Mobile Power Sources: Mechanisms & Applications
- Mobile power sources (MPSs) are energy resources whose effectiveness relies on mobility, location flexibility, and adaptive control, spanning battery management, wireless charging, and mobile storage units.
- Research shows that optimizing MPS performance involves user-aware power management, adaptive wireless delivery, and dynamic routing to extend battery life and enhance grid resilience.
- Coordinated frameworks integrating energy, mobility, and communication reshape power delivery, improving efficiency, safety, and system restoration in diverse environments.
Searching arXiv for recent and foundational papers on mobile power sources, wireless power delivery, and mobile energy storage. Mobile power sources (MPSs) denote a broad class of power-supply mechanisms in which usefulness depends on mobility, mobility-aware operation, or mobile delivery. In the literature considered here, the term spans three partially overlapping domains: battery-backed mobile devices whose usable runtime depends on software and communication efficiency, wireless power-delivery architectures intended to provide charging over the air to moving terminals, and physically relocatable generation or storage assets that are routed through transportation networks to support distribution systems and microgrids. Across these domains, the common research question is not only how much energy is stored or delivered, but where, when, and under what control architecture that energy becomes available (Lim et al., 2021, Liu et al., 2018, Lei et al., 2018).
1. Conceptual scope and taxonomy
MPS research does not refer to a single hardware category. In mobile-device studies, the “source” is usually the onboard battery, and the research focus is runtime extension through power management rather than chemistry improvement or battery enlargement. In wireless-power studies, the source is an infrastructure-side transmitter or power beacon delivering energy over the air. In resilient power-system studies, the source is a movable asset such as a mobile emergency generator, mobile energy storage system, power ship, or on-road plug-in electric vehicle acting as mobile energy storage. This suggests that MPSs are best understood functionally: they are energy resources whose value depends on mobility support, location flexibility, or mobility-aware control (Lim et al., 2021, Huang et al., 2014, Almousa, 2019).
| MPS modality | Representative mechanisms | Representative papers |
|---|---|---|
| Device-internal runtime extension | user-scheduled deep sleep, modem sleep prediction, workload-aware adaptation | (Lim et al., 2021, Brand et al., 2019) |
| Wireless mobile power delivery | RBC, ARBC, MPT/WPC, SLIPT | (Liu et al., 2018, Xiong et al., 2021) |
| Grid-side movable power assets | MEGs, MESSs, SMESSs, power ships, PEV-based MES | (Lei et al., 2018, Wang et al., 2020, Chen et al., 2019, Almousa, 2019) |
A recurring distinction in this literature separates stored-energy mobility from infrastructure-delivered mobility. Stored-energy mobility includes batteries, mobile storage trailers, EVs, and power ships, where energy is physically carried to the point of use. Infrastructure-delivered mobility includes microwave power transfer, resonant beam charging, and optical SLIPT, where energy remains centrally generated but is delivered wirelessly to moving devices. A third line of work treats mobility as a control variable rather than a hardware attribute, as in user-aware scheduling of idle intervals or proactive modem sleep.
2. Device-centric MPS efficiency in mobile electronics
A substantial part of MPS research argues that effective mobile power is not determined solely by battery capacity. One representative result is that a factory-reset dual-core smartphone with a 1650 mAh battery lost 19% of charge, or 315 mAh, over eight hours even when no user applications were running. Of that idle drain, 80% came from device operations and service daemons such as cell standby, device idle, and Android OS activity, while 20% came from application daemons such as Wi-Fi, screen-related activity, and Gmail synchronization. On that basis, a user-aware power manager was proposed with four components—user-space client, sleep time manager, sleep level controller, and battery timer—and three timer-driven low-power modes: suspend-to-RAM, suspend-to-disk, and complete-off. Under the tested idle conditions, this system extended battery lifespan by up to 18% over the existing system, with suspend-to-disk showing consumption similar to complete-off because in both cases only the battery timer meaningfully remained active (Lim et al., 2021).
This line of work is notable because it rejects the assumption that larger batteries alone solve mobile runtime constraints. The proposed method is time-based rather than battery-threshold-triggered, and “user-aware” means explicit user-supplied sleep and wake-up times rather than latent-context inference. A plausible implication is that, for predictable non-use periods, deep platform-wide suppression can be more effective than selectively killing applications. The corresponding limitation is equally clear: savings depend on user cooperation and on hardware timer support capable of surviving deep sleep or power-off states (Lim et al., 2021).
