Simulation and Validation of Green Hydrogen for the Production of Renewable Diesel: Case Study in La Guajira, Colombia

Abstract

This study validates green hydrogen (${\mathrm{H}}_2$) production from a 15 kWe wind–solar PV microplant under real operating conditions and quantifies the renewable diesel (RD) potential from oil hydroprocessing (with palm oil as the base case) via detailed stoichiometric balances. The electric output feeding two electrolyzers was monitored for six months (December 2024–May 2025). Three ${\mathrm{H}}_2$ production models were calibrated against the experimental results; the model with the best fit achieved ${\mathrm{R}}^2$ = 0.9848 and MSE = 130.05. Using the estimated ${\mathrm{H}}_2$ production, monthly balances were established for palm oil TAGs (POP, POO, POL, PLP, and SOS) across various deoxygenation routes—namely decarboxylation (DCX), decarbonylation (DCN), and hydrodeoxygenation (HDO)—with coproduct closure (propane, ${\mathrm{CO}}_2/\mathrm{CO}/{\mathrm{H}}_2\mathrm{O}$). The hybrid plant operated above the electrolyzers’ 2.88 kWe minimum, raising the effective ${\mathrm{H}}_2$ output (which peaked in February–March) and, thereby, the RD potential. The specific ${\mathrm{H}}_2$ demand followed the gradient of HDO > DCN > DCX; for POP, the global demand was 0.30 kg (saturation) + 1.20 kg (cracking) + 2.10 kg (DCN) or 2.55 kg (HDO), highlighting the carbon–hydrogen trade-off. The results indicate that ${\mathrm{green-H}}_2$–HDO integration is technically feasible and scalable in La Guajira; the choice of route (DCX/DCN vs. HDO) should align with local renewable availability to either maximize the liters of RD per kg ${\mathrm{H}}_2$ or conserve carbon.

1. Introduction

Global energy demand has risen steadily in recent decades, driven by population growth and industrialization [1,2]. Historically, this demand has been met largely with fossil fuels, whose combustion is a principal cause of increasing greenhouse gas (GHG) emissions and the resulting climate change challenge [3,4]. Moreover, dependence on these nonrenewable resources raises significant concerns about energy security and reserve depletion [5]. In this context, the search for sustainable, clean energy alternatives has become a global priority [6]. Biofuels have emerged as a promising solution to mitigate the environmental impact of conventional fuels and reduce the carbon footprint by providing a renewable energy source with lower environmental impact [5,7].

Within the biofuel portfolio, RD—also known as green diesel or hydrotreated vegetable oil (HVO)—has attracted considerable attention [8,9,10]. Unlike traditional biodiesel (FAME), which can suffer from low oxidative stability and poor low-temperature performance [2,11], RD exhibits physicochemical properties very similar to fossil diesel, making it a “drop-in” fuel compatible with existing transport infrastructure and engines without major modifications [12,13]. In quantitative terms, HVO typically exhibits a cetane number ≥ 70, a density on the order of 770–790 kg ${\mathrm{m}}^{-3}$, and a lower heating value of ∼42–44 MJ ${\mathrm{kg}}^{-1}$; moreover, after hydroisomerization, it shows improvements in low-temperature behavior, which is consistent with specifications for paraffinic fuels (EN 15940 [14]/ASTM D975 [15]) and with experimental data relative to commercial diesel. For example, reported values for HVO include a density of 779.3 kg ${\mathrm{m}}^{-3}$, a cetane number of 72.7, and a CFPP of −31 °C, compared with 826.6 kg ${\mathrm{m}}^{-3}$, 59.9, and −23 °C, respectively, for fossil diesel. It is also documented that soot oxidation with HVO is very similar to that with fossil diesel, so oxidation after-treatment/DPF devices designed for conventional diesel also operate properly with HVO [1,16,17,18]. Furthermore, when HVO is compared with biodiesel, improvements are also reported in aspects such as oxidative stability and emissions of NOx, particulate matter, and volatile organic compounds [19,20].

RD is produced mainly via the hydrotreating (also called hydroprocessing/hydrode-oxygenation, HDO) of lipids such as vegetable oils, animal fats, algal oils, and used cooking oils (UCOs) [21,22]. This process entails the catalytic conversion of triglycerides and fatty acids in the presence of hydrogen, removing oxygen atoms and saturating unsaturated bonds to form linear and branched paraffinic hydrocarbons (C15–C18) [23]. The principal reactions are hydrodeoxygenation (oxygen removal as water), decarboxylation (${\mathrm{CO}}_2$ removal), and decarbonylation (CO and water removal) [24]. Other feasible routes include hydrothermal liquefaction (HTL), Fischer–Tropsch (FT) synthesis from biomass gasification, and biological upgrading [3].

Numerous researchers and companies have advanced RD production via hydrotreating. Firms such as UOP/Eni, with Ecofining™, and Neste Oil, with NExBTL, pioneered commercialization [25]. Academically, diverse feedstocks and process conditions have been explored. For example, Douvartzides et al. (2019) [26] and Chia et al. (2022) [1] reviewed biomass sources and production technologies. Studies have addressed the hydrotreating of palm oil [9,27], jatropha oil [2,22], rapeseed oil [28], soybean oil [21], UCO [7], and animal fats [22], among others [1,25]. Process optimization often uses simulation tools such as Aspen Plus to determine mass and energy balances at industrial scale [9]. Tirado and Ancheyta (2022) [25] modeled multi-bed industrial reactors to analyze complex phenomena, including heat/mass transfer and hydrogen consumption to optimize catalyst bed configurations and maximize green diesel yields. Such kinetic models are essential to describe reactions, hydrogen demand, and product yields.

