Hydrogen Energy Storage Overview

• Hydrogen Energy Storage is a suite of technologies that leverages hydrogen’s high specific energy and diverse conversion pathways to store surplus renewable power and smooth grid fluctuations.
• It encompasses centralized, distributed, and geological systems, often integrating with battery storage to optimize both short-term response and long-duration energy balancing.
• Performance is assessed through process modeling, techno-economic analysis, and advanced control strategies, addressing challenges like low round-trip efficiency and material limitations.

Hydrogen energy storage (HES) refers to the suite of technologies and systems that use hydrogen as a reversible energy carrier to mitigate grid variability, store excess renewable energy, and provide multi-scale, cross-sectoral energy system flexibility. HES leverages hydrogen’s unique thermochemical properties—high specific energy, diverse conversion pathways, and long-duration storability—to fill temporal and spatial mismatches between variable renewable generation and demand.

1. Technological Architectures and System Components

HES architectures fall into three primary classes: centralized power-to-hydrogen-to-power (“power-to-gas-to-power,” P2G2P) systems; distributed behind-the-meter HES in buildings or microgrids; and large-scale subsurface storage infrastructures for seasonal or system-wide balancing. The canonical grid-connected HES chain incorporates an electrolyzer to generate hydrogen during periods of surplus electricity, a hydrogen buffer (via physical or materials-based storage), and a fuel cell or hydrogen combustion turbine for reconversion to electricity (Yu et al., 2023).

A representative topology for utility-scale renewable integration consists of:

• Alkaline electrolyzer (AE) for H₂ production at efficiency $\kappa_{\rm AE}$,
• Hydrogen buffer storage (HS), with state-of-charge and inventory management,
• Proton exchange membrane fuel cell (PEMFC) for electrical discharge at efficiency $\kappa_{\rm FC}$,
• Operating constraints including capacity expansion, energy balance, and power ramping limits.

Hybrid architectures increasingly pair HES with battery energy storage (BESS) to optimize both rapid-response (BESS) and long-duration (HES) balancing (Diabate et al., 2024). In such frameworks, a hierarchical EMS coordinates electrolyzer and battery charging—absorbing short-term renewables with BESS and longer surpluses as hydrogen for use during extended deficits. The technological mix is determined by round-trip efficiency, capital cost, degradation, and application timescale.

At the building/residential scale, H₂ storage is integrated with PV, BESS, heat pumps, and combined heat-and-power via PEMFCs, enabling seasonal shifting of renewable energy and backup power, with system-level efficiencies up to 75–90% when utilizing both power and heat outputs (Kovačević et al., 6 Nov 2025).

For long-duration (multi-GWh to TWh-scale) storage, geological storage in salt caverns, aquifers, and depleted hydrocarbon reservoirs enables multi-month to annual balancing, requiring robust systems modeling of cushion gas impacts, thermodynamics, and deliverability constraints (Franzmann et al., 16 Apr 2025, Zhao et al., 2023).

2. Thermodynamic and Process Modeling of Hydrogen Storage

Energy storage in hydrogen proceeds via sequential conversion processes:

• Water electrolysis: $P_{el} \xrightarrow{\kappa_{el}} H_2$
• Storage: pressurized, liquefied, or materials-based (adsorption, absorption), with associated losses and leakages
• Re-electrification: $H_2 \xrightarrow{\kappa_{fc}} P_{fc}$ or combustion

The mass and energy balances are governed by:

• $$q_{s,t}^{\rm H,prod} = \kappa_{\rm AE} P_{s,t}^{\rm AE}$$ (production),
• $$n_{s,t+1}^{\rm HS} = n_{s,t}^{\rm HS} + q_{s,t}^{\rm H,prod} \Delta T + q_{s,t}^{\rm H,purch} \Delta T - q_{s,t}^{\rm H,FC} \Delta T - q_{s,t}^{\rm H,sell} \Delta T - q_{s,t}^{\rm H,A} \Delta T$$ (inventory),
• $$P_{s,t}^{\rm FC} = \kappa_{\rm FC} q_{s,t}^{\rm H,FC}$$ (discharge).

Physical storage deploys compressed gas (CGH₂, up to 700–800 bar), liquid hydrogen (LH₂, 20 K), or solid-state storage (e.g., metal hydrides, physisorption, complex hydrides, high-entropy alloys). Associated performance metrics include gravimetric capacity (wt%), volumetric capacity (g H₂/L), enthalpy of formation, and kinetic parameters (Makridis, 2017, Mohammadi et al., 2023, Yadav et al., 3 Sep 2025, Yan et al., 2023).

