Technical and Economic Analysis of Rural Hydrogen–Electricity Microgrids

Abstract

China’s rural areas possess abundant renewable energy resources, but lack sufficient energy storage facilities. Hydrogen energy storage has been considered a potential green solution. This study, for the first time, constructed a planning model for a rural electric–hydrogen microgrid incorporating hydrogen and electricity storage, and conducted comprehensive technical and economic analysis under different time periods and combinations of technological elements. The levelized cost of electricity (LCOE) was employed as a key indicator, K-means clustering was employed to obtain typical source–load curves, and the curtailment/self-balancing rate was combined for evaluation. Off-grid energy storage schemes, grid-connected/off-grid modes, and hydrogen production methods were compared to determine the optimal solution. The simulation results show the following: in 2025, off-grid mode with alkaline water electrolyzer (AWE) hydrogen production, hydrogen–battery hybrid storage was the most cost effective (LCOE 0.2824 ¥/kWh) due to hydrogen sales profits and battery flexibility, while fuel cells were unfeasible. Grid-connected mode reduced LCOE by 0.008 ¥/kWh vs. off-grid. Currently, AWE’s LCOE is 0.0172 ¥/kWh lower than proton exchange membrane (PEM), but PEM may have a 0.0004 ¥/kWh lower LCOE by 2030, becoming preferred. The results are potential for cost effectiveness, aiding rural energy transition.

1. Introduction

1.1. Background and Motivation

In recent years, global carbon emissions have increased dramatically, necessitating a transition to carbon neutrality to avoid adverse impacts such as the greenhouse effect. In this global transition to carbon neutrality, rural energy systems have become a core area of sustainable development. Rural areas are rich in renewable energy resources, including photovoltaics, wind power, and biomass, offering significant potential for renewable energy development and playing a crucial role in sustainable development [1,2,3]. Rural rooftops, including those of houses and agricultural greenhouses, also offer ample space for photovoltaic installation, allowing for the large-scale deployment of photovoltaic systems [4]. However, China’s rural energy sector still faces numerous pressing challenges: low renewable energy penetration, high curtailment rates due to intermittent renewable energy output, and significant mismatches between energy supply and demand [2]. For example, during the peak agricultural season, irrigation loads on rural microgrids surge, while fluctuations in solar and wind power output often lead to insufficient energy supply or waste of excess electricity [5]. Additionally, off-grid rural communities—such as those in Southwest China or Northeast China’s power-starved regions—remain reliant on diesel generators, which incur high operational costs (e.g., 12.95 ¥/kWh for diesel-hydro systems) and significant carbon emissions [3,6]. These challenges highlight an urgent need for scalable, cost-effective energy storage solutions to unlock the potential of rural RE resources.

Hydrogen storage has emerged as a transformative green energy storage technology, offering unique advantages for rural microgrid applications. Unlike batteries, which are limited by short-duration storage (typically hours to days), hydrogen enables long-term, seasonal energy storage, critical for addressing the seasonal variability of rural RE (e.g., low solar output in winter, high irrigation demand in summer) [7,8]. Moreover, hydrogen can integrate with diverse rural energy systems: excess RE can be converted to green hydrogen via ELZ, stored in tanks, and later used for power generation (via fuel cells), heating, or even agricultural applications (e.g., green ammonia production) [9,10]. For example, a study in Baja California, Mexico, demonstrated that a hydrogen-based “Power-to-Gas-to-Power (P2G2P)” system reduced CO₂ emissions by 27% compared to diesel–battery solutions, underscoring hydrogen’s role in decarbonization [11]. In rural China, where land availability is greater than in urban areas, hydrogen storage systems can be deployed without significant spatial constraints, further enhancing their suitability for rural microgrids [6,12]. This study aims to guide the selection and planning of technology for rural electric–hydrogen microgrid systems in China to improve the economic efficiency of microgrid operation.

