Kilowatt-scale solar hydrogen production system using a concentrated integrated photoelectrochemical device

The work addresses the need to scale solar fuel technologies, particularly solar hydrogen production, beyond small laboratory demonstrations to pilot-scale, on-sun operation. While photoelectrochemical (PEC) cells, integrated photovoltaic-electrolyser (PV+EC) systems, particulate photocatalytic systems, and thermochemical cycles have shown promise, most demonstrations remain below 100 W output. Scaling these systems poses challenges involving achieving high efficiency and production rates, maintaining stability, managing costs, and ensuring sustainability. Solar concentration is identified as a promising route to higher power density and potentially improved efficiency while co-generating useful heat. Prior thermal integration of PV and proton exchange membrane (PEM) electrolysis has shown synergistic benefits from optimized thermal management but at limited hydrogen output power and often under laboratory illumination. Critical device-level challenges include careful thermal management, appropriate PV–EC coupling, operating condition control, and respecting material limits (PV, PEM). The authors present a pilot-scale, on-sun, thermally integrated concentrated PV–PEM system designed to demonstrate feasibility at kilowatt scale and to study control strategies and performance under real environmental conditions.

The paper situates its contribution within several strands of solar-to-fuels research. Integrated PV+EC and PEC devices have achieved high solar-to-hydrogen (STH) efficiencies at small scale, including demonstrations with tandem III–V photoabsorbers and integrated devices achieving STH efficiencies up to ~16–20% (depending on definition and conditions). Past integrated thermal designs exploited PV cooling and reactant/catalyst heating to reduce overpotentials, but experimental hydrogen output powers typically remained below ~32 W and often used solar simulators rather than on-sun conditions. Comparative large-scale demonstrations in related solar fuels domains include particulate PEC panels achieving ~0.76% STH at ~700 W output and thermochemical redox systems achieving a few percent system-level efficiency at a few hundred watts. Prior modeling highlighted the importance of thermal integration, dynamic control (e.g., flow-rate control to adjust operating points), and identified challenges under fluctuating irradiance and component failures. Economically, solar concentration enables higher power density and may justify more costly high-performance photoabsorbers, while offering co-generation of heat. The present work builds on a prior lab-scale integrated PEC demonstration (>15% STH at 32 W H2 power under high-flux simulator) by scaling to kilowatt-level, on-sun operation with concentrated light and thermal integration using commercially mature components, notably a PEM electrolyser.

