Resilience of urban public electric vehicle charging infrastructure to flooding

The study investigates how urban flooding, which creates geographically correlated outages, affects the resilience of public EV charging infrastructure and BEV mobility in cities. Although flooding’s impact on traffic has been widely studied, its implications for public EV charger availability and charging behavior remain underexplored. The authors focus on Greater London, a region with significant EV adoption targets, substantial areas at high flood risk, and a large share of drivers relying on public/on-street parking. The research question centers on whether flood-induced charger outages compromise BEV trip completion and how they redistribute charging demand across the network. The purpose is to quantify system-wide stress and accessibility changes and to propose siting strategies that enhance flood resilience of public charging networks. The importance lies in maintaining consumer confidence and facilitating sustained BEV adoption amid increasing climate-driven flood events.

Prior work has examined flooding impacts on traffic flows and road network vulnerability, as well as cascading effects across critical infrastructures, but has not addressed EV charging networks. Studies on charging adequacy assess infrastructure needs at various BEV penetrations without considering flood-induced outages. Policy and research have recommended avoiding charger siting in flood-prone areas, yet this may be infeasible in key commercial and residential centers and as flood zones evolve with climate change. Consequently, the literature provides limited insight into disruptions from simultaneous, geographically correlated charger failures and their implications for BEV adoption and user experience.

- Study area and data: Greater London. Building and road network data from OpenStreetMap (OSM). Public charger locations (slow <43 kW; rapid ≥43 kW) from the Mayor of London’s dataset, totaling 5925 chargers. Buildings were classified (residential/work/commercial), using OSM tags where available and probabilistic assignment otherwise. Road network connectivity was cleaned and contracted; buildings and chargers were mapped to nearest road nodes. - Travel and vehicle modeling: Personal BEV trips simulated over 24 h using Transport for London travel survey statistics (trip counts, purposes, departure-time distributions, speeds). Vehicles begin and end at home, with an average of 2.42 trips per BEV and a mean trip Haversine distance ~15.8 km. BEV fleet sized at 6 per charger for main results (consistent with 2020 stats), with three BEV models (Nissan Leaf 40 kWh, Leaf Plus 62 kWh, Tesla Model S 100 kWh) assigned uniformly. Initial state of charge (SOC) distributions reflect dependence on residential, public night-time, or daytime public charging. SOC is updated by trip energy consumption and charging events. - Charging behavior and assignment: Users initiate charging after trips if SOC < 0.5; if SOC < 0.3, they search for the nearest available charger at any distance; otherwise, they prefer an available charger within 300 m of the destination. First-come-first-served assignment. Charging only proceeds if ≥30 min available before next scheduled trip. Chargers modeled with 89% efficiency; slow chargers at 12.5 kW and rapid at 43 kW. A BEV is considered to have failed if SOC < 0.2 at any time; day-time public-charging dependent BEVs also fail if end-of-day SOC < 0.3. - Flood risk and scenarios: Regions at risk from flooding were identified using Climate Central’s coastal risk screening tool for 2030 (sea level rise + moderate flood, current trajectory, medium luck), filtered to Greater London and rasterized onto a 42 × 86 grid; all grids overlapping at-risk areas were flagged. For three flood scenarios, chargers in at-risk grids were taken offline with probabilities p = 0.5 (scenario-1), 0.7 (scenario-2), and 0.9 (scenario-3). Night-time public-charging users in flooded grids began with lower SOC [0.4, 0.6]. Flooded states were held constant throughout the simulated day. - Simulation resolution and metrics: 1-min time-step event-triggered simulation over 24 h. Key metrics: (i) charger utilization (fraction of the day a charger is occupied), and (ii) distance between a trip’s intended destination and the nearest available charger actually used (proxy for accessibility/comfort). Aggregations were reported by grid and as functions of distance from nearest flooded grid. Each result was averaged over 100 Monte Carlo simulations with randomized driving patterns and flooded charger realizations. - Mitigation strategies: Four siting strategies for additional chargers (all added as rapid chargers) tested primarily at +5% new chargers (with sensitivity at +2.5% and +10%): (1) ring-fencing: place new chargers just outside flooded regions; (2) usage-dependent: place in top 2% grids by baseline utilization; (3) distance-based: place near far-distance peaks observed in impact distributions; (4) random: distribute uniformly at random across the city.