1.2. Literature Review

Recent studies on rural microgrids integrated with hydrogen storage have made significant progress in optimizing energy storage configuration and operational stability. Zhang et al. [13] constructed an interval optimization model for wind-biomass-hydrogen integrated energy systems in rural Northeast China, reducing wind curtailment by optimizing storage charging/discharging schedules. Bouaouda et al. [14] optimized a standalone PV–wind–hydrogen system in rural Southwest Morocco using the Dandelion Optimizer (DO), demonstrating faster convergence than genetic algorithms (GA) or particle swarm optimization (PSO). Vimal et al. [15] proposed a standalone PV–hydrogen system for rural Indian households, optimizing storage capacity to meet basic loads but not analyzing the economic impact of battery vs. hydrogen storage replacement cycles. Khan et al. [16] optimized a PV-regenerative hydrogen fuel cell (RHFC) microgrid for rural electrification in Fiji’s Soa Village, focusing on capacity configuration to meet local loads. Younis et al. [10] developed a Reduced Fractional Gradient Descent (RFGD) algorithm for hydrogen-supercapacitor hybrid storage in rural Saudi Arabian PV microgrids, reducing hydrogen consumption by 15%.

In the study of economic analysis of a hydrogen energy storage system, Makhoukh et al. [17] simulated a hybrid system combining solar, wind, biomass, and hydrogen–battery storage for rural mountainous areas in eastern Morocco, achieving 100% RE self-sufficiency with an LCOE of ¥0.168/kWh. Elminshawy et al. [18] analyzed a PV–wind–battery–electrolyzer system across three Egyptian regions, calculating an LCOH of ¥15.4/kg in Ras Ghareb. Köprü et al. [9] analyzed a hybrid system for rural Turkey with Archimedes screw turbines (AST), PV, biogas generators, and hydrogen storage, calculating total net cost (¥1,125,271) and LCOE (¥1.400/kWh) but fixing the technology route and not considering AST efficiency improvements or future biogas upgrading cost declines. Hosseini Dehshiri et al. [19] designed a PV–hydrogen system for Iranian rural healthcare centers, reporting LCOE (¥1.96–2.38/kWh) but not simulating how PEM fuel cell cost declines would shorten payback periods or how grid extension costs might change over time. Seyhan et al. [20] developed a multi-criteria framework for biomass-to-hydrogen facility siting in rural Turkey, identifying optimal locations based on biomass availability but not analyzing how hydrogen storage costs will evolve to improve economic feasibility. Chakraborty and Bohre [21] proposed a green hydrogen-based hybrid renewable energy system (HRES) for rural India’s Siliguri region, optimizing sizing via HOMER Pro, but not considering lithium ion battery cost declines or future PV module efficiency gains.

1.3. Research Gaps

However, the above studies have two critical limitations:

Studies focusing on system structure and algorithm design lack a systematic evaluation of the overall life cycle economic indicators of the system, failing to fully reflect the long-term economic feasibility of rural hydrogen microgrids.

Studies focusing on economic analysis are all based on the discussion of a fixed route for the combination of hydrogen energy storage technology elements, and lack consideration of the development of different hydrogen energy storage technology elements at different stages (e.g., future cost changes of PEM electrolyzers), making it difficult to provide forward-looking technical selection references.

1.4. Contributions

Therefore, this study proposes a technical and economic analysis of rural hydrogen–electricity microgrids. The main contributions are as follows:

(1)

This is the first time that a comprehensive technical and economic analysis of rural electric–hydrogen microgrids under different time periods and technology combinations has been conducted, providing more targeted technology selection criteria for rural areas with varying resource conditions and grid connection capabilities.

(2)

A complete planning model for rural microgrids, including hydrogen energy storage and electrical energy storage, is established. Typical days are selected using the K-means algorithm, and the levelized cost of electricity (LCOE) is employed as the key indicator. The wind and solar curtailment rate and self-balancing rate are evaluation indicators to construct a techno-economic evaluation system for rural microgrids.