Workload-level studies reinforce the same point from a different angle. On an LG Nexus 5 with a 2300 mAh battery, local playback power rose from 783 mW at 360p to 1696 mW at 1080p, while streaming over WiFi increased total device power by 26% to 38% relative to internal memory and streaming over 3G increased it by 49% to 64%. For 1080p, total power rose from 1696 mW for internal playback to 2561 mW for 3G streaming, a 51% increase. The inferred wireless-interface burden itself increased with video quality, by up to 58% for WiFi and 72% for mobile networks. This establishes that mobile battery drain during multimedia use is jointly determined by decode/render load and transport-interface choice, not by CPU activity alone (Zsak et al., 2014).
A further step is proactive communication-centric control. For LTE devices, one paper modeled the receive chain as repeatedly waking each 1 ms transmission time interval to decode control information that is frequently irrelevant. It proposed two lightweight predictors—one supervised and one reinforcement-learning-based—to decide whether the modem could remain asleep. After accounting for predictor overhead, the reported achievable energy savings reached up to 17%. The RL variant was especially notable for low computational overhead, while the supervised model achieved lower false-negative rates on stable traces. This suggests that, in mobile electronics, MPS effectiveness increasingly depends on predicting when communication hardware can safely avoid “listen just in case” behavior (Brand et al., 2019).
3. Wireless and infrastructure-based mobile power delivery
A second major branch of MPS research seeks to replace episodic charging with ambient or infrastructure-based wireless power delivery. One influential systems vision is the Mobile Power Network (MPN), defined as a wireless power transfer network for mobile ranges from several meters to tens of meters, positioned as the power-side analogue of mobile Internet. In this view, power becomes a managed service delivered by access points connected to the power grid and Internet, with user equipment containing resonant-beam receivers. The enabling physical-layer technology is resonant beam charging (RBC), also called distributed laser charging in some references, because it combines meter-level range, watt-level power, self-alignment, inherent safety through obstruction-triggered shutdown, and concurrent charging of multiple devices. Reported RBC-related markers in that literature include 2 W electrical power delivery over 2.6 m, over 3 W over 2 m within over field of view, and theoretical 1 W transfer over 5 m under skin-safe regulations (Liu et al., 2018).
Microwave power transfer (MPT) offers a different infrastructure-based MPS paradigm. In wirelessly powered communications, base stations and especially power beacons become distributed mobile power nodes, while devices use RF energy harvesters. The paper on MPT emphasizes that practical transfer range is about 3–15 m, that transmitter DC-to-RF and receiver RF-to-DC efficiencies can each be around 80%, and that beam efficiency is the dominant bottleneck. It also highlights a critical safety constraint: average microwave power density should not exceed over a half-hour window, implying that useful wirelessly powered communications require active beam control and human-detection mechanisms rather than naive broadcast charging (Huang et al., 2014).
Optical wireless power delivery pursues higher power density. One mobile ARBC framework for IoT treats resonant beam charging as a “Wi-Fi-like” charging service and adapts source power to both battery-preferred output power and time-varying charging distance. In simulation with a 1,000 mAh Li-ion battery, the adaptive resonant beam charging scheme reduced average energy consumption to about 150 Wh, versus about 530 Wh for constant-power charging, about 320 Wh for profile-adaptive charging, and about 248 Wh for distance-adaptive charging. The reported savings were about 74.0% relative to constant-power charging, 53.8% relative to profile-adaptive charging, and 43.7% relative to distance-adaptive charging. This suggests that mobile wireless charging becomes materially more efficient when source power adapts jointly to battery profile and geometry (Zhang et al., 2020).
A more communication-centric optical architecture appears in mobile SLIPT based on a spatially separated laser resonator with intra-cavity second harmonic generation. In that design, the 1064 nm resonant beam supplies charging power while a 532 nm second-harmonic beam carries data over the same self-aligned path, avoiding receiver positioning and beam steering. At m, W, , and SHG crystal thickness mm, the reported maximum charging power was $1.05$ W and the achievable rate was $11.03$ bit/s/Hz. The paper also notes PT-branch efficiency around 2% for W and m, while identifying line of sight, safety engineering, and specialized optics as practical constraints (Xiong et al., 2021).