Solid-state storage achieves reversible uptake by physisorption (MOFs, COFs, 2D materials), with binding enthalpies $\Delta H_{ads}\sim$ 7–15 kJ/mol for practical cycling at 233–298 K and 1–10 bar (Pakhira et al., 2019, Shinde et al., 2016, Oliveira et al., 2024). Chemisorption (e.g., MgH₂, HEA hydrides) yields higher densities but requires engineered thermodynamics ($|\Delta H|$ in 15–24 kJ mol⁻¹ H₂ target range) and nanoconfinement to circumvent sluggish kinetics and high desorption temperatures (Makridis et al., 2013, Yan et al., 2023).

Subsurface storage in salt caverns is modeled by integrating the real-gas equation of state, mechanical stability, cavern leaching logistics, and storage cycle design, supporting multi-hundred TWh capacities at country scale (Franzmann et al., 16 Apr 2025).

3. Optimization, Control, and Economic Modeling

Optimal HES deployment requires solving high-dimensional, multi-stage expansion planning and operational problems, typically formulated as mixed-integer or stochastic linear/convex programs. Key variables include component sizing (electrolyzer, storage, fuel cell), dispatch schedules, and market interactions, under renewables uncertainty and emissions constraints (Yu et al., 2023, Kayacık et al., 2022).

Efficient solution of large-scale expansion planning models is enabled by Dantzig–Wolfe decomposition with column generation, exploiting block-angular structure to decouple annual subproblems linked by capacity-coupling constraints (solved for up to $6 \times 10^6$ variables in practice) (Yu et al., 2023).

Real-time and day-ahead operational scheduling of hybrid ESS (BESS + HES) is increasingly based on Markov decision processes (MDP), deep reinforcement learning (DRL) with interpretable policy networks, and multi-agent algorithms (Xiong et al., 2023, Samende et al., 2022, Veenstra et al., 2021). Objective functions account for revenue/cost, component degradation, contract fulfillment (PPA, hydrogen offtake), grid balancing, and carbon emissions.

Levelized cost of energy (LCOE) and levelized cost of storage (LCOS) serve as key economic metrics. Detailed studies demonstrate that integration of HES reduces LCOE by 12–15% versus battery-only systems for deep decarbonization scenarios (LCOE down to 0.4324 RMB/kWh under full carbon neutrality) (Yu et al., 2023), but the round-trip efficiency of HES ($\kappa_{\rm FC}$0–$\kappa_{\rm FC}$1) significantly trails batteries ($\kappa_{\rm FC}$2), shifting techno-economic preference to hydrogen as renewable penetration and seasonal mismatch increase (Le et al., 2022, Diabate et al., 2024).

Sensitivity analyses illustrate that HES becomes cost-effective under carbon taxes, high curtailment penalties, or the emergence of liquid hydrogen spot/offtake markets. Adoption is accelerated by multi-objective optimization frameworks (e.g., modified firefly algorithm) that co-optimize economic (NPV, LCOE) and reliability (self-sufficiency) metrics under component degradation and seasonal operation (Le et al., 2022).

4. Materials Science of Hydrogen Storage Media

HES performance is fundamentally limited by the materials used for hydrogen uptake and release. Key classes include:

• High-entropy alloys (HEAs) in Laves or BCC phases, engineered for interpolable binding energies near –0.1 eV/H for room-temperature reversibility, achieving 1.4–2.1 wt% H₂ and fast cycling ($\kappa_{\rm FC}$3 min) with 1000+ cycles stability (Mohammadi et al., 2023, Yadav et al., 3 Sep 2025).
• Nanoconfined hydrides (MgH₂ in rGO–organosilica): particle confinement ($\kappa_{\rm FC}$4) and carbon scaffolds reduce desorption temperature ($\kappa_{\rm FC}$5–160°C), increase rate (up to 4$\kappa_{\rm FC}$6 bulk MgH₂), and limit decrepitation (Yan et al., 2023, Makridis et al., 2013).
• Physisorptive frameworks (MOFs, COFs, TMDs): optimized via DFT for binding enthalpy, pore size, and accessible surface; practical implementation targets ΔH = 7–15 kJ mol⁻¹ and 5–7 wt% at 1–10 bar, 233–298 K (Pakhira et al., 2019, Shinde et al., 2016, Oliveira et al., 2024).
• Functionalized graphene, pillared and decorated for tunable physisorption or chemisorption (graphane, Kubas-type TM complexes), targeting theoretical gravimetric densities $\kappa_{\rm FC}$7 wt% (Tozzini et al., 2012, Pakhira et al., 2019).