(3)

A comparative analysis of different off-grid energy storage, different main grid access methods, and different hydrogen production methods in different time periods is constructed, and the following conclusions were drawn: compared with the pure electric energy storage, the LCOE when using electric-hydrogen hybrid energy storage in an off-grid state is reduced to 0.2824 CNY/kWh, a reduction of 33.06%; compared with now, the cost reduction of PEM in 2030 will reach 49.3%, and the LCOE of AWE will be 0.0004 ¥/kWh lower. In 2030, PEM will become the most preferred method of hydrogen production in rural electric–hydrogen microgrids.

In summary, the techno-economic analysis of rural hydrogen microgrids proposed in this paper fills the gap in existing research by providing a systematic assessment framework of the economic indicators throughout the system’s entire life cycle considering the combination and development of different hydrogen energy technology components at different stages. The research will provide a reasonable evaluation guideline for the planning of rural microgrids.

1.5. Organization

This paper is organized as follows. Section 2 analyzes the technical path selection for rural electric–hydrogen microgrids. Section 3 introduces a capacity planning model for rural electric–hydrogen microgrids. Section 4 introduces the principles and process of the K-means algorithm. Section 5 provides a case study for comparative analysis and identifies the optimal technical path for rural electric–hydrogen microgrids. Finally, Section 6 summarizes the conclusions reached in this paper, outlines the limitations and shortcomings of this study, and provides an outlook on future research.

2. Path Analysis of Rural Electricity–Hydrogen Coupled Microgrids

2.1. Comparison of Hydrogen Energy Storage Technologies

Water electrolysis driven by renewable energy is one of the most promising ways to produce hydrogen. Currently, there are many types of ELZ for producing hydrogen by water electrolysis, including AWE, PEM, anion exchange membrane (AEM), and solid oxide electrolyzers (SOEC) [22].

Nowadays, there is no specific electrolyzer technology outperforming the rest across all the features essential for the effective production of renewable hydrogen.

AWE and PEM stand out among hydrogen production technologies due to their superior compatibility with renewable energy and environmental performance, while also differing substantially in technical, economic, and operational characteristics. Compared to conventional steam reforming, a dominant industrial hydrogen production method, AWE and PEM avoid the critical limitation of insufficient carbon reduction, steam reforming of natural gas fails to truly cut overall CO₂ emissions [23], whereas AWE and PEM, when powered by solar, wind, or hybrid renewable systems, enable near-zero-carbon “green hydrogen” production. Against SOEC, which offers high conversion efficiency, AWE and PEM operate at ambient to moderate temperatures, eliminating SOEC’s drawbacks of long startup times and high-temperature operational constraints. Compared to AEM electrolyzers—which use low-cost non-noble metal electrocatalysts but suffer from low performance and short membrane lifetimes—AWE and PEM exhibit greater technical maturity, with more stable long-term operation and broader industrial application potential [24,25].

When comparing AWE and PEM directly, key differences emerge across multiple dimensions.

Table 1. Comparison of AWE and PEM technical roadmap parameters.

Performance Parameter	AWE	PEM
Current density (A/cm2)	0.25–0.45	1.0–2.0
Single Electrolyzer Voltage (V)	1.8–2.5	1.8–2.2
Operating Pressure (bar)	10–30	20–50
Minimum Load Rate (%)	20–40	0–10
Efficiency (%)	51–60	46–60
Operating Temperature (°C)	60–90	50–80
Service Life (kh)	55–120	60–100
Maximum Rated Power per Unit (MW)	6	3
Hydrogen Yield per Unit (Nm3/h)	1400	500
Investment Cost (CNY/kW)	4200–12,215	8400–14,000