Across these wireless-power studies, a common misconception is that over-the-air power automatically implies battery-free mobile computing. The reported evidence instead supports a narrower interpretation: infrastructure-delivered wireless power is presently better understood as a battery life extender, a continuous top-up mechanism, or a managed ambient charging layer rather than a wholesale replacement for wired fast charging or all onboard storage (Xiong et al., 2021, Huang et al., 2014).
4. Mobile storage and generation in resilient power systems
In distribution-system and microgrid literature, MPSs are physically relocatable assets deployed after faults, disasters, or prolonged outages. A foundational formulation co-optimized service restoration with repair crews and MPS dispatch by routing MPSs in the transportation network, scheduling them in time, and coordinating their injections with dynamic microgrid formation. That model explicitly distinguished mobile emergency generators (MEGs) and mobile energy storage systems (MESSs), linked nodal injections to location variables, and used radiality constraints based on spanning forests so that restored islands could be powered by MPSs and reshaped by repair actions and switch operations. The central operational insight was that MPSs are most valuable not as isolated backup devices but as sources for dynamic microgrids that form, expand, reconnect, and dissolve during restoration (Lei et al., 2018).
Later work refined the mobility model itself. A compact linear routing model for mobile energy resources represented each asset as either parked at a node or traveling toward a node, using binary variables 0 and 1, a residual-travel variable 2, and direction-consistency constraints. The model size scaled as 3 binary variables and 4 constraints, avoiding the quadratic dependence on node count typical of time-space-network formulations. In the hardest reported case—37 nodes with a 10-minute time span—the proposed model solved in approximately one tenth of the time of the general TSN. This matters for MPS practice because fine time discretization materially improved restoration value: in the 37-node test, the objective increased from 5 kWh at 30-minute resolution to 6 kWh at 10-minute resolution (Wang et al., 2020).
A further extension introduced the separable mobile energy storage system (SMESS), in which a carrier and multiple detachable storage modules are scheduled separately. A carrier can transport several modules, drop them at different nodes, and later pick them up again, allowing one transport asset to distribute storage across multiple outage islands. The same paper jointly modeled SMESS carriers, MEGs, and fuel tankers, explicitly representing node fuel inventories, MEG fuel use, and tanker routing. This is significant because prior generator-based MPS studies often assumed fuel availability away. Here, mobile power-source realism was extended to include the logistics that sustain MEG operation (Wang et al., 2020).
Battery-based MPS routing has also been studied in islanded microgrids using co-optimized load restoration and mobile storage routing. One MISOCP formulation tracked not only state of charge but also the evolution of lower and upper battery-capacity bounds as batteries moved between buses. In that model, mobile storage changed both local energy and local storage capability, so battery mobility altered the feasible set of future dispatch. This suggests that mobile batteries should not be modeled as stationary ESS with a travel delay added afterward; their physical relocation changes the network’s storage topology itself (Bose et al., 2022).
At a larger scale, power ships represent one of the largest mobile energy resources. A mixed-integer maritime-power co-optimization model treated power ships as routeable generators that move among ports and inject at connected buses through ship-to-grid arrangements. On the IEEE 118-bus high-loading case, the fully integrated model reduced total cost from \$10~\text{W/m}^2$72,037,004.10 and eliminated 4.276 MW of load shedding. Even in a lower-loading case with no baseline shedding, the same model reduced operating cost from \$10~\text{W/m}^2$82,004,269.73. The implication is that large MPSs can provide both resilience and routine economic value, provided their routing, berth constraints, and generation schedules are co-optimized with network operation (Almousa, 2019).
5. Coordination architectures: incentives, interdependence, and learning
Not all MPS coordination is utility-owned or centrally scheduled in a conventional dispatch sense. One alternative is to use privately owned plug-in electric vehicles as on-road mobile energy storage for charging-station overload compensation. In that framework, a power system operator posts a service price, and each participating vehicle chooses an energy-transfer amount by maximizing a utility function that balances service revenue, a motivation reward term, service-time cost, and battery-degradation cost. The resulting one-leader, multiple-follower Stackelberg game admits a unique equilibrium. This shows that some MPS categories are best modeled not only as engineering assets but also as economic agents whose participation thresholds and saturation prices determine effective dispatchable capacity (Chen et al., 2019).