Complementary to the HVO/RD hydrotreating approach, recent studies have advanced along the FAME route (biodiesel via transesterification) and in engine performance optimization using experimental designs and metaheuristics. For example, Ajith et al. [29] combined microwave-assisted transesterification with a CCD design coupled to GA/PSO/MOA/SBOA algorithms, achieving experimental yields close to 99% and demonstrating the computational advantages of SBOA over alternatives; in turn, Chao-zhe et al. [30] hybridized RSM with the Rao-3 algorithm to simultaneously optimize performance and emissions of a compression-ignition (CI) engine fueled with ZnO-doped biodiesel blends, attaining high desirability (∼0.95) and NOx levels on the order of 700 ppm. These advances confirm the usefulness of optimization frameworks and experimental validation in biofuels; however, they focus on the FAME/engine-use pathway rather than on hydrotreating (HVO/RD) or the explicit integration of green hydrogen. Our study is positioned precisely in that gap: it empirically validates the supply of ${\mathrm{H}}_2$ in a solar–wind microplant with AEM electrolysis and couples it to RD production via stoichiometric balancing in a territorial context.

A critical factor in RD’s environmental and economic performance is hydrogen production, since conventional steam methane reforming (SMR) from natural gas carries a substantial carbon footprint [31]. To minimize RD’s carbon intensity, green hydrogen (${\mathrm{H}}_2$), produced from renewable electricity (solar/wind) via water electrolysis, is the most sustainable alternative [12,21]. Life cycle assessments (LCAs) show that producing HVO using ${\mathrm{H}}_2$ and UCO with photovoltaic/wind electricity can significantly reduce environmental impacts relative to conventional scenarios [2]. Concas et al. (2022) [32] and Wang et al. (2021) [33] highlighted environmental gains when electrolysis is powered by wind and photovoltaics. Sadeghi et al. (2022) [34] reported a lower global warming potential (GWP) when solar energy is used for hydrogen production instead of SMR. Lorenzi et al. (2020) [35] investigated HVO from UCO using high-temperature electrolysis as the hydrogen source. Pressurized alkaline electrolysis (ALE-P), owing to its technological maturity, is considered a viable option for ${\mathrm{H}}_2$ production [2]. Although electrified SMR (E-SMR) can cut GHG emissions by using electricity instead of natural gas combustion [21], direct electrolysis with renewables offers the greatest footprint reduction [2].

The recent literature on hybrid microgrids for green ${\mathrm{H}}_2$ describes PV/WT/electrolyzer/storage integrations that are predominantly simulation- and optimization-based using various software tools, typically with PEM electrolyzers; these frameworks inform sizing and costs (LCOE/LCOH) but reveal a lack of site-specific experimental validation of ${\mathrm{H}}_2$ supply [36,37,38].

Within this context, the present work contributes three key innovations beyond the LCA/TEA and reactor modeling commonly reported: empirical validation of ${\mathrm{H}}_2$ production in an operational solar–wind hybrid microplant that supplies AEM-type electrolyzers at the University of La Guajira; the integration of the measured generation with an explicit stoichiometric balance of the hydrotreating process (HDO, decarboxylation, decarbonylation) to estimate conversion to RD from local palm oil; and sizing of the shares of renewable resources—palm oil, solar, and wind—needed to sustain the process, without performing an LCA/TEA, while providing a quantitative basis for sustainable scaling in the region. This approach contrasts with most hydrogen-based hybrid microgrid studies—typically simulation-driven and using PEM electrolyzers—by situating ${\mathrm{H}}_2$ validation within a specific territorial context.

Finally, the La Guajira case is relevant due to the co-location of high-potential solar and wind resources and the growing availability of palm oil in the rural areas of the department’s central–southern zone; at present, much of this oil is transported to other regions for biodiesel production, entailing logistical surcharges and reduced local margins. The on-site integration of renewable electricity (PV/WT), electrolysis, and conversion of palm oil to RD would shorten value chains and reduce transport emissions, enabling a territorially anchored production scheme.

Regarding the stoichiometric assumptions, verification is theoretical given the study’s scope; however, operational validation is supported by monthly ${\mathrm{H}}_2$ measurements from the microplant. Because Colombia is equatorial, with two seasons and less pronounced seasonal variability in radiation and wind, one semester of operation (December 2024–May 2025) captures relevant climatic variations to infer the operational sustainability of ${\mathrm{H}}_2$ supply; therefore, the chosen time window is deemed sufficient to draw conclusions about hydrogen production sustainability.

This study aims to experimentally validate green hydrogen production in a solar–wind hybrid microplant with AEM-type electrolyzers installed at the University of La Guajira, using operational data from December 2024 to May 2025, and to integrate the measured generation with a detailed stoichiometric balance of palm oil hydrotreating to estimate RD yields. Based on this integration, we quantify specific ${\mathrm{H}}_2$ and electricity demands, as well as the resource shares (palm oil, solar, and wind) required to sustain the process at micro-scale in a territorial context. This study is delimited to operational validation and stoichiometric balancing (without LCA/TEA or reactor kinetic modeling) and provides quantitative inputs for sizing and sustainable scaling in the region.