shows a comparison of the parameters for AWE and PEM technical routes [26,27]. In energy consumption and hydrogen production efficiency, PEM requires less voltage to generate the same amount of hydrogen; a 10 h experiment showed PEM achieved a daily hydrogen output of 10,300 mL, while AWE produced 3600 mL with KOH electrolyte and only 450 mL with seawater, meaning ~8 units of seawater are needed to match the daily production of KOH-based AWE. The PV-PEM system also has higher overall efficiency than PV-AWE, though total system efficiency decreases from 15% to 8% as incident solar irradiation increases [28]. Economically, AWE is more cost-competitive: its hydrogen production cost is CNY 13.86/kg (vs. CNY 19.04/kg for PEM), with a net present cost (NPC) of ¥184.17 million (vs. CNY 251.37 million for PEM) [29]. In capital expenditure (CAPEX), AWE ranges from CNY 4200–12,215/kW, while PEM is higher at CNY 8400–14,000/kW; however, AWE requires more land, whereas PEM features a more compact design [27]. Environmental performance favors PEM: AWE stacks emit 8434 kg CO₂, nearly 2.3 times that of PEM stacks (3695 kg CO₂) [30]. Dynamic response is another critical distinction: PEM adjusts power faster (0.21 s vs. 0.266 s for AWE) [31], with rapid cold-start capability, while AWE suffers from slow cold-start and power response. PEM also exhibits higher current density (1–2 A/cm²) and hydrogen purity (up to 99.9999%) [32], making it suitable for high-purity applications, while AWE’s compatibility with seawater enhances its adaptability in coastal areas. Finally, PEM better handles fluctuating renewable power [33], while AWE excels in stable power environments [34], making each suited to specific renewable energy coupling scenarios.

Although the current hydrogen production cost of PEM is much higher than that of AWE, in the long run, PEM has greater cost reduction potential. The cost gap between PEM and AWE is expected to narrow as PEM’s technological advancements will outpace AWE’s. Munther et al. noted that AWE currently has lower hydrogen production costs and net present costs [29]. However, Subramani et al. noted that PEM costs more now, the cost of PEM will decrease more dramatically due to improvements in current density and the reduction and/or replacement of expensive materials with cheaper alternatives, as well as the fact that PEM is not as mature as AWE [35]. Therefore, we need to choose the ELZ type that is more suitable for the current scenario at different time points.

Regarding the choice of hydrogen storage method, cryogenic liquid hydrogen storage technology suffers from drawbacks such as high technical difficulty due to the high energy consumption and low equipment and process efficiency of hydrogen liquefaction [36]. Organic liquid hydrogen storage suffers from problems such as high dehydrogenation temperature, low efficiency, and high energy consumption. Furthermore, this hydrogen storage technology has low dehydrogenation efficiency, and the dehydrogenation catalyst is prone to coking and deactivation, making it difficult to meet practical application requirements [37]. Solid-state hydrogen storage technology is not yet fully competitive in terms of cost and performance [38]. Due to low infrastructure costs, compressed gaseous hydrogen (CGH2) is the most suitable technology for short-distance transportation in rural areas [39].

As for the hydrogen-to-electricity method, hydrogen internal combustion engines and hydrogen gas turbines produce NO_X during combustion, which pollute the air [40,41]. Hydrogen fuel cells (HFCs) have the advantages of high efficiency and no CO₂ and NO_X emissions compared to internal combustion engines and gas turbines, so they can be used as a hydrogen-to-electricity technology in rural microgrids [42,43].

2.2. Technology Selection for Rural Electricity–Hydrogen Coupled Microgrids

The system structure diagram of the rural electric hydrogen microgrid in the on-grid/off-grid state is shown in Figure 1. In the technology selection of hydrogen energy storage, AEM or PEM is selected as the hydrogen production technology, CGH2 is selected as the hydrogen storage technology, and FC is selected as the hydrogen-to-electricity technology. In rural electric–hydrogen microgrids, WT and PV are the core renewable energy sources, and WT provides stable output across 24 h, while PV generates power during daytime (6:00–20:00) to match peak load; BMU use local straw/wood chips to generate electricity, filling gaps in renewable energy output during low wind/solar periods. BS is for short-term power balance; ELZ converts excess RE to green hydrogen, and HS stores hydrogen for rural hydrogen loads or FC. Rural power load covers residential, agricultural, and small enterprise needs; hydrogen load refers to the demand of local hydrogen refueling stations.

