EV Aggregation Is Reshaping How Grid-Scale Storage Gets Planned—And Why It Matters
In a quiet corner of China’s power sector, a subtle but powerful shift is underway. Engineers and researchers are beginning to treat electric vehicles not just as energy consumers—but as dynamic, mobile grid assets. This change in perspective is unlocking new strategies for energy storage deployment, particularly at the distribution level, where renewables’ volatility and load uncertainty have long complicated planning.
A groundbreaking study recently published in Energy Storage Science and Technology by Luo Shigang, Teng Jie, and Tan Zhuangxi offers one of the clearest blueprints yet for how this integration can—and should—be done. The paper introduces a novel two-stage stochastic optimization framework for siting and sizing grid-connected energy storage systems, with an especially notable feature: it explicitly embeds the dispatchable flexibility of aggregated electric vehicle (EV) charging stations into the planning process. And the results are not incremental—they’re transformative.
At first glance, the idea seems obvious: if millions of EVs are already connecting to the grid every day, why not treat them as a collective, controllable load (or, at times, as a distributed energy resource)? But the devil, as always, lies in the details—in modeling fidelity, computational tractability, and real-world operational constraints. The team’s method sidesteps many of the traditional roadblocks. Instead of drowning planners in high-dimensional optimization problems or requiring individual EV data, they use a second-order approximated Minkowski sum-based model to describe the aggregate feasible region of a charging station. This model captures the essential flexibility of a fleet of EVs—arrival times, departure times, state-of-charge requirements—but without the combinatorial explosion that has historically made such approaches impractical for large-scale planning.
What makes this work stand out in a crowded field is its operational realism. Too often, energy storage planning studies treat storage as a black box: charge when cheap, discharge when expensive. That approach, while economically intuitive, can backfire. In some cases, large-scale arbitrage-driven storage behavior can actually worsen grid congestion—introducing new peaks where none existed, or shifting stress from one feeder to another. Luo and his colleagues avoid this pitfall by making power flow temporal-spatial balance the core objective—not just cost minimization or pure profit maximization.
Think of it like urban traffic planning: you don’t just want fewer cars on the road; you want smoother, more evenly distributed traffic over time and across lanes. Similarly, the team’s balance indicator quantifies how evenly current flows across all branches and time intervals. Minimizing this metric leads to flatter, more resilient voltage profiles, reduced thermal stress on lines and transformers, and lower risk of cascading failures. It’s a systems-level metric—precisely the kind of forward-thinking metric that grid operators need as distributed energy resources multiply.
The model also cleverly sidesteps another classic headache: how to represent energy losses during charge/discharge cycles without crippling the solver with binary variables or nonlinearities. Their solution? Replace internal battery inefficiencies with a virtual resistance branch in the power flow equations—a small but elegant trick borrowed from advanced distribution modeling literature. This linearizes the loss representation while preserving physical fidelity, and it plays beautifully with conic relaxations of the Distflow equations, which the team uses to ensure computational scalability and solution accuracy.
But perhaps the most important insight from this research isn’t in the math—it’s in the implications.
When the authors tested their framework on a modified IEEE 33-bus distribution system, they found something striking: as the controllable proportion of EVs increased—from 20% to 80%—the optimal energy storage capacity installed actually rose slightly, contrary to the naive expectation that more EV flexibility would replace battery storage. Why?
Because EVs and stationary storage aren’t substitutes—they’re complements. EVs offer short-term, highly granular load-shifting capability, ideal for intra-day solar smoothing or rapid response to local voltage dips. Stationary storage, especially when co-located with photovoltaic generation, provides longer-duration energy shifting—moving excess midday solar to evening peak hours, for instance. Together, they cover different slices of the flexibility spectrum. The model recognizes this synergy and allocates resources accordingly.
More crucially, the carbon impact is dramatic. With just 50% EV controllability, daily carbon emissions dropped by over 18% in the test case—not from adding more generation, but from smarter coordination. Higher EV participation pushed that reduction to nearly 30%. Meanwhile, photovoltaic curtailment plummeted: solar utilization jumped from around 86% in the baseline case to over 97% when EVs and optimized storage operated in concert.
This isn’t theoretical. The study’s scenario generation uses real-world charging records—start times, stop times, total energy delivered—from a provincial EV charging network in China. And the second-order aggregation model proved remarkably accurate: when validated against full individual-EV scheduling, its mean squared error in aggregated power tracking fell to near-zero once more than 50 EVs were in a station’s pool. That’s the threshold where statistical smoothing kicks in—and where planners can confidently treat a charging hub as a predictably flexible load, much like a small, fast-responding demand-side resource.