2. Methodology

The methodological approach of this study was structured into five stages. First, the equipment at the Power-to-Gas (PtG) pilot plant was surveyed, including the 15 kWe hybrid wind–solar photovoltaic microplant that supplies electrical power to the on-site electrolyzers for green hydrogen production. Second, output power and electrical energy data from the microplant were collected over a six-month period. These data were used to analyze the energy production patterns and assess the reliability of the hybrid system. Third, research-informed mathematical models for hydrogen production were experimentally validated, with the aim of computing monthly hydrogen output from the microplant’s power data. Fourth, the most representative triglyceride models constituting palm oil—relevant to the RD production process—were identified. Finally, a detailed stoichiometric balance for RD production was performed based on the stages involved in producing this biofuel.

2.1. Equipment Recognition

A Power-to-Gas (PtG) pilot plant has been installed at the University of La Guajira, which was acquired through a research project funded by the Ministry of Science, Technology, and Innovation (Minciencias), as shown in Figure 1. The plant was designed to harness surplus electrical energy generated by a hybrid solar photovoltaic–wind system, given the region’s high solar irradiance and wind potential. From these renewable sources, green hydrogen is produced and stored as an energy carrier for subsequent conversion to methane. This is the plant’s current operating philosophy; however, because hydrogen can be transformed into various value-added products, we propose simulating its conversion toward RD production.

Figure 1. Diagram of the PtG plant located at the University of La Guajira, Colombia.

The pilot plant consists of three main sections. First, electric power generation comes from two renewable sources: a 10 kWe photovoltaic system and a 5 kWe wind system. This hybrid configuration addresses the need to improve generation reliability due to the inherent intermittency of these technologies under variable meteorological conditions.

Second, the electrochemical stage involves producing hydrogen via two electrolyzers, each with a capacity of 0.5 Nm³/h and a power rating of 2.4 kWe. In this stage, renewable electricity from the first section, together with water purified by auxiliary systems, enables water splitting into two streams: hydrogen, which is compressed to 35 bar and stored, and oxygen, which is currently vented to the atmosphere, although characterization studies are underway for its potential utilization.

Third, the thermochemical stage corresponds to methane production and hydrogen combustion, with the latter being used demonstratively to observe the characteristic flame of this gas.

The initial data collection system for the wind turbine was manual, which led to data losses and discontinuities (especially during nighttime periods and days without access). To overcome this limitation, the SAD.1—an automatic acquisition system based on RS-485/Modbus-RTU (MAX485 transceiver + Arduino microcontroller)—was developed, which continuously reads the wind inverter’s electrical variables (voltage, current, frequency, power, and energy). Before the start of the measurement campaign (December 2024–May 2025), an initial calibration was performed by comparing SAD.1 readings with the inverter’s front panel and with a reference multimeter. To minimize sampling errors and aliasing, the system was set to operate at 1 Hz and compute 5 min means from 300 one-second samples, which are stored in a 24 h circular buffer. Time synchronization was ensured using a real-time clock with periodic verification against an NTP source, aligning timestamps to 5 min boundaries and facilitating fusion with the photovoltaic log (APSystems ECU-C) into an integrated dataset.

The resulting set of calibrated and synchronized records enables a reliable assessment of the hybrid wind–solar system’s performance, its operating range, and its ability to meet the energy demand of the PtG pilot plant, quantifying both the grid contribution during deficit periods and the surplus energy injected into the power grid.

Additional details on the PtG pilot plant and its configuration can be found in our previous works [39,40,41].

2.2. Experimental Validation of Green Hydrogen Production

Since hydrogen is one of the fundamental inputs in RD production, it is essential to have mathematical models that can reliably predict its generation. At the University of La Guajira, experimental electrolyzers are available from which data have been collected on hydrogen production, as well as relevant electrical variables such as voltage and current. These experimental records provide a basis for benchmarking and validating different modeling approaches.

For this purpose, a review of the specialized literature was conducted to identify mathematical models applicable to predicting hydrogen production in electrochemical systems powered by renewable energy sources. Although the three models identified use the same equation to calculate the hydrogen production rate $\left({\dot{n}}_{H_2}\right)$ (see Equation (1)), they differ in how they determine the Faradaic efficiency (${\eta }_F$), as shown in Equation (2) (Model 1 by Castañeda et al. [42]), Equation (3) (Model 2 by Rangel et al. [43]), and Equation (4) (Model 3 by Li et al. [44]):

$${\dot{n}}_{H_2}={\displaystyle \frac{N_{elec}\ast P_{elec}}{2F\ast V_{elec}\ast N_{elec}}}\ast {\eta }_F\ast 3600={\displaystyle \frac{{\eta }_F\ast P_{elec}}{2F\ast V_{elec}}}\ast 3600$$

(1)

$${\eta }_F=96.5e^{\left({\displaystyle \frac{0.09}{I_{elec}}}-{\displaystyle \frac{75.5}{I_{elec}^2}}\right)}$$

(2)

$${\eta }_F={\displaystyle \frac{i-i_{Loss}}{i}}$$

(3)

$${\eta }_F=1$$

(4)

where $N_{elec}$ is the number of cells in series; $P_{elec}$ is the electrolyzer power; F (°C ${\mathrm{mol}}^{-1}$) is the Faraday constant; $V_{elec}$ is the electrolyzer voltage; ${\eta }_F$ is the Faradaic efficiency; $I_{elec}$ (A) is the electrolyzer current; and i and $i_{Loss}$ are the internal current and the hydrogen-loss current, respectively, with the latter arising from oxygen crossover from the anode to the cathode or hydrogen crossover from the cathode to the anode.