In the on-grid mode, when the microgrid has surplus renewable energy, such as wind and solar, the hydrogen energy storage system consumes the excess electricity, feeding it into the ELZ hydrogen production system to produce hydrogen. This hydrogen is then stored in high-pressure hydrogen storage tanks and supplied to the rural hydrogen load through the hydrogen energy supply chain, thus achieving full lifecycle application of hydrogen energy. When renewable resources are insufficient, the rural microgrid can purchase electricity from the main grid to meet rural electricity and hydrogen production needs, ensuring stable rural power and hydrogen supply.

In the off-grid mode, when the hydrogen energy storage system serves as a load, the operation is similar to that in the on-grid mode. When wind and solar resources are insufficient, the hydrogen energy storage system serves as a power source, primarily to meet the load requirements of the isolated microgrid. The hydrogen produced by the ELZ hydrogen production system not only meets the hydrogen needs of the nearby rural hydrogen industry, but also generates electricity through hydrogen fuel cells, thus achieving full lifecycle application of hydrogen energy.

3. Capacity Planning Model for Rural Electric–Hydrogen Microgrids

3.1. Objective Function

This paper selects LCOE as the planning objective function. LCOE is defined as the total lifetime cost of a generating unit, including capital costs and operation and maintenance costs, divided by the total amount of energy generated by the unit during its lifetime [44]. Compared to other evaluation metrics such as NPV, LCOE, measured in “cost per kilowatt-hour,” directly reflects the average cost of power generation over the entire lifecycle of a project. This allows for a direct comparison of the power generation costs of different technologies (such as wind power, solar power, and thermal power) or different projects, facilitating decision-makers’ quick assessment of which technology is more economical. LCOE can be used to more accurately and intuitively predict and compare the economic efficiency of power generation during the operation period, especially for renewable energy power generation, which is characterized by rapid technological development and complex system configuration [45].

The calculation method of LCOE is as follows:

$$\mathrm{LCOE}={\displaystyle \frac{I_0-\frac{V_0^{\mathrm{R}}}{{(1+r)}^n}+{\displaystyle \sum _{y=1}^n\frac{C_y^{\mathrm{O}\mathrm{M}}}{{(1+r)}^y}}}{{\displaystyle \sum _{y=1}^n\frac{E_y}{{(1+r)}^y}}}}$$

(1)

where $I_0$ represents the initial investment cost of the project; $V_0^R$ represents the residual value of the fixed assets at the end of the operation period, which is the depreciation cost; $C_y^{\mathrm{O}\mathrm{M}}$ represents the total operation and maintenance cost of the rural microgrid project in year y; $r$ represents the discount rate of the entire microgrid project; $n$ is the project life cycle; and $E_y$ represents the power generation of the project in year $y$.

The calculation method for $I_0$ is as follows:

$$I_0=c^{\mathrm{W}\mathrm{T}}{S}^{\mathrm{W}\mathrm{T}}+c^{\mathrm{P}\mathrm{V}}{S}^{\mathrm{P}\mathrm{V}}+c^{\mathrm{B}\mathrm{M}\mathrm{U}}{S}^{\mathrm{B}\mathrm{M}\mathrm{U}}+c^{\mathrm{B}\mathrm{S}}{S}^{\mathrm{B}\mathrm{S}}+c^{\mathrm{E}\mathrm{L}\mathrm{Z}}{S}^{\mathrm{E}\mathrm{L}\mathrm{Z}}+c^{\mathrm{H}\mathrm{S}}{S}^{\mathrm{H}\mathrm{S}}+c^{\mathrm{F}\mathrm{C}}{S}^{\mathrm{F}\mathrm{C}}$$

(2)

where $c^{\mathrm{W}\mathrm{T}/\mathrm{P}\mathrm{V}/\mathrm{B}\mathrm{M}\mathrm{U}/\mathrm{B}\mathrm{S}/\mathrm{E}\mathrm{L}\mathrm{Z}/\mathrm{H}\mathrm{S}/\mathrm{F}\mathrm{C}}$ indicates the unit investment costs of WT/PV/BMU/BS/ELZ/HS/FC; $S^{\mathrm{W}\mathrm{T}/\mathrm{P}\mathrm{V}/\mathrm{B}\mathrm{M}\mathrm{U}/\mathrm{B}\mathrm{S}/\mathrm{E}\mathrm{L}\mathrm{Z}/\mathrm{H}\mathrm{S}/\mathrm{F}\mathrm{C}}$ indicates the construction capacity of WT/PV/BMU/BS/ELZ/HS/FC.

