A research team including Can Wan and Xitong Mu of Zhejiang University has proposed a new way to use moving trains as an energy storage resource for the power grid. According to a peer-reviewed paper published in Nature Communications on October 19, 2026, adjusting how trains accelerate and decelerate within timetable and speed-limit constraints allows electrical energy to be temporarily stored as kinetic energy. The team estimates that, as of 2025, the world's electrified railways have a latent capacity equivalent to roughly 190 GW of discharge power. Using existing electrified railways could reduce some investment in new storage facilities. However, before the approach can be put into practice, it must be tested on actual lines to confirm how much regulating capacity can be secured at the times it is needed, without disrupting train operations.
How moving trains can act as energy storage
Electric trains consume power when accelerating with their motors, and convert part of their kinetic energy back into electricity through regenerative braking when slowing down. The team calls its approach "ERES" (Electrified Rail-Based Energy Storage). It exploits these characteristics by adjusting running speeds and the timing of acceleration and braking, so that the railway as a whole functions as a kind of mechanical energy storage system. Rather than installing new batteries on trains, it makes use of the kinetic energy of the trains themselves and the room available to raise or lower it.
When surplus power is available, trains are run so that they consume more electricity than their normal operating plan calls for, increasing their kinetic energy. Conversely, when power is short, the system reduces consumption by cutting back on "traction" (driving the train with its motors), or returns electricity through regenerative braking. Whether regenerated power can be sent from the railway's supply equipment back upstream to the grid depends on how the substations are configured. In this study, "discharge" includes not only actually sending power back, but also consuming less power than the normal operating plan.
Such adjustments must stay within limits that do not interfere with train operations. In the team's model, trains must keep to scheduled arrival times, stay within speed limits and motor output limits, and return to a specified motion state after the adjustment. Sections unsuited to adjustment, such as acceleration just after departure and braking just before stopping, were excluded. The researchers envisage coordinating multiple trains with different schedules and running states to regulate the amount of power absorbed or supplied across a region.
Conventional work on regenerative power in railways has centered on how efficiently the electricity generated during braking can be recovered. This study instead evaluates the flexibility in how trains are run as storage capacity in its own right. The idea is to make use of the latent regulating capacity of existing electrified railways before building new batteries or dedicated rail facilities.
190 GW of discharge power worldwide, up to 579 GW by 2050
The team first analyzed Chinese railway operating data to determine how much power trains could absorb and supply. The data comprised 2,688 regional timetables containing 326,034 inter-station runs. An inter-station run is a train's journey from one scheduled stopping station to the next, without breaking at stations it passes through. A round-trip service counts as separate runs, so the figure of 326,034 does not represent a number of trains or vehicles.
The analysis covers only electrified main-line railways and excludes subways and light rail. Those urban systems have short distances between stations and accelerate and brake frequently, so they require a different operating model. For each inter-station run, the team combined conditions such as vehicle mass, motor output and track speed limits, and calculated the adjustable power and energy by region.
The team then used a linear relationship observed in the Chinese data between regional storage capacity and variables such as annual electricity consumption for train propulsion to estimate the global potential. Regional differences in vehicle weight and operating speed were corrected for, and power consumed by auxiliary equipment not directly related to running was excluded. Railway development scenarios assumed by the International Energy Agency (IEA) were then applied to obtain storage capacity for 2025 and 2050. Note that this is an estimate based on a statistical relationship; it does not demonstrate a causal link in which more railway power consumption necessarily means more storage capacity.
The global estimates, shown in Supplementary Tables 12 and 13 of the paper, are as follows.
| Scenario | Global discharge power | Storable energy |
|---|---|---|
| 2025 | approx. 190 GW | 6,309.23 MWh |
| 2050 (baseline rail growth) | approx. 315 GW | 10,448.42 MWh |
| 2050 (expanded rail use) | approx. 579 GW | 19,214.34 MWh |
The discharge power for 2025 was estimated at about 190 GW, with storable energy of about 6.3 GWh. For 2050, discharge power is projected at about 315 GW under baseline rail growth, and about 579 GW if rail use expands further. All of these are model-based estimates of potential, and even the 2025 figure is not a record of storage equipment actually installed and operating. The range in the 2050 figures reflects differences between railway growth scenarios, not measurement error in the estimates.
It is important to distinguish between GW, which expresses power output, and GWh, which expresses the amount of energy that can be stored. A figure of 190 GW alone sounds like an enormous storage system, but the storable energy is only about 6.3 GWh. This combination suggests that ERES is better suited to absorbing short-term fluctuations in supply and demand than to supplying power over long periods. How long and at what output the system can actually be sustained, however, will vary with the operating conditions and control settings of the trains at the time.
Uncertainty remains in the global estimates. The underlying operating data come from Chinese railways, and even with corrections for regional vehicle weight and operating speed, they cannot fully reflect differences in operating practices and equipment configurations among countries. Power consumption by auxiliary equipment also relies on estimated values, so the global potential should be distinguished from the capacity actually available on individual lines.