All models are grounded in the same principle—Faraday’s law—which links hydrogen production to the current passing through the electrolyzer; operationally, the measured electrical power is converted to current using an effective cell voltage. In Model 1, Faraday’s law is retained and the Faradaic efficiency is modeled explicitly as a function of current, following empirical/experimental correlations reported for electrolyzers that capture losses due to physicochemical phenomena; under standard operating assumptions (quasi-constant temperature), the electrolyzer can be represented by an equivalent electrical circuit with first-order dynamics for the production response. In Model 2, the Faraday structure is preserved, but the Faradaic efficiency is condensed into a compact parametrization (monotonic and bounded), consistent with phenomenological fitting of highly nonlinear current–voltage characteristics and with decomposing the electrolyzer’s electrical efficiency into current efficiency × voltage efficiency; this choice is physically consistent and mathematically identifiable for system-level explorations. In Model 3, the same functional form as in Model 1 is used, but the Faradaic efficiency is fixed at 100% (unity), a common simplification in analysis and sizing when working with time averages; under that assumption, and taking a nearly constant effective per-cell voltage, Faraday’s law reduces to a direct relationship of power → current → hydrogen, which even allows for the calculation of the production of an electrolyzer with a given nominal power [42,43,44].

To validate the experimental results against the mathematical models, three key relationships were assessed: ${\mathrm{H}}_2$ production vs. electrical power (see Figure 2), ${\mathrm{H}}_2$ production vs. efficiency (see Figure 3), and electrical power vs. efficiency (see Figure 4).

Figure 2. Validation of ${\mathrm{H}}_2$ production versus electrical power.

Figure 3. Validation of ${\mathrm{H}}_2$ production versus electrolyzer efficiency.

Figure 4. Validation of electrical power versus electrolyzer efficiency.

Model performance was analyzed using goodness-of-fit statistics—coefficient of determination (${\mathrm{R}}^2$) and mean squared error (MSE). These results are shown in Table 1, where it can be observed that Model 2 provides the best fit to the experimental data (${\mathrm{R}}^2$ = 0.9848; MSE = 130.0476), exhibiting higher overall accuracy in predicting ${\mathrm{H}}^2$ production as a function of electrical power. This performance positions it as the most reliable model to represent the real operating conditions of the pilot plant under renewable energy supply.

Table 1. Adjustment error.

Error Metric	Model 1	Model 2	Model 3
Mean Squared Error (MSE)	835.9485	130.0476	187.3681
R2; (Coefficient of Determination)	0.9023	0.9848	0.9781

It is important to note that extensive validation details are not provided herein because the results are currently in a paper that is yet to be published.

2.3. General Model of Palm Oil Triglycerides

The general triglyceride model of palm oil is based on representative species such as POP, POO, POL, PLP, and (as a limiting case) SOS, whose acyl-chain combinations (C16:0, C18:1, C18:2, C18:0) determine the empirical formula, molar mass, and degree of unsaturation; these parameters govern the hydrogen requirement and the hydroprocessing pathways—hydrogenation, hydrodeoxygenation (HDO), decarboxylation (DCX), and decarbonylation (DCN)—thereby conditioning selectivity toward C15–C18 n-paraffins (RD), the formation of propane and ${\mathrm{CO}}_x/{\mathrm{H}}_2\mathrm{O}$, and the carbon–hydrogen efficiency of the process. Table 2 reports each TAG’s fatty acid composition, empirical formula, and molar mass, together with its typical abundance in palm oil, which enables the mapping of the feed’s effective H/C and closing stoichiometric balances to be consistent with the monthly availability of ${\mathrm{H}}_2$ [45,46].

Table 2. Most representative models of triglycerides in palm oil [45,46].

Triglyceride (TAG)	Composition (Fatty Acids)	General Formula	Molar Mass (g/mol)	Comment
POP (Palmitoyl– Oleoyl–Palmitoyl)	C16:0–C18:1–C16:0	${\mathrm{C}}_{55}{\mathrm{H}}_{104}{\mathrm{O}}_6$	807–808	Most representative model for crude palm oil (predominates due to high palmitate and oleate content).
POO (Palmitoyl– Oleoyl–Oleoyl)	C16:0–C18:1–C18:1	${\mathrm{C}}_{57}{\mathrm{H}}_{106}{\mathrm{O}}_6$	831–832	Highly abundant in the liquid fraction (palm olein).
POL (Palmitoyl– Oleoyl–Linoleoyl)	C16:0–C18:1–C18:2	${\mathrm{C}}_{57}{\mathrm{H}}_{104}{\mathrm{O}}_6$	833–834	Present at lower levels; contributes unsaturations (fluidity).
PLP (Palmitoyl– Linoleoyl–Palmitoyl)	C16:0–C18:2–C16:0	${\mathrm{C}}_{55}{\mathrm{H}}_{102}{\mathrm{O}}_6$	809–810	Provides a balance between saturation and unsaturation.
SOS (Stearoyl– Oleoyl–Stearoyl)	C18:0–C18:1–C18:0	${\mathrm{C}}_{59}{\mathrm{H}}_{110}{\mathrm{O}}_6$	891–892	More typical of cocoa butter fats, but occurs in some palm oil fractions.