The calculation method for $V_0^{\mathrm{R}}$ is as follows:

$$\begin{array}{l}{V}_0^{\mathrm{R}}={\alpha }_{\mathrm{R}}^{\mathrm{W}\mathrm{T}}{c}^{\mathrm{W}\mathrm{T}}{S}^{\mathrm{W}\mathrm{T}}+{\alpha }_R^{\mathrm{P}\mathrm{V}}{c}^{\mathrm{P}\mathrm{V}}{S}^{\mathrm{P}\mathrm{V}}+{\alpha }_{\mathrm{R}}^{\mathrm{B}\mathrm{M}\mathrm{U}}{c}^{\mathrm{B}\mathrm{M}\mathrm{U}}{S}^{\mathrm{B}\mathrm{M}\mathrm{U}}\\ +{\alpha }_{\mathrm{R}}^{\mathrm{B}\mathrm{S}}{c}^{\mathrm{B}\mathrm{S}}{S}^{\mathrm{B}\mathrm{S}}+{\alpha }_R^{\mathrm{E}\mathrm{L}\mathrm{Z}}{c}^{\mathrm{E}\mathrm{L}\mathrm{Z}}{S}^{\mathrm{E}\mathrm{L}\mathrm{Z}}+{\alpha }_{\mathrm{R}}^{\mathrm{H}\mathrm{S}}{c}^{\mathrm{H}\mathrm{S}}{S}^{\mathrm{H}\mathrm{S}}+{\alpha }_{\mathrm{R}}^{\mathrm{F}\mathrm{C}}{c}^{\mathrm{F}\mathrm{C}}{S}^{\mathrm{F}\mathrm{C}}\end{array}$$

(3)

where ${\alpha }_{\mathrm{R}}^{\mathrm{W}\mathrm{T}/\mathrm{P}\mathrm{V}/\mathrm{B}\mathrm{M}\mathrm{U}/\mathrm{B}\mathrm{S}/\mathrm{E}\mathrm{L}\mathrm{Z}/\mathrm{H}\mathrm{S}/\mathrm{F}\mathrm{C}}$ indicates the depreciation rate of WT/PV/BMU/BS/ELZ/HS/FC.

$C_y^{\mathrm{O}\mathrm{M}}$ is mainly composed of equipment operation and maintenance cost $C_y^{\mathrm{D}\mathrm{O}\mathrm{M}}$, electricity purchase cost $C_y^{\mathrm{P}\mathrm{u}\mathrm{r}}$, biomass purchase costs $C_y^{\mathrm{b}\mathrm{i}\mathrm{o}}$, and hydrogen product trading income $C_y^{\mathrm{H}}$. The calculation formula is as follows:

$$C_y^{\mathrm{O}\mathrm{M}}=C_y^{\mathrm{D}\mathrm{O}\mathrm{M}}+C_y^{\mathrm{P}\mathrm{u}\mathrm{r}}-C_y^{\mathrm{H}}$$

(4)