System architecture: A 7 m-diameter dual-axis tracking parabolic dish (38.5 m2 aperture) focuses sunlight to a reactor located at the focal point. The reactor contains a water-cooled flux homogenizer, a triple-junction III–V concentrated PV (CPV) module, and two 16-cell PEM electrolyser (EC) stacks thermally integrated via a shared deionized (DI) water loop. A water-cooled shield protects against excess concentrated light due to an oversized dish relative to reactor aperture. Hydraulic design and thermal integration: Two water loops are used: a global DI water flow from ground-level storage through deionizers to the reactor, and a PV recycle loop via a small centrifugal pump to recirculate water across the CPV heat exchanger. This decouples PV cooling flow from EC stoichiometric needs, enabling adequate CPV heat removal, controlled temperature rise to the EC (typically 30–90 °C depending on flow), and high stoichiometric water excess to the EC. Wetted copper surfaces in the CPV heat sink were ALD-coated (Al2O3/TiO2, ~50 nm) to ensure DI water compatibility. The homogenizer (stainless steel with internal water channels and reflective inner faces) converts the near-Gaussian flux to a rectangular, more homogeneous profile matching the CPV active area. Product handling and heat recovery: Anodic (O2 + H2O) and cathodic (H2 + H2O) streams are routed to ground-level. The anodic stream is cooled via a liquid–liquid heat exchanger; custom liquid–gas separators remove water for recycle to the storage tank. A back-pressure regulator maintains H2 delivery pressure (1–30 bar); O2 is near atmospheric pressure. Gases are stored in compressed quads or vented. Co-generated heat is recovered as warm water; system outlet temperatures are measured upstream of the heat exchanger. Instrumentation and control: Temperatures (K-type thermocouples in reactor, PT100 at in/out), pressures, conductivities, and flow rates are monitored. H2 crossover is monitored by an inline flammable gas sensor. Electrical sensors record integrated device current and voltage. A supervisory control and data acquisition system (LabVIEW) automates operation of ~30 valves, ~60 sensors, and two pumps; dish tracking is controlled by a dedicated PLC. Commissioning and optical characterization: Individual PV and EC components were characterized using a 15 kW bidirectional power supply; PV performance was validated at 1 sun and 700 suns. The incident flux distribution was mapped using a water-cooled Lambertian target, CCD camera, and flux gauge, with coaxial imaging to avoid geometric correction and sub-mm spatial resolution. Flux maps across planes characterized the dish–homogenizer optics. Experimental campaign: On-sun tests over >13 days (Aug 2020; Feb–Mar 2021) covered varying ambient conditions (∼8–20 °C) and DNI, including clear and hazy/cloudy periods. Typical operating points: global flow ~4.92 l/min, PV recycle flow ~10.3 l/min; anode pressure ~3.5 bar, cathode ~29 bar; system inlet ~14.4 °C and outlet ~45.1 °C; reactor outlet 60–70 °C. External auxiliary electrical demand was ~0.58 kW. Performance metrics: Solar input power Qsolar = DNI × Adish. Fuel power Qfuel = I × NEC × ΔnH2/F (with NEC=32 cells in series, Faradaic efficiency ≈1). ΔnH2 is multiplied by either reaction enthalpy (286 kJ/mol, HHV) or Gibbs free energy (237 kJ/mol) to compute HHV- or Gibbs-based fuel power. Heat power Qthermal = ṁg × Cp × (Toutlet − Tinlet). System fuel efficiency = Qfuel/(Qsolar + Qexternal). System thermal efficiency = Qthermal/(Qsolar + Qexternal). Diagnostic device efficiency (IPEC device-level) = Qfuel/QPV, where QPV is solar power at the reactor aperture; ratio Qsolar/QPV estimated at 27.5% from flux calibration. Modeling: A zero-dimensional steady-state model with componentwise mass/energy balances, detailed PV and EC electrical models (including light inhomogeneity), and connections by energy streams (light, heat, electricity) was developed and validated against experimental data. Parametric studies explored effects of flow rates, EC cell count, PV capture fraction/area scaling, and light homogeneity on fuel and heat outputs.

- First on-sun kilowatt-scale demonstration of a thermally integrated concentrated PV–PEM system co-generating hydrogen and heat. - Device-level (diagnostic) STH efficiencies: average 20.3% (Gibbs) and 24.4% (enthalpy) over the campaign; best day 22.6% (Gibbs) and 27.2% (enthalpy). Achieved >20% at >2.0 kW H2 power, representing a >100× increase in H2 power compared to prior thermally integrated lab-scale devices (~32 W). - System-level efficiencies (including 0.58 kW auxiliaries): fuel efficiency 5.5% ± 0.5% (Gibbs) or 6.6% ± 0.6% (HHV); thermal efficiency 35.3% on average. - Hydrogen production performance: instantaneous up to 0.9 Nm3/h; mean 0.59 Nm3/h (49.7 g/h) over >13 days, corresponding to ~41.3 A EC current on average. Peak over 5 min: 14.0 Nl/min (1.26 g/min). Total H2 produced >3.2 kg. - Heat co-generation: average thermal output 10.6 kW at ~45.1 °C; peak 14.9 kW over 5 min; total 679 kWh thermal over the campaign. Thermal integration reduced auxiliary electrical demand by eliminating a separate EC heater (estimated ~0.6 kW), lowering balance-of-plant complexity and cost. - Operating conditions: global water flow ~4.92 l/min (λ≈1460 at 60 A), PV recycle flow ~10.3 l/min; anode pressure ~3.5 bar, cathode ~29 bar; typical inlet/outlet water temperatures 14.4 °C/45.1 °C; reactor outlet 60–70 °C. - Correlations and model validation: Fuel and heat powers scale approximately linearly with DNI: Qfuel [kW] = 3.1955 × DNI [kW/m2] − 0.4963 (R2=0.9137); Qthermal [kW] = 14.4998 × DNI [kW/m2] − 0.6145 (R2=0.8649). Higher operating temperatures reduce overpotentials and EC voltage. The model reproduces observed trends. - Optical/thermal losses and scale compromises: Only ~27.5% of total solar power reached the PV aperture; ~52.1% was reflected/absorbed by the shield and homogenizer due to an oversized dish vs reactor and non-ideal optics. Pipe heat losses were significant, lowering thermal output and outlet temperatures, especially in winter. - Control insights: Water flow-rate control can stabilize outlet temperature under fluctuating DNI; increased flow slightly reduces fuel power (−0.03 kW per l/min at ~4.9 l/min) but increases heat power (+1.37 kW per l/min) while lowering outlet temperature (−4.4 °C per l/min). - Optimization projections (validated model): Redirecting shield light to PV, scaling PV area to maintain concentration, and improving light homogeneity by 90% can raise system STH efficiency to ~15.9% (Gibbs)/~19.2% (enthalpy) with component efficiencies of ~72.6% optical, 37.3% PV, 59.7% (Gibbs)/71.8% (HHV) EC, and 98.5% balance-of-plant. Dish optical efficiency improved to ~90% (state-of-the-art mirrors) could yield ~19.7% (Gibbs)/~23.8% (HHV) system STH. With 20 mm pipe insulation (UA≈6 W/K), projected system heat efficiency is ~49.3% with outlet temperatures ~80 °C. - Dynamic performance: Startup/shutdown ~5 minutes; stable operation across summer/winter conditions with no observed degradation over the test periods.

