1. Introduction
Against the backdrop of the global transition in the energy landscape and increasingly stringent carbon-reduction targets, harnessing renewable electricity to electrolyze water and generate hydrogen has emerged as a pivotal technology for realizing green, low-carbon development [
1]. Among the available options, alkaline water electrolysis (AWE) has been extensively deployed in large-scale hydrogen-production projects owing to its technological maturity and significant cost advantages [
2]. Nevertheless, the stochastic power fluctuations inherent to wind and solar resources directly undermine the operational stability of AWE systems. Conventional AWE installations are typically engineered for quasi-steady, constant-power conditions. When the supply power varies rapidly, critical cell parameters—temperature, pressure, and hydrogen-in-oxygen impurity (HTO)—become difficult to stabilize, thereby inducing fluctuations in gas yield and purity [
3]. Consequently, the intermittent and variable nature of renewable generation impedes the long-term continuous operation of AWE, and the attendant frequent load swings and start–stop cycles further challenge system stability.
Under fluctuating operating conditions, controlling HTO becomes particularly challenging because its dynamic behavior is characterized by pronounced inertia and hysteresis. Experimental studies show that abrupt cell current changes trigger rapid electrical and pressure responses within seconds, whereas the HTO concentration exhibits a delay of several minutes due to inertia [
4]. To prevent HTO from surpassing its safety threshold (2%), industrial AWE plants impose a minimum load limit of roughly 10–40% [
3]. Operating below this limit causes the HTO content to rise rapidly to hazardous levels, necessitating an emergency shutdown. Recent studies have proposed various strategies and models to address dynamic control challenges associated with HTO. Haug et al. [
5,
6,
7] first established a comprehensive gas-purity mechanistic model incorporating diffusion, convection, and alkali mixing, and quantified the influence of operating variables—current, temperature, pressure, and electrolyte flow rate—on HTO. Qi et al. [
8] simplified the gas-purity framework and, for the first time, captured dynamic impurity accumulation, proposing an MPC-based pressure-control scheme that markedly reduces HTO. Zhang et al. [
9] systematically reviewed mechanistic models (diffusion, convection, and electro-osmosis) and empirical correlations (temperature—current and pressure coupling) for HTO in alkaline electrolyzers, and introduced a dynamic accumulation model to overcome the limitations of steady-state descriptions. This model predicts HTO evolution under power fluctuations, enables the timely adjustment of operating parameters, and prevents impurity excursions, thereby providing a theoretical basis for real-time prediction and optimization. Gu et al. [
4] characterized the dynamic responses of current, voltage, temperature, HTO, and pressure during cold-start, hot-start/stop, and variable-load operations on a 250 kW industrial AWE plant across three time scales—hours, minutes, and seconds. Their results elucidated the distinct time delays and inertial traits associated with each transient. Jang et al. [
10] assessed, via numerical modeling, the impact of operating pressure on gas purity and determined that elevating the pressure from atmospheric to 20 bar significantly enhances hydrogen purity and suppresses HTO; further pressurization, however, yields diminishing returns and must be balanced against the additional balance-of-plant energy demand. Ren et al. [
11] performed cold-start and wind-powered operating tests on a 250 kW demonstration system and developed a semi-empirical model that captures the coupled effects of pressure, temperature, and current density. Model parameters were identified by nonlinear regression, and validation against demonstration data yielded a maximum HTO prediction error of only 3.9%. Tromans et al. [
5] derived accurate fundamental parameters for gas dissolution and mass transfer by systematically measuring the solubility and diffusion coefficients of hydrogen and oxygen in KOH solutions over a range of temperatures and concentrations, thereby furnishing reliable inputs for subsequent mechanistic HTO models that quantify dissolved-gas permeation through membranes. Although numerous studies have examined the influence of operating parameters on HTO through laboratory-scale models and rigs, investigations based on industrial-scale alkaline electrolyzers remain scarce.
