Why Central Planning Doesn't Break the Grid: New Research on EV Charging Flexibility
The Grid's Secret Problem
Somewhere in the near future, when you plug in your electric car to charge overnight, you might assume that clever software somewhere is making sure everything runs smoothly. That the grid knows what it's doing. And mostly, you'd be right—but only mostly. A significant gap lurks between how the power system is planned and how it's actually used, and that gap is starting to matter as electric vehicles flood onto our roads.
Here's the tension: when engineers plan the national electricity grid, they think in enormous abstractions—gigawatts of solar power, massive wind farms spanning hundreds of kilometers, the flow of electrons across thousands of kilometers of transmission lines. But when you actually charge your car at home, physics doesn't care about those abstractions. The electricity arrives through your neighborhood's distribution lines, through a specific transformer serving your block, at a voltage that can sag if too many neighbors plug in at once.
A new study from researchers at ETH Zürich and Delft University of Technology cuts straight to this problem. Published on arXiv, it asks a deceptively simple question: if you optimize when millions of electric vehicles should charge—scheduling their energy intake to balance the grid and absorb renewable power—does that optimization actually survive contact with reality at the local level? Or does it fall apart when rubber meets road, when thousands of individually owned vehicles respond to the same signals in ways that create new problems on your street?
The answer, according to this research, is more hopeful than you might expect. But it's not a clean victory—it's a careful accounting of trade-offs, a map of when central planning works and where it needs help.
The Science: Two Scales of the Same Problem
To understand what the researchers did, you first need to grasp why this problem exists at all. The electricity grid operates on two nested levels that most people never think about. The transmission grid is the interstate highway system—high-voltage lines carrying power across continents, connecting massive power plants to regional substations. The distribution grid is the local road network—lower-voltage lines branching into neighborhoods, eventually reaching individual homes and businesses.
When researchers study how to integrate renewable energy, they typically focus on one level or the other. Transmission-level studies ask: how do we move power from wind farms in Wyoming to cities on the coasts? Distribution-level studies ask: how do we keep voltage stable on a particular feeder line when fifty houses install solar panels?
But electric vehicles are peculiar because they exist at both levels simultaneously. From the transmission perspective, they're flexible loads—devices that can shift when they draw power without changing the total energy they consume. This flexibility is incredibly valuable. Solar panels generate electricity during the day when demand is often lower than supply. Wind often peaks at night. Without flexibility, the grid has to either waste this clean energy or fire up fossil fuel plants to balance supply and demand. Electric vehicles, sitting parked most of the day, could soak up that excess renewable energy and return it during peaks. It's a beautiful theoretical solution.
From the distribution perspective, though, those same vehicles are physical devices drawing power through wires and transformers designed decades ago for a world without EVs. Charge too many cars on one street simultaneously, and you get voltage drops, transformer overheating, and lines stressed beyond their limits. The aggregate looks clean at the transmission level; the reality on the ground looks messier.
The research team—Ambra Van Liedekerke, Lorenzo Zapparoli, María Parajeles Herrera, Blazhe Gjorgiev, and Gabriela Hug—designed a study to test whether optimization done at the transmission level actually translates into good outcomes at the distribution level. They built a modeling framework that captures both scales and how they interact.
Their methodology involved creating a synthetic but realistic test system. They modeled a portion of the European grid: four national transmission systems (Belgium, the Netherlands, Luxembourg, and a region representing France's northern neighbors) connected through high-voltage lines, with distribution networks beneath them containing the actual households and vehicles. They populated this virtual world with realistic patterns of vehicle ownership, driving behavior, and grid infrastructure.
The driving data came from the German Mobility Panel—a large-scale survey that tracks actual travel patterns. This matters because EV charging optimization is only useful if it respects how people actually use their cars. If your optimization assumes drivers only need their car between 8 AM and 6 PM but your customers work night shifts, you've built a system that fails in practice. The researchers used these travel patterns to establish when each vehicle was available to charge, how much energy it needed, and what departure time it had to be ready.