Although palm oil is a mixture of co-reacting triacylglycerols (TAGs), total hydrogen consumption can be obtained as a weighted sum of the contribution of each species. Since differences between TAGs—mainly due to the degree of unsaturation and deoxygenation pathway—are limited in palm-type feedstocks, calculating for representative species separately and then combining them in a weighted manner is a conservative approximation that can be refined in the future using the analytical TAG distribution and a sensitivity analysis across deoxygenation routes.

2.4. Stoichiometric Balance of ${\mathrm{H}}_2$ Requirements in DR Production

The stoichiometric balance of the triglyceride hydrotreating process enables the quantification of the net hydrogen demand and prediction of the distribution of main products and coproducts, irrespective of kinetics or reactor model. To this end, we begin with a generic triglyceride whose acyl chains may contain unsaturations and consider a sequence of three conceptual stages: (i) saturation of the triglyceride, (ii) triglyceride cleavage (hydrogenolysis of the ester bonds), and (iii) formation of n-paraffins via the deoxygenation of the fatty acids. The equations presented below represent stage-wise overall balances and provide the basis for elemental closure of C, H, and O and, therefore, for estimating the process ${\mathrm{H}}_2$ requirements [47,48].

The balance presented here constitutes a first step to defining the initial conditions and net ${\mathrm{H}}_2$ requirements from elemental C/H/O balances; by design, it does not incorporate reaction rates, catalyst effects, or operating conditions. Consistent with the focus of this work—validating the sustainable supply of low-emission ${\mathrm{H}}_2$ and integrating that supply into the hydrotreating scheme—the kinetic and selectivity assessment (HDO/DCN/DCX) is proposed as follow-on work. In particular, once the periods of the highest availability of green ${\mathrm{H}}_2$ have been established and characterized, kinetic models and reactor simulations (together with the corresponding experimental campaign) can be developed to estimate yields, specific consumptions, and product distributions under different catalytic and operating configurations.

2.4.1. Triglyceride Saturation

In this first stage, the double bonds present in the acyl chains are saturated, stabilizing the substrate and preventing parallel reactions associated with unsaturations. In overall terms, this step can be represented as follows [47]:

$${\mathrm{C}}_3{\mathrm{H}}_5{\left(\mathrm{COO}\right)}_3{\mathrm{C}}_{3\mathrm{n}}{\mathrm{H}}_{6\mathrm{n}-3} + 3{\mathrm{H}}_2\to {\mathrm{C}}_3{\mathrm{H}}_5{\left(\mathrm{COO}\right)}_3{\mathrm{C}}_{3\mathrm{n}}{\mathrm{H}}_{6\mathrm{n}+3}$$

(5)

2.4.2. Triglyceride Cleavage (Hydrogenolysis)

Once the chains are saturated, the three ester bonds that link glycerol to the acyl chains are cleaved. This step releases propane as an invariant coproduct and generates three saturated fatty acids that feed the deoxygenation stage. The overall reaction can be represented as follows [47]:

$${\mathrm{C}}_3{\mathrm{H}}_5{\left(\mathrm{COO}\right)}_3{\mathrm{C}}_{3\mathrm{n}}{\mathrm{H}}_{6\mathrm{n}+3} + 3{\mathrm{H}}_2\to {\mathrm{C}}_3{\mathrm{H}}_8+3{\mathrm{C}}_{\mathrm{n}}{\mathrm{H}}_{2\mathrm{n}+1}\mathrm{COOH}$$

(6)

2.4.3. Formation of n-Paraffins (Deoxygenation of Fatty Acids)

Each saturated fatty acid is converted into an n-paraffin via one of three overall routes. These routes differ in carbon retention and ${\mathrm{H}}_2$ demand, and they define the oxygenated coproducts [47].

Decarboxylation (DCX)

This route involves removing the carboxyl group as ${\mathrm{CO}}_2$, yielding an n-paraffin with one fewer carbon in the chain. It reduces the hydrogen requirement, although it entails carbon loss [47]:

$${\mathrm{C}}_{\mathrm{n}}{\mathrm{H}}_{2\mathrm{n}+1}\mathrm{COOH}\to {\mathrm{C}}_{\mathrm{n}}{\mathrm{H}}_{2\mathrm{n}+2} + {\mathrm{CO}}_2$$

(7)

Decarbonylation (DCN)

This route generates CO and water, along with the loss of one carbon from the hydrocarbon chain. It requires less hydrogen than HDO and offers a compromise between ${\mathrm{H}}_2$ consumption and carbon retention [47]:

$${\mathrm{C}}_{\mathrm{n}}{\mathrm{H}}_{2\mathrm{n}+1}\mathrm{COOH}+{\mathrm{H}}_2\to {\mathrm{C}}_{\mathrm{n}}{\mathrm{H}}_{2\mathrm{n}+2} + {\mathrm{H}}_2\mathrm{O}+\mathrm{CO}$$

(8)

Hydrodeoxygenation (HDO)

This route removes oxygen exclusively as water and preserves the total chain length (n + 1 carbons when counting the carboxyl carbon), favoring drop-in fuel properties. It is the most hydrogen-intensive route [47]:

$${\mathrm{C}}_{\mathrm{n}}{\mathrm{H}}_{2\mathrm{n}+1}\mathrm{COOH}+3{\mathrm{H}}_2\to {\mathrm{C}}_{\mathrm{n}+1}{\mathrm{H}}_{2\mathrm{n}+4} + 2{\mathrm{H}}_2\mathrm{O}$$

(9)

It should be noted that the notation n denotes the number of carbon atoms in the fatty acid’s hydrocarbon chain.

