%, respectively.
Across the case studies, the results show that the impact of PEV charging depends not only on the number of vehicles connected but also on when and where they charge. Voltage quality emerges as the principal constraint in both impact case studies, despite differences in country, network topology, voltage level and regulatory framework. The impact does not grow in proportion to PEV adoption because network violations depend on charging events coinciding in time and concentrating at electrically weak locations. Mitigation and charging control are also not inherently beneficial: measures that neglect broader network effects can aggravate operating conditions. On the network side, distribution-transformer tap adjustments can shift undervoltage into overvoltage, whereas low-voltage regulators substantially reduce voltage violations and provide a short-term option as PEV adoption increases. On the demand side, charging that minimises owners' costs against a common electricity price can synchronise demand. Under the published case-study scenarios, this reduces hosting capacity, the largest share of PEVs the network can accommodate, from 27% under uncontrolled charging to 8%. Combining the same price signal with a network peak-demand constraint instead raises it to 49%, while also reducing network stress and, at higher adoption levels, charging costs.
Increasing PEV adoption also changes the characteristics of the demand that distribution-system planning seeks to predict. At secondary substations, load coincidence initially decreases and then increases as electrification progresses. This non-monotonic, U-shaped behaviour changes the relationship between annual energy consumption and coincident peak demand, so patterns derived from historical demand do not remain representative throughout the transition. Consequently, conventional peak-demand estimation methods become less reliable, particularly during intermediate stages of electrification. Data-driven models that explicitly account for PEV-related demand characteristics reduce peak-demand prediction error by up to 70% relative to Velander's formula, traditionally used in Sweden to estimate peak demand. Prediction accuracy alone, however, is not sufficient for choosing an estimation method. Methods that err in the same direction with similar frequency can have substantially different consequences for transformer sizing because the magnitude of those errors also matters. Estimation methods for planning should therefore be judged by how often they under- or overestimate the required capacity and by how large those errors are, particularly as increasing PEV adoption reduces transformer capacity margins.
Overall, the thesis shows that accommodating transport electrification requires methods that account for uncertainty from individual charging behaviour to distribution-system operation and planning. First, probabilistic assessment is needed to identify where and when network constraints emerge. Second, flexibility and mitigation measures must be evaluated with those constraints represented explicitly. Third, peak-demand estimation methods must remain representative as electrification changes load coincidence. The thesis provides methods that meet these three requirements, so that planners can assess when existing infrastructure can accommodate further electrification, when operational flexibility or targeted mitigation may be sufficient, and when reinforcement or revised dimensioning should be considered. In this way, the methods help PEV adoption grow without compromising the reliable operation of the network.