They then simulated three scenarios. The first was uncontrolled charging—the dumb default where everyone plugs in when they get home and the car starts drinking power immediately. The second was a top-down approach: a central planner optimizes charging across the entire transmission system to maximize renewable integration, then that aggregate plan gets disaggregated down to individual vehicles, respecting distribution constraints and individual driving requirements. The third was a bottom-up approach: each local distribution network optimizes its own EV charging using electricity prices derived from the centralized planning model, with no direct coordination from above.
The key question was whether these approaches produced charging patterns that were simultaneously good for the transmission system (absorbing renewable energy) and good for the distribution system (not causing local overloads). The bottom-up approach tested whether price signals alone—essentially, telling drivers "charge now because power is cheap"—could coordinate millions of individual decisions to achieve the global optimum without any central controller dictating each individual vehicle's schedule.
What They Found: Planning Translates, With Nuance
The results tell a story of qualified success for central planning—but with important subtleties that matter for how we build the EV-friendly grid of the future.
The top-down approach delivered exactly what it promised: EV charging schedules that respected both the grid's needs and individual driving requirements. When the central planner decided it was optimal to shift charging to the middle of the night—when wind power was plentiful and demand was low—that shift actually happened. The disaggregation process ensured that each vehicle's constraints (it needs to be charged by 7 AM for the commute, it can only accept a certain charging rate, etc.) were honored while still hitting the system-wide targets. Compared to uncontrolled charging, the top-down approach dramatically reduced what the researchers call "grid violations"—situations where voltage sagged below acceptable levels or equipment was loaded beyond safe thresholds.
This matters because it's not obvious that it would work. A central planner might decide, in aggregate, to shift 40% of evening charging to midnight. But when that decision gets translated into specific instructions for specific vehicles, it has to respect all those local constraints. A car that needs to leave at 5 AM can't be told to wait until 3 AM to start charging. A vehicle on a feeder line that's already heavily loaded can't be told to draw more power during a peak. The fact that this translation worked smoothly suggests that centralized optimization isn't as brittle as skeptics might fear.
The bottom-up approach told an equally interesting story. When distribution networks were given price signals from the centralized model and allowed to optimize locally, they converged on a charging profile remarkably similar to what the central planner had prescribed. In other words, the price mechanism worked. The invisible hand of electricity prices guided millions of individual charging decisions toward the globally optimal outcome, without requiring any central authority to micromanage each vehicle.
This is a significant finding for anyone worried about the scalability of grid management. Coordinating millions of vehicles through direct control—sending instructions to each car telling it exactly when to charge—is technically possible but administratively and politically complicated. It raises questions about data privacy, about who controls the infrastructure, about what happens when the coordination system fails. Price signals offer an alternative: just make it cheaper to charge when the grid needs you to, more expensive when it doesn't, and trust that people will respond rationally.
The study found that they do respond rationally. The bottom-up optimization identified that the centrally optimized profile was also locally optimal—there's no fundamental conflict between what helps the transmission system and what helps your local distribution transformer. This alignment suggests that a market-based approach could work, at least in the scenarios modeled.
However—and this is a crucial however—the bottom-up approach didn't perfectly replicate the top-down results. Some differences remained. The local optimizers, working with incomplete information about the broader system, sometimes made choices that were good locally but not quite optimal globally. The gap was small, but it existed. This suggests that pure price signals might need some backup: a system that handles the edge cases, the unusual conditions, the scenarios where the locally optimal choice isn't quite the globally optimal one.
Why This Changes Things: The Architecture of an EV-Friendly Grid
To understand why this matters, you need to understand the stakes. The International Energy Agency projects that there could be 350 million electric vehicles on the world's roads by 2030—up from about 30 million today. Each one represents a potential flexible load, a distributed battery that could help balance an increasingly renewable grid. Or each one represents a potential headache, a new demand that stresses aging infrastructure and forces expensive upgrades.