Another control-oriented strand treats plugged-in EVs as bidirectional mobile storage inside a virtual power plant. In a residential-scale RL study with 4 households, 4 charging stations, 16 kW PV, and 12 kW wind, the best RecurrentPPO controller reduced RE2V unused energy from 17,475.87 kWh in uncontrolled charging to 5828.97 kWh, reduced grid energy used from 39,397.31 kWh to 2035.55 kWh, and improved self-consumption and autarky in the selected validating configuration from 58.1% and 43.9% to 90.0% and 98.3%, respectively. The paper interprets EVs as mobile storage mainly through stochastic arrival and departure rather than explicit spatial routing, but it nonetheless demonstrates that mobile storage assets can materially improve local autonomy when coordinated with renewables (Maldonato et al., 2024).
A key recent development is the recognition that mobile-power restoration is fundamentally an interdependent power–transportation–information problem. In a PTIN-based restoration framework, EVs and MESSs restore electrical supply while UAVs restore communication coverage needed for feeder automation and EV dispatch. The paper argues that communication coverage determines EV deployment capability and PDN controllability, while traffic operation efficiency determines whether mobile resources reach V2G stations quickly enough. Its two-stage scheme therefore prioritizes CN and UTN loads before general PDN loads, because restoring communication and traffic facilities enlarges the feasible set of later MPS dispatch and switch control. This suggests that, in disaster recovery, early restoration of enabling infrastructure can improve the productivity of later mobile power deployment more than immediate restoration of ordinary loads (Zhong et al., 2024).
6. Decentralization, limits, and research directions
A recurrent limitation across MPS research is the assumption of intact centralized coordination. Recent work on microgrid resilience enhancement via MPSs and repair crews relaxes that assumption by studying decentralized operation under degraded communication. In that setting, MEGs and MESSs are modeled as agents in a hierarchical Dec-POMDP, and the proposed H2MAPPO method uses a high-level action to switch between transport and power decision contexts and a low-level hybrid policy to generate discrete routing and continuous scheduling actions. On the IEEE 33-bus system, H2MAPPO achieved an average resilience index of 21.34, compared with 19.56 for IPPO, 20.15 for MAPPO, 19.93 for MPC, and 22.12 for centralized MILP, while keeping average computation time per day at 0.54 s versus 76.21 s for MILP and 1028.93 s for MPC. On the IEEE 69-bus system, MESSs contributed 2068 kWh, MEGs 10,299 kWh, and repair crews 26,495 kWh. This suggests that decentralized MPS control can approach centralized restoration quality when communication is impaired, especially if routing and power decisions are separated hierarchically (Wang et al., 24 Jul 2025).
The literature also reveals several persistent constraints. In mobile-device power management, user-scheduled or predictive methods are limited by hardware timers, estimation errors, and unquantified QoS penalties such as missed calls, resume latency, or synchronization delay (Lim et al., 2021, Brand et al., 2019). In wireless MPS architectures, the dominant barriers remain end-to-end efficiency, safety certification, line-of-sight dependence, deployment density, and hardware complexity (Huang et al., 2014, Xiong et al., 2021). In grid-resilience applications, major open issues include fuel logistics, uncertain travel times, damaged roads, limited communication, scalable multi-resource coordination, and initial siting or pre-allocation of mobile assets (Wang et al., 2020, Zhong et al., 2024, Wang et al., 24 Jul 2025).
A common misconception is that MPS research is only about adding portable energy to an already defined system. The surveyed work indicates something broader: MPSs reshape the feasible operating region by coupling energy with mobility, communications, and topology. In device contexts, this means usable runtime depends on idle-state exploitation and communication-aware sleep. In wireless-power contexts, it means the charger is an infrastructure service rather than a standalone accessory. In power-system contexts, it means that routing, network reconfiguration, communications, and transportation are inseparable from power dispatch. The most plausible research direction, therefore, is not a single universal MPS architecture, but increasingly integrated models in which energy delivery, movement, control, and infrastructure dependency are optimized jointly.