May partly replace existing storage, but additional investment is needed
The team also examined how rail storage capacity would affect the power system as a whole. Using the power system planning model SWITCH, they compared capital and operating costs for the power systems of China, Europe and India with and without ERES. The analysis period runs from 2025 to 2054, divided into three investment periods represented by the years 2030, 2040 and 2050. Operations in each year are reproduced at hourly resolution for four representative days chosen by season, which makes long-term investment planning computationally tractable.
The simulations set two railway development scenarios and evaluated three levels of progress in renewable energy and storage technologies. Carbon price trajectories were also varied to examine how ERES's value changes under different policy and technology conditions. In the cost-focused analysis, total cost was minimized while keeping CO2 emissions at or below the baseline case. In the emissions-focused analysis, total cost was held at or below the baseline, so the maximum benefits obtained in each analysis cannot necessarily be achieved at the same time.
The results indicate that using the regulating capacity of existing electrified railways could allow some investment in conventional storage facilities to be avoided. However, this is the result of re-optimizing the configuration of the whole power system, including generation and transmission, with ERES added; it is not an effect observed on an actual grid. The comparison included batteries assumed to supply power for about four hours, as well as longer-duration storage such as compressed air energy storage, pumped hydro and hydrogen-based storage. ERES is not meant to replace these facilities wholesale; the premise is to combine multiple technologies that handle supply-demand fluctuations on different time scales.
Using existing railways does not mean no additional investment is needed. The paper estimates the initial cost of introducing ERES at $5.8 per kW for AC railways and $180 per kW for DC systems, in 2025 US dollars. These figures assume systems that link train and grid control commands, and, for DC systems, substation upgrades to allow power to flow in the reverse direction. They are planning estimates calculated with reference to the costs of existing technologies, and do not represent the amounts actual construction work would require.
There are equipment and operational hurdles between having an electrified railway and being able to use its regulating capacity on the grid. In particular, sending regenerated power back upstream requires confirming that substations can handle reverse power flow and that protection devices operate correctly. Costs for additional control equipment and its maintenance must also be considered. Only once actual installation and operating costs are known can the economics be compared with building new batteries.
Simulation shows reduced frequency drop; real-line demonstrations are still to come
To confirm that train schedules could be maintained while operations were adjusted, the team ran running simulations for a high-speed railway linking Shanghai and Kunming and a parallel conventional line. The study covered eight inter-station runs on each, 16 in total, with speed limits of 310 km/h or 350 km/h on the high-speed line and 120 km/h or 160 km/h on the conventional line. Using the trains' equations of motion to adjust acceleration and deceleration, the team could calculate running patterns that kept scheduled arrival times while respecting motor output and speed constraints. This is a numerical simulation result, however, and was not confirmed by running actual trains.
The effect on the grid was examined separately from the 16 representative runs, using a model that aggregates the railways of all of East China. Specifically, the team assumed an incident in which 4.9 GW of generation drops out of a 50 Hz power system with a total load of 138 GW. The railway side was set to supply 1.61 GW of regulating power for five minutes, then recharge at 0.36 GW for 25 minutes. The aim was to see whether such operation could limit the frequency drop when a sudden power shortage occurs.
In the simulation, the lowest frequency after the incident improved from 49.592 Hz without ERES to 49.726 Hz with it. These values were obtained by taking the minimum of the time-series data in columns B and C of the "Fig. 8" sheet in the team's published Source Data, rounded to three decimal places. Both are calculation results under the same incident conditions and system parameters, but they do not necessarily compare frequencies at the same moment, and they are not measured values. Moreover, this effect comes from the model aggregating all of East China's railways, not from the 16 inter-station runs described above alone.
Challenges also remain that simulations alone cannot assess. For example, on an actual grid, power exchanged with railways could cause local voltage fluctuations and harmonics, so coordination with protection devices must be verified. The real-time control question of whether power adjustment can continue while prioritizing on-time operation when trains are delayed or disrupted has also not been evaluated. There are institutional issues as well: how to return to railway operators the cost savings that ERES brings to the power system as a whole, and how to meter and settle the regulating capacity they provide.
So that third parties can reproduce the analysis, the team has also published the data and programs used in the calculations. A Zenodo dataset distributes the 2,688 railway operation cases across 12 ZIP files. A GitHub repository contains the storage capacity calculation code and SWITCH configuration files, along with input data for 315 scenarios covering China, Europe and India. Reproducing the calculations, however, requires the specified software environment and a valid license for the optimization software Gurobi. Publishing code and data makes third-party verification possible, but it does not mean independent researchers have reproduced the same results. The materials available for this article show no results from third-party replication or demonstration operations on commercial lines.
The paper is an accepted manuscript that has passed peer review and been published ahead of final production, so further editing by the publisher may follow. The next steps the team cites are trial operations on representative lines and verification of a mechanism linking control commands on the railway and power sides. If it can be confirmed that on-time operation can be maintained, the technical constraints at the points of connection to the grid can be met, and railway operators' additional costs can be recovered, a path would open to using existing electrified railways for short-term power supply-demand balancing. The key question going forward is whether grid flexibility can be improved not only by building large amounts of new storage, but also by making smarter use of transport infrastructure that already exists.