Meanwhile, frequent renewable-power fluctuations and electrolytic cell start–stop cycles create safety hazards and significantly reduce system efficiency and lifespan. Operating at partial—or particularly low—loads substantially degrades electrolytic hydrogen-production efficiency [
12]. Kojima et al. [
12] demonstrated that when the electrolysis current falls below 50% of its rated value, auxiliary energy demands and parasitic losses precipitate a sharp decline in system efficiency; even if the cell continues to operate, the specific energy consumption per unit of hydrogen rises markedly. Moreover, when wind or photovoltaic supply is insufficient, the plant must shut down; repeated cold- and hot-start operations incur heat losses and additional energy expenditure to re-heat the electrolyte and restore the operating temperature, thereby further diminishing overall energy efficiency. Frequent load swings and start–stop sequences also accelerate cell ageing and shorten service life. During power transients, the electrode potential changes abruptly; at shutdown, reverse currents induced by polarity reversal trigger redox cycling on the electrode surface and expedite degradation of the catalytic layer. San Martín et al. [
13,
14] reported that the lifetime of Ni-based electrodes diminishes significantly under repeated start–stop cycles. Experimental data show that electrode performance deteriorates markedly after roughly 5000–10,000 start–stop cycles, well below the life obtainable under continuous operation. Direct integration with intermittent sources like photovoltaics may subject a single electrolyzer stack to tens of thousands of start–stop cycles over 20–30 years. Thus, persistent input-power fluctuations impose periodic mechanical and electrochemical stress on the cell, accelerating material fatigue and failure. Such degradation reduces hydrogen-production efficiency by increasing electrode overpotential, and also raises maintenance and replacement frequency, reducing the economic viability and availability of AWE systems.
Existing research on HTO concentration has largely focused on steady-state mechanisms, employing highly complex simulation models that pose challenges in calibration. Additionally, there is a notable lack of research on the dynamic behaviors of HTO under conditions of fluctuating power input. Studies addressing optimization and control strategies remain scarce, typically concentrating on single-parameter adjustments to improve HTO while neglecting adverse effects on other critical performance indicators, such as electrolysis efficiency. A literature analysis on the origins and influencing factors of HTO reveals that system pressure and electrolyte flow rate are the primary operational parameters. While reducing system pressure lowers hydrogen solubility, it inevitably raises compressor energy consumption and increases bubble coverage on electrode surfaces, consequently elevating electrolyzer energy consumption. Research has demonstrated that elevating operating pressure increases hydrogen concentration within oxygen streams. Sharshir et al. [
15] noted that while higher pressure minimally impacts electrolyzer voltage, it significantly reduces hydrogen purity due to increased hydrogen crossover from the cathode to the anode. In experiments with a 120 bar high-pressure electrolyzer, Janssen et al. [
16] found HTO rose from approximately 1.53% to 2.45% when pressure was elevated from 70 bar to 130 bar at 60 °C. On the other hand, the electrolyte inherently dissolves small amounts of generated gas. Zhou et al. [
17] confirmed this mechanism, highlighting that increased electrolyte circulation rates significantly raise HTO by transporting more dissolved hydrogen toward the oxygen side. Consequently, under high-pressure conditions, increased hydrogen solubility in the alkali solution and elevated pressure differentials across the diaphragm facilitate hydrogen diffusion and migration. Furthermore, pressure imbalances between the anode and cathode can abruptly drive significant gas volumes across the diaphragm. Numerous studies indicate that varying alkali solution flow rates, particularly in mixed circulation modes, substantially impact dissolved hydrogen ingress to the anode. Zhao et al. [
18] employed three-dimensional CFD to compare single- and multi-inlet concave–convex flow channels in alkaline electrolyzers. A three-inlet square design improved the electrolyte-velocity uniformity index to 0.80–0.88, markedly shrinking low-velocity vortex zones and, by facilitating bubble detachment, lowering energy losses and gas-purity penalties. Cheng et al. [
19] emphasized that mixed electrolyte circulation continually transfers dissolved hydrogen from the cathode to the anode, with flow rates influenced by electrolyte circulation speed and pressure differentials. Haug et al. [
5] systematically compared mixed and separated circulation modes under normal pressure in a zero-gap electrolyzer, demonstrating that reducing average flow rates in mixed circulation decreased HTO from 1.35% to 0.75% at low loads, while separated circulation at low flow rates further reduced it below 0.5%. Innovatively, Hu et al. [
20] proposed a coordinated control strategy for the electrolyte flow rate and system pressure to reconcile the conflicting demands of minimum load limits and efficiency in a 250 kW industrial alkaline electrolysis system under renewable energy fluctuations. Employing steady-state and dynamic gas purity models, they elucidated the mechanisms governing HTO. Their approach, involving strategic reductions in system pressure and electrolyte flow rate, increased wind and solar energy utilization rates to 98.3% and 95.6%, respectively, effectively mitigating elevated HTO levels at low loads. Haug et al. [
6] developed a simulation model accounting for alkali flow rate impacts, predicting and validating experimentally that a fivefold increase in alkali flow rate raised HTO from 0.2% to 1.1% under mixed circulation at 0.2 A/cm
2. They concluded that implementing separated circulation significantly reduces the influence of alkali mixing, rendering HTO predominantly determined by diaphragm diffusion. Finally, Brauns et al. [
21] experimentally compared three dynamic strategies under conditions of 7 bar and 60 °C: constant high-flow mixed circulation, periodic switching between mixed and separated circulation, and a linear adjustment of alkali flow rate with current density. Results indicated that the linear load-flow adjustment strategy maintained HTO concentrations between 1.2% and 2.0%, substantially lower than the fixed high-flow strategy (1.5–2.4%), thereby demonstrating the effectiveness of load-adaptive flow control in stabilizing HTO amidst renewable energy fluctuations.