3. Results and Discussion

In this section, we discuss the results obtained over six months (December 2024–May 2025), beginning with electric generation from the wind–photovoltaic microplant, continuing with green hydrogen production based on the model validated in Section 2.3, and, finally, estimating RD via a detailed stoichiometric balance using the ${\mathrm{H}}_2$ data generated at the University of La Guajira’s PtG plant.

The compiled data enable a comparison of the standalone performance of the photovoltaic and wind systems and, subsequently, of the hybrid system. Figure 5 shows that the photovoltaic contribution dominates total generation during the evaluation period, explained by the larger nameplate capacity of the PV field relative to wind (10 kWe vs. 5 kWe) and by the lower intraday variability in irradiance compared with wind speed. Nonetheless, both resources are complementary: the wind component partially buffers PV production valleys, reinforcing the hybrid operating philosophy in a setting with high solar radiation and favorable winds, as reported in the literature [49,50]. The month-by-month behavior shown in Figure 5 corroborates that microplant sizing conditions not only the monthly energy available but also the stability with which that energy can be supplied to the electrolyzers.

Figure 5. Electricity generation of the hybrid wind–solar photovoltaic microplant during the study months: (a) December 2024, (b) January 2025, (c) February 2025, (d) March 2025, (e) April 2025, and (f) May 2025.

Building on the validated ${\mathrm{H}}_2$ model, Figure 6 illustrates the direct role of wind speed and solar irradiance resources in green hydrogen production. Analyzing the systems separately, the intermittency of each resource prevents sustained ${\mathrm{H}}_2$ output because the electrolyzers must operate above a minimum power threshold (2.88 kWe for the pair)—a level not stably reached with a single resource over the day. Operating below this threshold would require grid support and forfeit the ${\mathrm{H}}_2$ “green” attribute. By contrast, when operating as a hybrid system, Figure 6 shows that the fraction of useful time above the threshold increases and, with it, the effective ${\mathrm{H}}_2$ production; in practical terms, “the more net renewable energy, the more ${\mathrm{H}}_2$,” and hence, the greater RD potential in the subsequent stoichiometric stage.

Figure 6. Green hydrogen production as a function of wind speed and solar radiation during the study months: (a) December 2024—wind speed; (b) December 2024—solar radiation; (c) January 2025—wind speed; (d) January 2025—solar radiation; (e) February 2025—wind speed; (f) February 2025—solar radiation; (g) March 2025—wind speed; (h) March 2025—solar radiation; (i) April 2025—wind speed; (j) April 2025—solar radiation; (k) May 2025—wind speed; (l) May 2025—solar radiation.

With the availability of green hydrogen, its use is considered with vegetable oils—particularly palm oil available in Lower Guajira—and used cooking oils in a hydrotreating process to produce a drop-in fuel (RD) compatible with conventional diesel engines. The process occurs in three steps: triglyceride scission (release of fatty acids and propane); hydrogenation of double bonds; and oxygen removal via DCX, DCN, or HDO. In an actual plant, the proportions among these routes define the product distribution and ${\mathrm{H}}_2$ consumption; strategically, DCX minimizes ${\mathrm{H}}_2$ use (but sacrifices carbon as ${\mathrm{CO}}_2$), HDO conserves carbon at the expense of higher specific hydrogen demand, and DCN lies in between (CO + ${\mathrm{H}}_2\mathrm{O}$).

To determine palm oil compositional diversity, representative TAGs (POP, POO, POL, PLP, and SOS) were considered, and a month-by-month stoichiometric balance was performed using the modeled monthly green ${\mathrm{H}}_2$. Table 3, Table 4, Table 5, Table 6, Table 7 and Table 8 report, for each TAG and month, the oil mass required, estimated RD, coproducts (propane from triglyceride scission), and total ${\mathrm{H}}_2$ required by the deoxygenation route (DCX/DCN/HDO). The cross-cutting trend aligns with known kinetics: months with higher ${\mathrm{H}}_2$ availability—e.g., February and March—concentrate the largest calculated RD masses across all TAGs, whereas months with lower hybrid energy—e.g., May—show proportional reductions. In compositional terms, TAGs with higher effective molecular weight, such as SOS, generally yield larger absolute RD masses under the same conditions than relatively lighter TAGs (POP/POO/POL/PLP), holding the assumed stoichiometric conditions by route constant. Additionally, the calculated propane mass remains in the expected order for one triglyceride scission per molecule and represents a potential ${\mathrm{C}}_3$ valorization stream if integrated.

Table 3. Stoichiometric balance for RD production in the month of December 2024.