The results directly address the central challenge of scaling solar hydrogen technologies by demonstrating robust, on-sun, kilowatt-scale operation of a thermally integrated concentrated PV–PEM system with high device-level STH efficiency and meaningful co-generated heat. Thermal integration achieves a synergistic improvement: PV cooling sustains high electrical performance while elevating feed water temperature to the PEM stack, reducing overpotentials and stack voltage. The validated model captures key dependencies, such as linear scaling of fuel and heat with DNI and the strong influence of operating temperature on EC voltage, and explains underperformance at low DNI due to circumsolar radiation outside the effective acceptance angle of the dish–homogenizer optics. A Sankey-based energy accounting highlights that present system efficiency is limited primarily by optical mismatches (oversized dish, shield/homogenizer losses) and thermal losses in piping, not by intrinsic PV or EC conversion. Control strategies, particularly water flow regulation, mitigate fluctuations in irradiance by stabilizing outlet temperatures with limited penalties to fuel power. The modeling framework identifies straightforward optimization pathways—better optical matching to increase PV capture, improved light homogeneity, targeted flow control, and enhanced thermal insulation—that can substantially raise system-level STH efficiency toward device-level values. These findings validate thermal integration for co-production of hydrogen and low-grade heat and outline practical steps for advancing toward commercial viability, including potential integration with storage to smooth DNI variability and expansion to coupled power/heat/fuel delivery.

A pilot-scale, on-sun, thermally integrated concentrated PV–PEM system was designed, constructed, and operated, delivering large-scale co-generation of hydrogen and heat. The system achieved >20% device-level STH efficiency at >2.0 kW hydrogen power, with system-level fuel efficiency of ~5.5% (Gibbs)/~6.6% (HHV) and thermal efficiency of ~35%. Over >13 days of operation, the plant produced >3.2 kg H2, demonstrated rapid startup/shutdown (~5 minutes), and maintained stable operation across varying meteorological conditions without observable degradation. Detailed diagnostics and a validated steady-state model identified dominant energy losses (optical and piping heat losses) and provided clear, practical optimization routes—improving optical capture/matching and light homogeneity, reducing thermal losses, and refining flow control—that can elevate system STH efficiency to >16% (Gibbs) and potentially approaching ~20% with state-of-the-art optics. Future work should pursue multi-year durability studies, advanced control for flexible co-production of hydrogen, power, and heat, and integration with storage and downstream fuel synthesis to enhance capacity factor and extend application to residential and industrial heat use cases.

- Optical mismatch due to an oversized dish relative to reactor aperture led to substantial shield/homogenizer losses (only ~27.5% of solar power reached the PV), lowering system efficiency. - Significant thermal losses in piping reduced delivered outlet temperatures and heat efficiency, particularly in cooler ambient conditions; insulation was not optimized. - Performance under low DNI was reduced due to increased circumsolar radiation outside the optical acceptance of the dish/homogenizer. - Light inhomogeneity at the PV limited benefits from increasing electrolyser cell count beyond current design (~32 cells). - Some balance-of-plant components and operating parameters were not optimized for efficiency at pilot scale (e.g., auxiliary loads, optical surfaces, shield design). - The experimental campaign, while spanning seasons, covered ~13 operating days; long-term durability and degradation under intermittent operation require multi-year validation. - Control complexity arises from coupled PV–EC thermal and electrical behavior; only initial flow-rate control strategies were explored.