- Trip completion resilience: Even with flood-induced outages, >99.7% of rides completed successfully under all scenarios, likely due to intra-city travel distances. This held despite over 34% of public chargers taken offline in some simulations. - Network stress redistribution: Mean charger utilization across the entire network decreases during flooding (reflecting offline chargers), while maximum utilization increases, indicating uneven stress with some chargers becoming heavily loaded. Flood scenario maximum utilizations rose from a baseline of 59.7% to 62.7%, 63.2%, and 64.9% (scenarios 1–3). Minimum utilization remained 0% across scenarios. - Geographic propagation: Outside flooded grids, most areas experience increased utilization and reduced accessibility. Peaks in stress occur both immediately outside flooded regions and surprisingly far away (~10–13 km). In these distant peaks, grids can see up to +50.9% charger utilization and +269.9% increase in walking distance from destination to the nearest available charger as users defer charging from flooded zones to later trips. - Concentration on already-stressed areas: Changes in both utilization and access are strongly positively correlated with baseline values in grids where impacts are positive. Pearson correlations between baseline utilization and change in utilization: r = 0.65 (p = 6.21e-65), 0.60 (p = 1.71e-53), 0.52 (p = 1.30e-38) for scenarios 1–3. For distance to nearest available charger: r = 0.74 (p = 8.59e-234), 0.73 (p = 1.83e-238), 0.71 (p = 3.71e-229). Impacts also correlate with building density, indicating demand-driven vulnerability. - Mitigation performance with +5% new chargers: All strategies improved accessibility (reduced destination-to-charger distances). Random placement yielded city-wide improvements with peak distance reductions of 26.7% (near flooded region) and 45.3% (farther). Ring-fencing and distance-based strategies achieved targeted reductions at their respective peaks: 35.9% (near peak) and 42.8% (far peak), with limited effect on the other peak. For utilization stress, usage-dependent siting performed best, reducing the far-peak utilization by up to 24.4%; ring-fencing reduced the near-peak by ~5.7%. Across the entire network, peak utilization decreased under all strategies; usage-dependent also increased mean utilization, suggesting more efficient network use. Distance-based siting had minimal effect on utilization peaks even with +10% chargers (~0.3% reduction), likely due to poor overlap with stressed areas. - Policy-relevant quantitative guidance: Approximately 5–10% additional chargers are needed to restore peak utilization to baseline values under flooding. Usage-dependent siting best enhances resilience, accessibility, and efficiency; ring-fencing central London (areas around the Thames) provides targeted relief; random placement can bolster accessibility at the city’s edges.

The study shows that while urban flooding does not significantly compromise in-city BEV trip completion, it creates geographically correlated outages that propagate stress through the public charging network, especially burdening already-busy areas and diminishing user accessibility. This redistribution underscores a critical vulnerability: recurrent flood events may undermine user confidence among those depending on the most affected chargers, potentially dampening BEV adoption. Mitigation requires pre-emptive placement of additional chargers outside flood-prone regions. Usage-dependent siting most effectively reduces peak utilization where demand is high, aligning new capacity with need and improving overall efficiency. Ring-fencing the flood-prone core (e.g., central London near the Thames) can blunt near-field peaks, while random placement improves accessibility in peripheral areas; elements of these strategies can be combined to balance equity, resilience, and feasibility. As flood risks evolve with development and climate change, city-specific analyses should be periodically updated, co-planning upgrades to distribution infrastructure as charging capacity expands. The findings directly address the research question by quantifying how flooding redistributes demand and identifying targeted interventions that restore resilience without materially impacting BEV mobility.

This work quantifies the resilience of Greater London’s public EV charging network to flood-induced, geographically correlated outages. It finds negligible impact on BEV trip completion but substantial, uneven increases in charger stress and reduced accessibility, including peaks 10–13 km from flooded zones. The impacts concentrate on already-stressed locations. Four mitigation strategies were proposed and evaluated, with usage-dependent siting performing best for reducing utilization peaks and improving efficiency; ring-fencing effectively suppresses near-field peaks, and random placement improves accessibility more broadly. Approximately 5–10% additional rapid chargers can restore peak utilization to baseline under flooding. Future research should incorporate high-resolution, city-specific cost, revenue, and flood-risk data to optimize siting via multi-criteria methods; account for dynamic flooding; and consider interactions with power systems and evolving travel patterns to holistically plan resilient charging infrastructure.

- Travel demand assumptions: The number of rides and trip destinations are assumed unchanged by flooding, appropriate where floods are an inconvenience rather than a deterrent to travel. - Flood dynamics: Flooded grids and charger outages are static over the day; dynamic flood evolution is not modeled. - Charger-BEV interoperability: Assumes no compatibility issues; any BEV can use any charger. - Charger characteristics: All added chargers for mitigation are modeled as rapid chargers; real-world mixes may vary. - Data granularity and vintage: Travel patterns are based on older survey data; some OSM building classifications were inferred probabilistically. - Scope: Focus on intra-city travel; does not model power distribution constraints, grid impacts, or economic factors (installation costs, revenues). Flood risk was primarily based on coastal projections; other sources are discussed in supplementary analysis.