Triglyceride (TAG)	Palm Oil Mass (g)	RD Mass (g)			Propane Mass (g)
		DCX	DCN	HDO
POP	16,452.96	13,130.73	13,130.73	13,961.29	870.11
POO	15,181.10	12,243.97	12,243.97	12,987.10	778.52
POL	13,703.27	11,077.87	11,077.87	11,750.23	704.37
PLP	14,685.68	11,748.55	11,748.55	12,491.68	778.52
SOS	17,560.37	14,238.14	14,238.14	15,068.70	870.11

Table 4. Stoichiometric balance for RD production for the month of January 2025.

Triglyceride (TAG)	Palm Oil Mass (g)	RD Mass (g)			Propane Mass (g)
		DCX	DCN	HDO
POP	21,713.52	17,329.06	17,329.06	18,425.17	1148.31
POO	20,035.01	16,158.77	16,158.77	17,139.51	1027.44
POL	18,084.66	14,619.84	14,619.84	15,507.17	929.59
PLP	19,381.18	15,504.95	15,504.95	16,485.68	1027.44
SOS	23,175.01	18,790.55	18,790.55	19,886.66	1148.31

Table 5. Stoichiometric balance for RD production for the month of February 2025.

Triglyceride (TAG)	Palm Oil Mass (g)	RD Mass (g)			Propane Mass (g)
		DCX	DCN	HDO
POP	28,183.05	22,492.24	22,492.24	23,914.94	1490.45
POO	26,004.42	20,973.26	20,973.26	22,246.21	1333.56
POL	23,472.97	18,975.81	18,975.81	20,127.52	1206.55
PLP	25,155.79	20,124.63	20,124.63	21,397.58	1333.56
SOS	30,079.98	24,389.17	24,389.17	25,811.88	1490.45

Table 6. Stoichiometric balance for RD production for the month of March 2025.

Triglyceride (TAG)	Palm Oil Mass (g)	RD Mass (g)			Propane Mass (g)
		DCX	DCN	HDO
POP	31,035.52	24,768.73	24,768.73	26,335.43	1641.30
POO	28,636.39	23,096.02	23,096.02	24,497.80	1468.53
POL	25,848.72	20,896.40	20,896.40	22,164.67	1328.67
PLP	27,701.87	22,161.49	22,161.49	23,563.28	1468.53
SOS	33,124.45	26,857.66	26,857.66	28,424.36	1641.30

Table 7. Stoichiometric balance for RD production for the month of April 2025.

Triglyceride (TAG)	Palm Oil Mass (g)	RD Mass (g)			Propane Mass (g)
		DCX	DCN	HDO
POP	16,836.69	13,436.98	13,436.98	14,286.91	890.40
POO	15,535.17	12,529.53	12,529.53	13,290.00	796.68
POL	14,022.87	11,336.24	11,336.24	12,024.28	720.80
PLP	15,028.20	12,022.56	12,022.56	12,783.02	796.68
SOS	17,969.93	14,570.22	14,570.22	15,420.14	890.40

Table 8. Stoichiometric balance for RD production for the month of May 2025.

Triglyceride (TAG)	Palm Oil Mass (g)	RD Mass (g)			Propane Mass (g)
		DCX	DCN	HDO
POP	10,657.48	8505.49	8505.49	9043.49	563.62
POO	9833.63	7931.09	7931.09	8412.45	504.29
POL	8876.35	7175.74	7175.74	7611.27	456.26
PLP	9512.72	7610.18	7610.18	8091.54	504.29
SOS	11,374.81	9222.82	9222.82	9760.82	563.62

Finally, the six-month overall balance (see Table 9) itemizes ${\mathrm{H}}_2$ consumption by stage and explicitly compares the impact of the selected route on process hydrogen intensity. One limitation of Table 8 is that the exact contents of TAG types cannot be established; therefore, we assume that all oils required by the available ${\mathrm{H}}_2$ correspond to a single TAG type for the analysis. Accordingly, individual ${\mathrm{H}}_2$-demand balances are performed for each TAG.

Table 9. Overall stoichiometric balance for RD production over the six-month study period.

Triglyceride (TAG)	${\mathrm{H}}_2$ Mass (kg)					Total ${\mathrm{H}}_2$ Mass (kg)	Palm Oil Mass (kg)	RD Mass (kg)			Propane Mass (kg)	${\mathrm{CO}}_2$ Mass (kg)	CO Mass (kg)	Water Mass (kg)
	Saturation	Fatty Acid Decomposition	DCX	DCN	HDO			DCX	DCN	HDO		DCX	DCN	DCN	HDO
POP	0.3	0.9	0.00	0.9	1.351	2.552	124.879	99.663	99.663	105.967	6.604	19.813	12.608	8.105	16.210
POO	0.537	0.806	0.00	0.806	1.209	2.552	115.226	92.933	92.933	98.573	5.909	17.727	11.281	7.252	14.504
POL	0.729	0.729	0.00	0.729	1.094	2.552	104.009	84.082	84.082	89.185	5.346	16.039	10.206	6.561	13.123
PLP	0.537	0.806	0.00	0.806	1.209	2.552	111.465	89.172	89.172	94.813	5.909	17.727	11.281	7.252	14.504
SOS	0.3	0.9	0.00	0.9	1.351	2.552	133.284	108.068	108.068	114.372	6.604	19.813	12.608	8.105	16.210

