Research report: Performance evaluation of sodium-ion batteries for microgrid and off-grid energy storage applications

Sodium-ion batteries have long been touted as a cheaper, more resource-friendly alternative to lithium-ion chemistry for stationary energy storage, but most laboratory evidence has come from standardized cycling that bears little resemblance to how a battery actually behaves inside a working power system. A new study now closes that gap by taking a single, fully characterized sodium-ion cell design and pushing it through two of the most demanding stationary applications at once: a grid-connected microgrid and a standalone off-grid power supply. The results, published in Frontiers in Chemistry, show that the same chemistry can deliver strong efficiency and reliability in both settings, but that the duty cycle imposed by each application leaves a very different fingerprint on the battery's long-term health.

The research, conducted by Xingyu Hu of Imperial College London, began with twenty-four 5.0-ampere-hour laminate pouch cells built around a layered sodium transition-metal oxide cathode and a hard-carbon anode. The cells used a 1.0 molar sodium electrolyte in a propylene carbonate and ethyl methyl carbonate blend with 2 percent fluoroethylene carbonate, a 2.00-to-3.65-volt operating window, and a negative-to-positive capacity ratio of 1.10 chosen to preserve cathode utilization while avoiding local sodium saturation at the anode. Twelve cells were reserved for electrochemical characterization and model identification, while the remaining twelve were split into three four-cell aging groups representing constant-current reference cycling, microgrid duty, and off-grid duty.

Baseline testing revealed how strongly operating conditions shape performance. Discharge capacity fell from roughly 4.98 ampere-hours at 0.2 C to 4.28 ampere-hours at 2 C at 25 degrees Celsius, a retention of about 86 percent, and most of that loss proved reversible when the rate returned to 0.5 C. Temperature mattered even more: at 10 degrees Celsius the 2 C capacity dropped to 3.72 ampere-hours, while at 40 degrees Celsius it rose to 4.42 ampere-hours. Crucially, Coulombic efficiency stayed above approximately 99 percent across all tested rates, meaning the chemistry itself remained highly reversible; the losses showed up instead in voltage polarization, which dragged energy efficiency down from about 94.8 percent at 0.2 C to 87.2 percent at 2 C. At 2 C, model-based attribution of the 12.8-point efficiency deficit assigned 5.9 points to instantaneous ohmic drop, 4.7 points to polarization dynamics, and 2.2 points to residual voltage-path and hysteretic losses.

To bridge the laboratory and the power system, the study identified a second-order Thevenin equivalent-circuit model directly from pulse and impedance measurements taken across 10 to 90 percent state of charge and 10 to 40 degrees Celsius. The open-circuit voltage, rising monotonically from about 2.34 volts at 10 percent state of charge to 3.58 volts at 90 percent, was captured by a cubic polynomial with an overall error of just 2.9 millivolts. Resistance parameters showed a shallow U-shaped dependence on state of charge, lowest near mid-range and rising toward both extremes, and fell consistently with temperature: at 50 percent state of charge, the instantaneous resistance dropped from about 14.5 milliohms at 10 degrees Celsius to 9.1 milliohms at 40 degrees Celsius.

When the recorded 24-hour microgrid and off-grid current traces were replayed through the model, voltage errors remained in the millivolt range, with root-mean-square errors of 7.6 and 9.2 millivolts respectively, and no simulation ever strayed outside the experimentally validated domain. The calibrated cell model was then scaled, without any lithium-ion surrogate parameters, to two full system designs. The grid-connected microgrid pack, arranged as 240 cells in series and 140 parallel strings, delivered 504 kilowatt-hours at 720 volts nominal and served a configuration with 300 kilowatts of solar, 150 kilowatts of wind, and a 350-kilowatt peak load.

The standalone pack, at 192 series and 42 parallel cells, provided 120.96 kilowatt-hours for a 90-kilowatt solar and 30-kilowatt wind system with an 85-kilowatt peak load and a 60-kilowatt diesel-equivalent backup generator. Both systems were simulated over a full 365-day year at 15-minute resolution using synthetic renewable and demand profiles built from fixed diurnal and seasonal components plus autoregressive perturbations, with the random seed fixed so that every comparison used identical input data. The annual results were striking.