From a utility perspective, the economics check out too. In the case study, the installed storage—2.25 MWh at Node 16 and 3.32 MWh at Node 21—cost roughly ¥2.3 million annually (amortized), but delivered ¥3.46 million in annual benefits: ¥2.67 million in reduced energy purchases, ¥0.12 million in deferred grid upgrades, and ¥0.67 million in avoided carbon compliance costs. That’s a 150% return on investment—and it includes the cost of the storage itself.
This kind of result flips the traditional narrative. Energy storage is often seen as a cost center—a necessary buffer to accommodate renewables. Here, it’s a value multiplier: it enables better coordination with EVs, which in turn unlocks more renewable penetration, which then justifies even more storage. It’s a virtuous cycle.
Of course, real-world implementation isn’t trivial. The model assumes centralized coordination of charging stations—something that requires robust communication infrastructure, standardized APIs (like OCPP 2.0), and regulatory frameworks that allow aggregators to bid flexibility into grid services markets. It also presumes accurate forecasting of EV arrival/departure patterns, which remains challenging in unstructured urban environments.
But signs point to feasibility. In California, utilities like PG&E are already piloting “managed charging” programs where EVSEs respond to price or reliability signals. In the UK, National Grid’s “Dynamic Frequency Response” service now accepts bids from EV fleets. And in China—where this research originates—the State Grid Corporation has been aggressively deploying vehicle-to-grid (V2G) pilot zones in cities like Shenzhen and Nanjing. The pieces are falling into place.
What’s missing, until now, was a planning methodology sophisticated enough to anticipate this convergence—and optimize the grid for it, years in advance. Most storage siting studies still treat loads as passive, renewables as exogenous, and EVs as noise. This paper changes that. It forces planners to ask: Where do we put storage not just to buffer solar, but to enable EVs to become active grid participants?
The answer, it turns out, is strategic co-location. In the IEEE-33 test case, storage was placed at Nodes 16 and 21—not coincidentally, the same feeders hosting photovoltaic arrays and EV charging clusters. This creates local micro-hubs of flexibility: solar charges batteries by day; batteries and coordinated EV charging jointly support the evening peak; excess EV discharge capability (when available and consented) provides ancillary services.
It’s a vision of the distribution grid not as a one-way pipeline, but as a mesh of intelligent, interacting flexibility nodes—where assets talk, trade, and adapt in real time.
Critically, the authors avoid over-claiming. They don’t assume all EVs are V2G-capable (a still-nascent technology); instead, they focus on unidirectional smart charging, which is far more deployable today. They don’t require perfect prediction—uncertainty in solar and load is handled via K-Medoids scenario reduction, a robust clustering algorithm that preserves distributional features better than simple k-means. And they don’t ignore hardware limits: battery state-of-energy constraints, converter ratings, and thermal line limits are all explicitly enforced via second-order cone formulations, which guarantee near-global optimality.
The method’s use of Benders decomposition is also noteworthy. By separating investment decisions (first stage: where and how much to build) from operational decisions (second stage: how to dispatch storage and EVs across hundreds of scenarios), it achieves both scalability and realism. The master problem converges in under 30 iterations—a blink of an eye for such a complex system—and each subproblem (one per scenario) can be solved in parallel, making the approach viable even for larger feeders.
Looking ahead, this framework could easily be extended. What if some EVs do support V2G? The model can absorb bidirectional power curves with minimal modification. What if hydrogen-fueled heavy-duty trucks join the mix? Their longer refueling times and higher energy demands would simply alter the aggregate feasible region parameters. The core architecture is modular and extensible.
More profoundly, this work challenges a deep-seated assumption in utility planning: that passive, worst-case load modeling is “conservative”—and therefore safe. In reality, in a world of distributed flexibility, ignoring controllable resources is the riskiest stance of all. It leads to over-investment in wires and transformers, underutilized storage, stranded renewables, and missed decarbonization opportunities.
Luo, Teng, and Tan’s contribution is to provide a rigorous, implementable alternative—a planning methodology that doesn’t just tolerate uncertainty, but harnesses it. By treating EVs not as a threat to grid stability, but as a latent reservoir of adaptive capacity, they’ve opened a new design space for the distribution grid: one that’s not just cleaner or cheaper, but more intelligent, more resilient, and fundamentally more human-centered—designed around how people actually live, drive, and charge.
In the end, the most important number in this study isn’t the 150% ROI or the 30% emissions cut. It’s the implicit message: the grid of the future won’t be built despite electric vehicles. It will be built with them—in active partnership.
And for the first time, we have a blueprint for how to do that right.
Luo Shigang¹, Teng Jie², Tan Zhuangxi³
¹ Economic and Technological Research Institute of State Grid Gansu Electric Power Co., Ltd., Lanzhou 730030, China
² State Grid Gansu Electric Power Co., Ltd., Lanzhou 730030, China
³ Hunan University of Science and Technology, Xiangtan 411100, China
Energy Storage Science and Technology
DOI: 10.19799/j.cnki.2095-4239.2023.0310