Choosing POP as a representative example, the ${\mathrm{H}}_2$ contributions to saturation and fatty acid scission for 124.89 kg of oil amount to 0.30 kg and 1.20 kg, respectively; DCX requires no additional ${\mathrm{H}}_2$, whereas DCN and HDO add 2.10 kg and 2.55 kg, respectively, assuming that the 124.89 kg of oil corresponds to POP. This establishes the ${\mathrm{H}}_2$-demand order by reaction as HDO > DCN > DCX. The operational takeaway is direct: under ${\mathrm{H}}_2$ limitations, DCX/DCN is favored, increasing the RD mass per kg of ${\mathrm{H}}_2$ but with a carbon penalty (${\mathrm{CO}}_2$ in DCX); with abundant ${\mathrm{H}}_2$, HDO maximizes carbon conservation and the strictest drop-in compatibility. Table 8 also quantifies the coproducts and by-products (propane, ${\mathrm{CO}}_2$, CO, and ${\mathrm{H}}_2\mathrm{O}$), which are useful for mass closure and environmental assessment.

Moreover, ${\mathrm{H}}_2$ requirements can be related to the number of double bonds and total carbon count of the oil’s TAG types; hence, each oil will have particular hydrogen needs for RD production.

Table 10 normalizes the hydrogen requirements and RD production to one metric ton of oil, enabling comparison across TAGs and establishing the degree of unsaturation regarding ${\mathrm{H}}_2$ demand. Under this consideration, SOS is the least unsaturated TAG and therefore requires less ${\mathrm{H}}_2$ to saturate its unsaturations. Although POP and SOS each contain one unsaturation, the greater carbon number of SOS results in lower ${\mathrm{H}}_2$ requirements per ton of oil, thereby giving the appearance of being less unsaturated than POP.

Table 10. Required amounts of ${\mathrm{H}}_2$ and DR production for one ton of palm oil.

Triglyceride (TAG)	${\mathrm{H}}_2$ Mass (kg)				Total ${\mathrm{H}}_2$ Mass (kg)	RD Production (Gallons)
	Saturation	Fatty Acid Decomposition	DCN	HDO		DCX	DCN	HDO	Propane Mass (kg)
POP	2.40	9.62	16.83	20.43	32.45	247.87	247.87	263.55	52.88
POO	4.66	11.66	18.65	22.14	38.46	250.50	250.50	265.70	51.28
POL	7.01	14.02	21.03	24.53	45.56	251.08	251.08	266.32	51.40
PLP	4.82	12.05	19.28	22.89	39.76	248.47	248.47	264.19	53.01
SOS	2.25	9.01	15.77	19.14	30.41	251.83	251.83	266.52	49.55

This study is confined to a six-month monitoring window (December 2024 to May 2025) in La Guajira; given the weaker seasonality of solar irradiance and wind in equatorial settings, this time horizon is informative for the operational validation of green hydrogen, although a full-year window would allow a more robust quantification of interseasonal variability. RD yields were estimated via a stoichiometric balance (elemental closures of carbon, hydrogen, and oxygen), assuming complete conversion and pathway selectivity (hydrodeoxygenation, decarbonylation, and decarboxylation) and using representative triglycerides of palm oil. The analysis did not incorporate kinetics, catalyst effects, or heat and mass transport at the reactor scale, nor techno-economic or life-cycle assessments. The following aspects are proposed as future work: (i) full annual validation; (ii) modeling and kinetic testing under different catalyst families and operating conditions (temperature, pressure, and space velocity); (iii) reactor simulation with energy balances and pressure drop; and (iv) techno-economic and life-cycle analyses coupled to the local supply of green ${\mathrm{H}}_2$.

4. Conclusions

Under real operating conditions, this study validated the viability of coupling a hybrid wind–solar PV microplant with two electrolyzers to supply green hydrogen for RD production via hydrotreating. The hybrid configuration increased the fraction of time above the 2.88 kWe joint operating threshold, thereby enhancing the effective ${\mathrm{H}}_2$ output from December 2024 to May 2025. The selected ${\mathrm{H}}_2$ prediction model showed the best performance, enabling the closure of monthly stoichiometric balances for representative TAGs of palm oil (POP, POO, POL, PLP, SOS) and quantification of RD yields and coproducts (propane) under DCX/DCN/HDO pathways. The overall balance confirms the gradient in specific hydrogen consumption (HDO > DCN > DCX) and reveals the carbon–hydrogen trade-off; for example, for POP, the requirement is 0.30 kg ${\mathrm{H}}_2$ for saturation and 1.20 kg for cleavage, with an additional consumption of 2.10 kg (DCN) or 2.55 kg (HDO). These results guide operational decisions based on local renewable availability: when ${\mathrm{H}}_2$ is limiting, DCX/DCN maximizes liters of RD per kilogram of ${\mathrm{H}}_2$ but incurs a ${\mathrm{CO}}_2$ penalty; when ${\mathrm{H}}_2$ is abundant, HDO preserves carbon and ensures strict drop-in compatibility. Collectively, the findings provide experimental and modeling evidence that integrating green hydrogen with RD is technically feasible and scalable in La Guajira; moreover, the approach is transferable to other regions where renewable (or otherwise low-carbon) electricity, a sustainable supply of oils/fats, and access to hydrotreating capacity are available, provided that system sizing and the choice of reaction pathway (DCX/DCN vs. HDO) are tailored to the local energy profile and feedstock.