The microgrid battery achieved a round-trip efficiency of 91.6 percent, tracked its net-load smoothing target with a root-mean-square error of 18.7 kilowatts, and curtailed only 4.3 percent of available renewable energy. The off-grid system, forced to juggle renewable-first supply, battery dispatch, and generator hysteresis thresholds, achieved 89.8 percent round-trip efficiency, a loss-of-power-supply probability of just 1.6 percent per year, and 6.4 percent curtailment. Sensitivity checks across twenty different profile realizations kept microgrid efficiency between 90.9 and 92.2 percent and off-grid loss-of-power-supply probability between 1.1 and 2.3 percent, while a finer 5-minute dispatch step barely moved the headline numbers.

Increasing battery energy by 20 percent cut the off-grid supply-loss probability to 0.9 percent, and adding 20 percent more renewable capacity pushed it down to 0.7 percent, offering a clear roadmap for designers of remote systems. The most consequential findings emerged from the aging experiments, where the three duty-cycle groups were cycled at 25 degrees Celsius to 500 equivalent full cycles or an 80 percent capacity endpoint, with reference performance tests every 100 cycles. Capacity state of health fell to 85.2 percent under reference cycling, 81.9 percent under microgrid duty, and 79.8 percent under off-grid duty, with the off-grid group just crossing the predefined end-of-life threshold at the final test point.

Resistance growth told an even starker story, reaching approximately 39, 58, and 74 percent in the three groups respectively. Capacity-fade rates climbed from 2.96 percent per 100 equivalent full cycles for reference cycling to 3.62 percent for microgrid duty and 4.04 percent for off-grid duty, and impedance spectroscopy showed the largest shifts toward higher resistance in the off-grid group, with fitted interfacial and charge-transfer resistance rising to 31.9 milliohms compared with 18.0 milliohms in the reference group. The authors attribute the faster off-grid degradation to a combination of deeper daily discharge excursions, longer residence at low state of charge, higher current severity, and greater energy throughput rather than duty-cycle shape alone.

The trace descriptors bear this out: off-grid cycling averaged a 61.3 percent daily depth of discharge against 41.6 percent for the microgrid, spent nearly twice as much time below 20 percent state of charge, and ran at a higher root-mean-square C-rate. Repeated wide state-of-charge swings and prolonged low-state-of-charge residence can accelerate hard-carbon interphase growth and charge-transfer polarization, while sustained layered-oxide cycling can promote structural rearrangement at the cathode. Because the measurements were full-cell impedance spectra without electrode-specific post-mortem analysis, the relative contributions of anode and cathode cannot be uniquely separated, a limitation the study acknowledges explicitly.

Several caveats frame the conclusions. The laboratory population was small, with only four aging replicates per group, and the cycling horizon stopped at 500 equivalent full cycles. Calendar aging was not included, even though the two application groups accumulated cycles at different daily rates and occupied different state-of-charge ranges, factors that could alter the long-term degradation ranking.

The annual profiles were synthetic rather than measured field data, and the pack simulations assumed scaled cell behavior without resolving cell-to-cell heterogeneity, thermal gradients, or balancing losses. A literature-parameterized lithium iron phosphate benchmark run on the same profiles produced somewhat higher round-trip efficiencies of 93.0 and 91.5 percent and capacity retention of 94.1 and 92.6 percent for the microgrid and off-grid duties, though this comparison is contextual rather than a matched head-to-head experiment. Even with those limits, the study delivers a rare and practically valuable message: sodium-ion batteries are ready to be judged as system components, not just as laboratory curiosities.

Under realistic, application-specific duty cycles, the tested layered-oxide and hard-carbon chemistry kept Coulombic efficiency near 99 percent, delivered round-trip efficiencies near 90 percent at pack level, and maintained supply reliability in a standalone configuration with under 2 percent annual loss-of-power probability. For engineers weighing sodium-ion storage against incumbent lithium-ion technology, the findings suggest the chemistry is technically credible for stationary duty today, while also warning that sizing, dispatch strategy, and state-of-charge management will decisively shape how long these batteries last in the field.

Subject of Research: Performance evaluation of sodium-ion battery pouch cells for grid-connected microgrid and standalone off-grid energy storage

Article Title: Performance evaluation of sodium-ion batteries for microgrid and off-grid energy storage applications

Article References: Hu, X. (2026). Performance evaluation of sodium-ion batteries for microgrid and off-grid energy storage applications. Frontiers in Chemistry, 14, 2007977. https://doi.org/10.3389/fchem.2026.2007977

Image Credits: AI Generated

DOI: 10.3389/fchem.2026.2007977

Keywords: sodium-ion batteries, microgrid, off-grid energy storage, stationary storage, battery degradation, round-trip efficiency, hard carbon anode, layered oxide cathode, equivalent-circuit model, renewable curtailment, loss-of-power-supply probability, electrochemical-thermal modeling