Hydrogen storage with gravel and pipes in lakes and reservoirs

The study addresses the challenge of large-scale, seasonal energy storage required to enable high shares of renewable energy, particularly for seasonal balancing of solar generation. While salt caverns and depleted gas reservoirs are leading options for large-scale hydrogen storage, their geographic availability is limited. The authors propose and assess an alternative: storing hydrogen underwater in lakes, hydropower, and pumped storage reservoirs using gravel-ballasted HDPE pipe tanks. The research questions are whether such systems are technically feasible, cost-competitive, environmentally compatible, and what the global storage potential could be. The purpose is to expand viable hydrogen storage siting options, support seasonal storage, and facilitate a future hydrogen economy.

Prior work on underwater compressed gas energy storage (UWCGES) has focused mainly on the deep ocean and on compressed air rather than hydrogen. Two main accumulator types are identified: flexible accumulators (e.g., Energy Bag, Hydrostor’s flexible bags) that vary volume but face durability challenges in harsh marine environments; and rigid accumulators (steel/reinforced concrete) that displace water to maintain near-constant pressure but may suffer long-term issues like concrete cracking and corrosion. Underwater storage concepts leveraging ambient hydrostatic pressure remove the need for thick pressure vessels. Similar ideas have been proposed for storing hydrogen in pipelines filled with sand in the deep ocean. However, studies on lakes and reservoirs are comparatively sparse. This work fills that gap by proposing lake/reservoir siting and pipe-with-gravel tanks, with an explicit techno-economic assessment and global potential estimate.

Concept and system: Hydrogen is stored in HDPE pipeline tanks placed on the bottom of lakes or reservoirs and ballasted with gravel to counter buoyancy. The tank internal pressure is maintained equal to the surrounding hydrostatic pressure by simultaneously exchanging water and hydrogen during charging/discharging. Hydrogen’s low solubility in water allows contact with water without membranes.

Materials and properties: HDPE (PE100) is selected for tanks due to low H2 permeation (permeability coefficient ~1e−15 mol m m−2 s−1 Pa−1), ease of handling, and cost. A nominal wall thickness of 10 cm and HDPE density ~945 kg m−3 are assumed. Gravel (>5 cm granularity, density ~2666 kg m−3, porosity ~40%) from mine waste is used as ballast and does not degrade.

Operation: During charging, hydrogen is injected at the top of the reservoir and water is withdrawn from the tank bottom; during discharging, hydrogen is withdrawn and water is injected. Two analog pressure relief valves regulate water inflow/outflow to keep internal pressure equal to ambient. Hydrogen dissolution losses per cycle are negligible (mole fraction solubility at 0 °C ranges ~0.00004–0.0009 xH2(%) for 1–50 bar).

Thermodynamics and storage density: Tank pressure PT = Pa + D/10.2, where Pa ≈ 1 bar and D is depth in meters. At 15 °C, hydrogen density increases with depth/pressure (e.g., ~0.838 kg m−3 at 100 m; ~7.864 kg m−3 at 1000 m). Hydrogen density decreases with temperature (~0.36% per °C near 10 bar). Energy density rises with depth (e.g., ~27.9 kWh m−3 at 100 m; ~261.9 kWh m−3 at 1000 m, assuming 33.3 kWh kg−1 H2).

Buoyancy and ballast sizing: Gravel mass/volume is sized so the combined weight of gravel, HDPE, contained water, and hydrogen exceeds the buoyant force. An inequality relating volumes and densities (water, gravel solids, hydrogen, HDPE) determines required gravel to prevent flotation. At 200 m depth, the maximum hydrogen volume fraction (including voids) without submergence is ~62.53% of tank total volume.

Costing and LCOS: A reference tank uses a 10 m diameter, 100 m length HDPE pipe (internal volume ~7850 m3). At 200 m depth and 15 °C, usable hydrogen volume is ~4836 m3 at ~20.6 bar. Capital cost components include pipe, gravel, valves/sensors, and construction. O&M is 5% of CAPEX per year, lifetime 30 years, interest rate 8%, and an assumed 3 cycles per year for LCOS. CAPEX and LCOS are evaluated across depths; deeper siting reduces unit costs via higher storage density.