In summary, despite notable advancements in mechanistic modeling, predictive-control methodologies, and flexible system integration, the majority of studies remain confined to low-power experimental rigs or numerical simulations. Systematic empirical datasets and methodological frameworks for defining the dynamic safety boundary of HTO, elucidating efficiency-degradation mechanisms, and formulating multi-parameter cooperative control laws for industrial-scale AWE under renewable-energy fluctuations are still absent. In particular, the coupling effects of key variables—current, pressure, and alkali flow rate—across different load regimes, along with the time-delay and inertial characteristics of gas post-treatment, remain insufficiently characterized. To address these gaps, this study establishes an industrial AWE test rig with a hydrogen-production capacity of 2 Nm3/h. By systematically varying current, pressure, and alkali flow, this study investigates the dynamics and governing factors of HTO in industrial-scale alkaline electrolyzers. In a simulated current ramp-down scenario, the rig evaluates how pressure and alkali flow-rate adjustments influence HTO dynamics and controllability. The research results provide an empirical basis for deeply integrated system design, real-time predictive control strategy development, and life cycle economic assessment of industrial AWE combined with intermittent renewable energy.
2. Materials and Methods
This experimental study was performed on a 10 kW industrial-scale AWE system. The system configuration is illustrated in
Figure 1, comprising the electrolytic cell, separation subsystem, electrical subsystem, control subsystem, purification subsystem, and auxiliary units. The electrolytic cell exhibits a hydrogen-production capacity of 2 Nm
3/h. The cell comprises 26 individual chambers, rated for a working pressure of 1.6 MPa, an operating temperature of 90 °C, a current density of 3500 A/cm
2, a nominal current of 200 A, and a cell voltage of 47.8 V. The separation subsystem supports stable cell operation and consists of a gas–liquid separator, wash tank, condensing tank, alkali circulation pump, deionized-water feed pump, heat exchanger, and control valves. Its primary function is to separate product gas from the alkaline electrolyte, facilitate electrolyte supply and forced-circulation cooling, and regulate the cell’s temperature and pressure. The hydrogen-production control system is implemented on a programmable logic controller (PLC) and acquires real-time sensor data—including pressure, temperature, flow rate, actuator position, and gas composition. The HTO analyzer used in this study (YF-H02, Xi’an Yunfeng Intelligent Technologt Co., Ltd, Xi’an, China) has a detection range 0–4%, accuracy 2%, and a response time of approximately 30 s. Therefore, HTO concentrations are now reported to two decimal places, and response delays to the nearest tenth of a minute (e.g., 2.0 min), reflecting the reliable accuracy and resolution of the sensor under the experimental conditions. This control system enables automated operation, continuous monitoring and fault diagnosis, and comprehensive data acquisition and analysis.
Industrial AWE systems typically employ gas chromatography or dedicated hydrogen analyzers to measure HTO concentration at the separation-system outlet. The electrolyte effluent from the cell traverses the gas–liquid separator, wash tank, and condenser before sampling by the HTO analyzer. As illustrated in
Figure 2, this sequence entails dynamic impurity accumulation via electrolyte recirculation through the cell and separator, compounded by separation-induced delays. Delay Step 1 denotes the cell’s oxygen purge time, primarily dictated by the alkaline flow rate and chamber volume. Delay Stage 2 represents the oxygen–alkali separation time in the gas–liquid separator, governed by its separation capacity. Inertia Link 1 corresponds to the total residence time from separation through condenser egress, influenced by oxygen flow rate, system pressure, temperature, and chamber volume.