The difference between these outcomes depends almost entirely on how we orchestrate charging. Done wrong, EVs become a problem—forcing utilities to spend billions on grid upgrades, potentially requiring new fossil fuel plants to meet evening demand spikes when everyone gets home and plugs in simultaneously. Done right, EVs become part of the solution—absorbing excess solar and wind energy, providing grid services, smoothing out the renewable intermittency that currently makes high levels of clean power difficult to integrate.
The conventional wisdom in the field has been pessimistic about central planning's ability to work at both scales. The thinking goes: transmission-level optimization is necessary but not sufficient. You need distributed intelligence, local control, price signals, vehicle-to-grid services, and a thousand other mechanisms working in concert. The paper's findings suggest this pessimism may be overstated. Centralized optimization can work, and it can work without requiring heroic coordination or perfect information about every individual vehicle's state.
This has implications for how we build the control systems of the future. The researchers' top-down approach demonstrates that you can optimize at the scale of a continent and then translate those decisions down to individual households without the system falling apart. The disaggregation step—taking a global charging target and turning it into specific vehicle instructions—worked smoothly because the researchers explicitly designed it to respect local constraints. This suggests that future EV charging management systems might not need to be perfectly hierarchical, with all the intelligence at the top. They could be more like federal systems: national targets, local implementation, with clear rules about how lower-level decisions respect higher-level objectives.
The bottom-up findings are equally important because they point toward a market mechanism that might avoid the need for such hierarchical control in the first place. If electricity prices accurately reflect the grid's needs—if cheap power really does correspond to times when the system has excess renewable energy—then autonomous agents (smart chargers, vehicle management systems, aggregation algorithms) can make individually rational decisions that are collectively optimal. This is the promise of transactive energy: a grid where devices negotiate with each other through price signals, achieving coordination without coordination.
The study also offers reassurance about something practitioners worry about: whether distribution grids can handle high penetrations of EVs without massive infrastructure investment. The top-down approach explicitly accounts for grid constraints during the disaggregation process, ensuring that no distribution feeder gets overloaded. The bottom-up approach does the same through local optimization. In both cases, the controlled charging scenarios produced fewer grid violations than uncontrolled charging. This suggests that smart charging management might be a substitute for some infrastructure upgrades—not a complete substitute, but enough to buy time and reduce the capital requirements of the EV transition.
What's Next: From Modeling to Reality
The study has limitations that its authors acknowledge. Like all modeling work, it operates in a synthetic world that approximates but never perfectly replicates reality. The test system uses simplified distribution networks and aggregated loads. Real distribution grids are messier: they have three phases that don't always balance perfectly, voltage regulation equipment with its own dynamics, and interaction effects between solar panels, batteries, and EV chargers that are hard to capture in a model. The driving patterns come from German surveys—behavior may differ in other countries with different car cultures and commuting patterns.
The scenario modeled also assumes high penetration of distributed energy resources—solar panels on rooftops, local generation that complicates the flow of power through the distribution network. The results may differ in systems with less DER penetration, where the grid's constraints are different and the optimization landscape looks different.
What remains to be tested? First, the temporal dynamics the model doesn't fully capture: what happens when thousands of vehicles respond to a price signal simultaneously, creating a new peak where the model assumed they'd be spread out? Second, the human factor: real drivers don't behave like model agents. They have irrational preferences, inertia, distrust of automated systems, and habits that don't match the theoretically optimal patterns. Third, the market design question: electricity prices in most jurisdictions don't currently reflect the true cost of consumption in real time and location. Building a price mechanism that actually guides behavior requires regulatory changes and market reforms that are far from guaranteed.
The path forward probably involves learning from this research while extending it. Field trials with real EVs and real distribution networks would test whether the model's findings hold in the messy real world. Market designers need to grapple with the practical question of how to create price signals that are both accurate and understandable—reflecting grid conditions without confusing consumers. And utility regulators need to develop frameworks that allow smart charging to happen while protecting consumers from predatory pricing and ensuring that the benefits of flexibility are shared equitably.