Case study (Oroville Lake, CA): Assumes bottom temperature 8 °C. A single 10 m × 100 m tank stores ~4836 m3 H2 at ~1.65 kg m−3 (≈7983 kg H2, ≈186 MWh electricity assuming 70% generation efficiency). The bottom area allows 462 such tanks, totaling ~3.68 million kg H2 (~86 GWh electricity). A 400 MW PV plant (19% capacity factor) supplies a constant 70 MW load with a storage system comprising batteries (daily cycling), electrolysis, and fuel cells (330 MW conversion capacity) and 86 GWh H2 storage; curtailment is ~14.6%.

Global potential estimation: Databases of artificial reservoirs and lakes are screened, using total area, volume, location, and average depth. Lacking bathymetry and maximum depth, the analysis assumes: storage depth equals average depth; usable storage area equals 10% of reservoir/lake surface, with 90% of that deployable; tanks are anchored; candidates have average depth ≥30 m (to allow parts to reach ≥100 m in some sites). Resulting sets: 3403 reservoirs and 1760 lakes. Regional potential and cost curves are derived, including sensitivity to cost uncertainties.

Environmental/LCA considerations: HDPE cradle-to-gate emissions are ~1.6 kg CO2/kg. A 300-ton tank implies ~480 t CO2 to produce. For comparison, generating 186 MWh in a gas plant emits ~77 t CO2 (413 kg/MWh). With seasonal operation, ~6.2 years are needed for the tank to store an amount of energy equivalent to the gas plant’s CO2 emissions. With renewable feedstocks and CO2 utilization (e.g., MTO), tanks could embody negative emissions (≈85.7% carbon by mass; ~257 t C ≈ 942 t CO2 stored).

• Technical feasibility: Hydrogen’s solubility in water is extremely low (e.g., ~0.0004% molar at 20 bar), making dissolution losses per cycle negligible (≤0.0009% at 0 °C up to 50 bar). Maintaining tank pressure equal to ambient hydrostatic pressure enables thinner, cheaper HDPE tanks.
• Storage density: Hydrogen density increases with depth (e.g., ~0.838 kg m−3 at 100 m; ~7.864 kg m−3 at 1000 m at 15 °C). Energy density scales with depth (~27.9 to ~261.9 kWh m−3 from 100 to 1000 m).
• Volume fractions: At 200 m, maximum hydrogen volume (including voids) without submergence is ~62.53% of total tank volume.
• Cost competitiveness: At 200 m depth, the levelized cost of hydrogen storage (LCOS) is ~0.17 USD kg−1. CAPEX for the 10 m × 100 m reference tank: ~30,000 USD total (pipe 12k, gravel 8k, other equipment 2k, construction 8k); hydrogen storage CAPEX ~3.76 USD kg−1. Costs decrease with depth due to higher storage density.
• Case study (Oroville Lake): One 10 m × 100 m tank stores ~4836 m3 H2 at ~20.6 bar and ~1.65 kg m−3 (~7983 kg H2), equivalent to ~186 MWh electricity (70% efficiency). The site can host 462 tanks, totaling ~3.68 million kg H2 (~86 GWh). A 400 MW PV plant (19% CF) serving 70 MW, with batteries + electrolysis + 330 MW fuel cells and 86 GWh H2 storage, meets demand with ~14.6% solar curtailment. Hydrogen density (and stored energy) varies with reservoir level.
• Global potential: Lakes offer ~12 PWh hydrogen storage capacity; reservoirs add ~3 PWh. The Caspian Sea contributes ~6.4 PWh. Excluding the five largest lakes, lake potential is ~1.9 PWh. Top lakes by potential include Baikal (~1.96 PWh), Tanganyika (~1.57 PWh), Superior (~1.02 PWh), and Malawi (~0.65 PWh). By region, potential is led by Former Soviet Union (~8.83 PWh), North America (~2.61 PWh), and Sub-Saharan Africa (~2.60 PWh). Sensitivity analyses show a range of CAPEX outcomes.
• Comparison with alternatives: LCOS (USD kg−1): lakes/reservoirs ~0.17; salt caverns ~0.11; depleted gas fields ~1.07; rock caverns ~0.23; pressurized containers ~0.17. Capacity ranges and cycle profiles indicate suitability for seasonal to multi-week storage.
• Land/water footprint: For the Oroville scenario, the hydrogen storage area (~0.46 km2) is ~38× smaller than the PV generation area (~17.4 km2).