Experiments were designed to study HTO behavior under various transient operating conditions. Each key experimental scenario (e.g., current drop/rise tests or pressure and flow rate changes) was performed only once, and the repeatability of the experiments needs to be verified. In a cold-start test, the electrolyzer current was increased from 0 to full load in 20 A increments (10 steps) over approximately 3.8 h, while monitoring HTO evolution from startup to steady state. We then carried out controlled current-fluctuation tests at a steady operating temperature of 90 °C. In a baseline scenario (Case 1), the system was maintained at 1.4 MPa pressure with a 7 L/min alkaline flow, and the cell current was stepped from 200 A (100% load) down to 120 A (60% load) and back up to 200 A. In Case 2, the same current profile was applied but with the system pressure reduced from 1.4 MPa to 1.0 MPa during the low-load period (followed by re-pressurization to 1.4 MPa). In Case 3, pressure was lowered to 1.0 MPa and held at that level throughout the current rise. Similarly, to investigate the effect of electrolyte circulation, we conducted a set of tests with varying flow rates. In flow Case 1, the flow rate was held constant at 7 L/min. In Case 2, the flow was reduced from 7 L/min to 3.5 L/min during the current downturn and then returned to 7 L/min for the ramp-up. In Case 3, the flow was lowered to 3.5 L/min and kept at that lower rate for the entire test. Throughout all experiments, the HTO concentration at the oxygen outlet was continuously monitored using a hydrogen analyzer (connected to the PLC data acquisition system), providing real-time measurements of the impurity level under each condition.
4. Discussion
Lowering the system pressure was found to directly decrease HTO by reducing hydrogen solubility in the electrolyte. At a reduced pressure of 1.0 MPa, less hydrogen dissolves into the KOH solution, and so less dissolved hydrogen can migrate to the oxygen side; as a result, the HTO concentration in the oxygen stream remains lower. Conversely, when pressure is raised back to 1.4 MPa, the increased solubility causes a rapid resurgence of hydrogen impurity, explaining the quick rebound in HTO. This indicates that a moderate pressure reduction can effectively suppress HTO in the short term, although it must be balanced against the energy cost and practicality of pressure changes. Likewise, reducing the alkaline solution flow rate markedly lowers HTO by limiting the convective carryover of dissolved hydrogen. A lower circulation rate (e.g., 3.5 L/min and 7 L/min) means the electrolyte transports less dissolved hydrogen from the cathode to the anode side, thereby decreasing the impurity level in the oxygen output. Additionally, we observed an HTO response delay of approximately 3 to 5 min following abrupt current changes. This inertia is attributable to the system’s gas residence time and the sequential purge and separation delays, which cause HTO to lag behind the current response. Recognizing these delay and inertia effects is important for control: it suggests that predictive strategies are needed to manage HTO spikes before they approach the 2% safety threshold.
Furthermore, the findings presented in this study have direct practical implications for industrial alkaline water electrolysis system operation, particularly regarding maintaining operational safety and efficiency under renewable energy fluctuations. Understanding the dynamics and time delays associated with hydrogen-in-oxygen concentrations enables plant operators to proactively implement control strategies, such as pressure adjustments and optimized electrolyte circulation rates, to maintain gas purity and prevent hazardous conditions. For instance, strategically lowering system pressure or adjusting electrolyte flow during low-load conditions can significantly mitigate impurity accumulation, thereby enhancing the safety and reliability of electrolyzer operation.
Intelligent control and management systems enhance stability and reliability, reduce O&M costs, and improve efficiency in alkaline electrolyzers. Intelligent control is realized through integration of high-precision sensors and advanced controllers for real-time monitoring and regulation of cell operation. Sensors monitor key parameters—temperature, pressure, and liquid level—in real-time, providing accurate and timely data to the control unit. The control unit performs real-time analytics, executes precise control decisions, and dynamically adjusts operating setpoints via embedded algorithms, thereby mitigating human-induced or parameter-driven fluctuations.
A digital-twin model centered on HTO integrates mechanistic and data-driven models to simulate system behavior in real time, predict HTO trends, and support fault diagnosis and control optimization. Jiang Yue [
22] developed a digital twin for alkaline electrolyzers leveraging their operational characteristics. By combining high-precision sensors, IoT connectivity, and simulation modeling, the twin provides real-time evaluation of impedance, temperature profiles, power-regulation behavior, and hydrogen-production efficiency. Emphasizing tank temperature as a critical indicator, the model simplifies mechanistic complexity and underpins precise operation and control of electrolysis cells. Building on this, Jiang Yue [
23] proposed a four-layer digital-twin architecture—physical, data-acquisition, simulation, and application layers—and detailed its use in fault diagnosis, parameter optimization, and asset-cluster management. They also identified challenges, including data-collection accuracy limitations, incomplete monitoring dimensions, and real-time model-update constraints. The work underscores how digital twins enhance monitoring fidelity and decision timeliness in O&M, offering robust digital tools for large-scale hydrogen-production system optimization.