Quantum-level calculations decompose H₂ binding energy into dispersion, electrostatics, Pauli, and orbital terms, guiding ligand/metal selection and framework design. Aggregation, toxicity (e.g., BeO), kinetics at ambient temperature, and volumetric capacity are the major practical bottlenecks.

5. Macro-Scale Infrastructure: Geological Storage and Sector Integration

For grid-level and seasonal balancing, hydrogen storage using subsurface salt caverns offers the only currently demonstrated TWh-to-PWh scale, with high mechanical integrity, rapid injection/withdrawal, and exceptional cycle life (Franzmann et al., 16 Apr 2025, Makridis, 2017). Global screening studies find geologically and land-eligible salt cavern capacity sufficient to balance 43–66% of annual electricity demand worldwide, potentially rising to 85% with international storage-sharing (Franzmann et al., 16 Apr 2025).

Cushion gas selection (CO₂ vs. CH₄/N₂) in aquifer storage is shown to mediate trade-offs between hydrogen recovery rate, purity, and long-term pressure support (Zhao et al., 2023). Integration with hydrogen backbone pipelines and grid infrastructure is critical for sector coupling and energy transition goals.

6. Catalysis, System Efficiency, and Future Research Directions

Efficient HES requires ongoing advancement in electrolysis and fuel cell catalyst performance and cost. Generative machine learning—e.g., Catalyst GFlowNet—now accelerates the search for stable, low-overpotential HER/OER catalysts by coupling materials-structure generation with fast ML predictors for formation and adsorption energy, yielding validated rankings of elemental and multi-element catalysts for HES (Podina et al., 2 Oct 2025).

System-level round-trip efficiency for generation–H₂–electricity cycling remains lower than batteries; hybrid control logic and DRL frameworks are under active research to optimize HES dispatch in the presence of renewable/price uncertainty, minimize degradation losses, and exploit heat integration opportunities (Xiong et al., 2023, Kovačević et al., 6 Nov 2025).

Future directions include:

• Engineering hydride and adsorbent materials for higher capacity and lower enthalpy (targeting $\kappa_{\rm FC}$8 in 15–24 kJ/mol H₂),
• Scale-up and techno-economic analysis of subsurface and distributed HES,
• Regulatory and market design for hydrogen as an energy commodity,
• System integration studies to regionally optimize the mix of batteries, HES, and sector-coupled hydrogen applications (industry, heating, transport).

7. Quantitative Performance and Comparative Metrics

A summary of major hydrogen storage methods and key parameters:

Storage Type	Typical Capacity (wt%)	Operating Conditions	Key Limitation
Pressurized Gas	n.a. (120–140 MJ/kg)	200–700 bar, 300 K	Volumetric density, safety
Liquid H₂	n.a.	20 K, 1 bar	Boil-off, liquefaction energy
Metal Hydrides	1–6.5	20–400°C, 1–30 bar	Kinetics, heat management
Physisorption	2–10	77–298 K, 1–50 bar	Low binding at 298 K
HEA (Laves/BCC)	1.4–2.1	RT–300°C, 1–50 bar	Moderate capacity, temp
Porous MOFs/COFs	5–7 (theor.)	298 K, 1–10 bar	Synthesis, stability
Salt Caverns	n.a. (up to 100 GWh+)	200–1000 m depth, 30 bar	Geography, brine disposal

Performance and design must be matched to application (hourly/daily/off-grid/seasonal)—with BESS dominating for short-duration, and HES/salt caverns required to achieve deep decarbonization and grid autonomy on annual time scales.

---

References are to key arXiv papers: (Yu et al., 2023, Diabate et al., 2024, Yadav et al., 3 Sep 2025, Pakhira et al., 2019, Veenstra et al., 2021, Makridis, 2017, Le et al., 2022, Yan et al., 2023, Xiong et al., 2023, Kovačević et al., 6 Nov 2025, Zhao et al., 2023, Samende et al., 2022, Kayacık et al., 2022, Tozzini et al., 2012, Podina et al., 2 Oct 2025, Franzmann et al., 16 Apr 2025, Mohammadi et al., 2023, Makridis et al., 2013, Shinde et al., 2016, Oliveira et al., 2024).