The proposed lake/reservoir-based hydrogen storage addresses geographic limitations of subsurface caverns by leveraging abundant deep freshwater bodies. By matching internal and ambient pressures, tanks avoid large cyclic stresses, potentially improving durability. Low hydrogen solubility allows direct contact with water, simplifying design without membranes. The system is well suited for seasonal storage while also enabling daily/weekly compressed air storage in empty tanks, enhancing flexibility.

Integration with hydropower and pumped storage offers additional benefits: on-site renewable generation for electrolysis reduces transmission losses; reservoir depth and stable hypolimnion temperatures (6–15 °C below ~40 m) support stable storage conditions; and co-benefits such as freshwater-based district cooling or improved liquefaction efficiency are possible using cold deep water. The cost results (LCOS ~0.17 USD kg−1 at 200 m) are competitive with other large-scale options, with deeper siting further lowering costs via higher storage densities.

Policy implications include the need for standards and permitting for underwater hydrogen storage, incentives for deployment, and regional coordination to connect multiple storage sites into resilient renewable grids. Globally, the approach complements clean energy transitions by providing scalable, cost-effective seasonal storage. Risk mitigation (e.g., monitoring vessel traffic, protective nets) can reduce mechanical damage risks. Environmental and life-cycle aspects are discussed, including pathways to lower or even negative embodied emissions in tank materials with renewable production routes. Overall, the findings indicate that lake/reservoir hydrogen storage can meaningfully expand siting options, reduce area footprints relative to generation, and support high-renewable systems.

This work proposes and evaluates hydrogen storage in gravel-ballasted HDPE pipe tanks placed at depth in lakes, hydropower, and pumped storage reservoirs. The concept is technically feasible due to hydrogen’s very low solubility in water and offers competitive costs (LCOS ~0.17 USD kg−1 at 200 m). A detailed case study (Oroville Lake) demonstrates practical seasonal storage scales (~86 GWh) with modest seabed footprint compared to generation area. A first global assessment indicates substantial potential (~12 PWh in lakes and ~3 PWh in reservoirs), with notable contributions from large/deep water bodies (e.g., Caspian Sea).

Main contributions include: (i) defining a practical storage container (HDPE pipe with gravel ballast) and pressure-equalized operation; (ii) cost modeling with depth-dependent storage density; (iii) a case study integrating solar generation and hydrogen storage; and (iv) a global potential and regional cost-curve analysis. Future research should focus on: acquiring bathymetric data to refine site-specific potentials and costs; detailed engineering, reliability, and environmental assessments; optimization of tank geometry and materials; integration planning with hydropower and grid operations; and policy frameworks to enable safe, scalable deployment.

• Data constraints for global potential: lack of detailed bathymetry and maximum depths necessitated using average depth and assumed deployable area (10% of surface, 90% usable). Some reservoirs with ≥30 m average depth may still lack sufficiently deep zones (≥100 m) for cost-effective storage, introducing uncertainty.
• Case-specific assumptions: Oroville bottom temperature (8 °C) was assumed; tank spacing, anchoring, and protection measures are idealized.
• Cost uncertainties: Pipe transport and underwater construction costs can vary substantially; gravel availability and transport can increase costs; connection to future hydrogen grids was excluded.
• Operational variability: Reservoir level fluctuations change tank pressure and storage density; potential energy-water conflicts may alter optimal operation.
• Material and environmental aspects: While hydrogen solubility losses are negligible, long-term material performance and environmental impacts (e.g., HDPE life-cycle impacts) warrant further study; risk of physical damage from heavy objects requires mitigation measures.
• Minor inconsistencies in reported parameters (e.g., cycles/year text vs. value; tabulated hydrogen density units) do not affect overall conclusions but indicate areas for data clarification.