The researchers suggest that centrally optimized EV charging remains effective when implemented at the distribution level, even in systems with high DER penetration. That's a significant result. It means that ambitious national targets for renewable integration—charging EVs when the wind blows, for instance—can be achieved without requiring micromanaged control of every vehicle on the road. The optimization doesn't break when it meets the real world.
This matters because the alternative would have been daunting. If centralized planning only worked in theory, we'd need to build a new infrastructure of distributed intelligence, local markets, and transactive energy systems before we could safely integrate millions of EVs. The finding that existing optimization approaches can work across scales doesn't eliminate the need for that infrastructure, but it suggests we have more runway than we thought. We can start deploying smart charging systems based on what we know while continuing to research the edge cases and exceptions.
The grid that emerges from this transition won't be either fully centralized or fully distributed. It will be a hybrid, a federated system where national planning sets the framework and local implementation fills in the details. The study from ETH Zürich and Delft suggests that this hybrid can work—that the invisible hand and the visible hand can cooperate rather than conflict. Whether that cooperation actually emerges depends on the policy choices, market designs, and technical standards that are being decided right now, in the offices of utilities and regulatory commissions and software companies building the systems that will manage our electrified future.
What this research confirms is that the engineering is not the obstacle. The obstacle is the human stuff: the coordination, the regulation, the markets, the trust. The physics allows for a clean energy future with electric vehicles playing a central role. The question is whether we can build the institutions to make it happen.
This digest is based on a preprint published on arXiv (Van Liedekerke et al., 2026). The full paper is available at https://doi.org/10.48550/arXiv.2607.24423.
Key Takeaways
- Central planning survives contact with reality: EV charging schedules optimized at the national transmission level translate effectively to the local distribution grid without causing widespread overloads or voltage violations.
- Prices can do the work: Bottom-up optimization using electricity price signals converged on nearly the same outcome as direct central coordination—suggesting market mechanisms could substitute for some hierarchical control.
- Flexibility enables renewables: Smart charging that shifts EV demand to times of abundant renewable generation is crucial for integrating high shares of wind and solar power.
- Hybrid approaches show promise: Neither pure centralization nor pure decentralization appears optimal; the solution likely involves national targets with local implementation.
- Grid upgrades can be deferred: Smart charging management significantly reduces distribution grid stress, potentially buying time before major infrastructure investment is needed.
References
Van Liedekerke, A., Zapparoli, L., Parajeles Herrera, M., Gjorgiev, B., & Hug, G. (2026). Does Central Planning Fail Locally? Evaluating EV Charging Flexibility Optimization Across Grid Levels. arXiv preprint arXiv:2607.24423. https://doi.org/10.48550/arXiv.2607.24423
Figures
-
Overview of the top-down and bottom-up approaches used in the study
Test system configuration with transmission and distribution network representation
Comparison of charging profiles under uncontrolled, top-down, and bottom-up scenarios
Grid violation reduction achieved by controlled charging approaches
Sensitivity analysis of DER penetration on optimization performance
Price signal evolution across the simulation period
Distribution of individual vehicle charging decisions in bottom-up optimization
Temporal alignment between renewable generation and optimized EV charging
Charts
Grid Violation Reduction by Control Approach
| Label | Value |
|---|---|
| Uncontrolled Charging | 100 |
| Top-Down Controlled | 25 |
| Bottom-Up Controlled | 30 |
Optimization Performance Metrics
| Label | Value |
|---|---|
| Transmission Efficiency | 94 |
| Distribution Efficiency | 91 |
| Renewable Integration | 87 |
Optimal Charging Load Profile
| Label | Value |
|---|---|
| Midnight | 15 |
| 2 AM | 35 |
| 4 AM | 55 |
| 6 AM | 25 |
| 8 PM | 80 |
| 10 PM | 60 |
Note: Charts and figures reference the full paper's data visualizations. Actual data values should be extracted directly from the original publication for precise numerical reporting.