$$\begin{array}{l}{C}_y^{\mathrm{D}\mathrm{O}\mathrm{M}}={\alpha }_{\mathrm{D}\mathrm{O}\mathrm{M}}^{\mathrm{W}\mathrm{T}}{c}^{\mathrm{W}\mathrm{T}}{S}^{\mathrm{W}\mathrm{T}}+{\alpha }_{\mathrm{D}\mathrm{O}\mathrm{M}}^{\mathrm{P}\mathrm{V}}{c}^{\mathrm{P}\mathrm{V}}{S}^{\mathrm{P}\mathrm{V}}+{\alpha }_{\mathrm{D}\mathrm{O}\mathrm{M}}^{\mathrm{B}\mathrm{M}\mathrm{U}}{c}^{\mathrm{B}\mathrm{M}\mathrm{U}}{S}^{\mathrm{B}\mathrm{M}\mathrm{U}}+{\alpha }_{\mathrm{D}\mathrm{O}\mathrm{M}}^{\mathrm{B}\mathrm{S}}{c}^{\mathrm{B}\mathrm{S}}{S}^{\mathrm{B}\mathrm{S}}\\ +{\alpha }_{\mathrm{D}\mathrm{O}\mathrm{M}}^{\mathrm{ELZ}}{c}^{\mathrm{ELZ}}{S}^{\mathrm{ELZ}}+{\alpha }_{\mathrm{D}\mathrm{O}\mathrm{M}}^{\mathrm{HS}}{c}^{\mathrm{HS}}{S}^{\mathrm{HS}}+{\alpha }_{\mathrm{D}\mathrm{O}\mathrm{M}}^{\mathrm{FC}}{c}^{\mathrm{FC}}{S}^{\mathrm{FC}}\end{array}$$

(5)

$$C_y^{\mathrm{Pur}}={\displaystyle \frac{365{\displaystyle \sum _{t=1}^T{c}_t^{\mathrm{Pur}}{P}_t^{\mathrm{Pur}}}}{4}}$$

(6)

$$C_y^{\mathrm{bio}}={\displaystyle \frac{365{\displaystyle \sum _{t=1}^T{c}^{\mathrm{bio}}{N}_t^{\mathrm{bio}}}}{4}}$$

(7)

$$C_y^{\mathrm{H}}={\displaystyle \frac{365{\displaystyle \sum _{t=1}^T{c}^{\mathrm{H}}{H}_t^{\mathrm{l}}}}{4}}$$

(8)

where ${\alpha }_{\mathrm{DOM}}^{\mathrm{W}\mathrm{T}/\mathrm{P}\mathrm{V}/\mathrm{B}\mathrm{M}\mathrm{U}/\mathrm{B}\mathrm{S}/\mathrm{E}\mathrm{L}\mathrm{Z}/\mathrm{H}\mathrm{S}/\mathrm{F}\mathrm{C}}$ indicates the operation and maintenance rate of WT/PV/BMU/BS/ELZ/HS/FC; $c_t^{\mathrm{Pur}}$ and $P_t^{\mathrm{Pur}}$ indicate the electricity price and electricity purchase amount; $c^{\mathrm{bio}}$ and $N_t^{\mathrm{bio}}$ indicate the biomass price and biomass purchase amount; $c^{\mathrm{H}}$ and $H_t^{\mathrm{l}}$ indicate the hydrogen selling price and hydrogen demand; T indicates the total scheduling duration. This paper selects four typical days as scheduling periods, with T = 96 h and a scheduling interval of 1 h $t\in \{1,2\dots ,T\}$.

3.2. Constraints

3.2.1. Power Balance Constraints

The microgrid needs to maintain a balance between power supply and demand during operation. The power balance constraint is shown in the following formula:

$$P_t^{\mathrm{WT}}+P_t^{\mathrm{PV}}-P_t^{\mathrm{ELZ}}+P_t^{\mathrm{FC}}+P_t^{\mathrm{BMU}}-P_t^{\mathrm{BS},\mathrm{c}}+P_t^{\mathrm{BS},\mathrm{d}}+P_t^{\mathrm{Pur}}=P_t^{\mathrm{l}}$$

(9)

where $P_t^{\mathrm{WT}/\mathrm{PV}}$ indicates the power of WT/PV generation at time t; $P_t^{\mathrm{ELZ}}$ indicates the power consumption of the ELZ at time $t$; $P_t^{\mathrm{FC}}$ indicates the power of the FC generation at time t; $P_t^{\mathrm{BMU}}$ indicates the power of BMU generation at time t; $P_t^{\mathrm{BS},\mathrm{c}/\mathrm{BS},\mathrm{d}}$ indicates the charge/discharge of the BS at time t; $P_t^{\mathrm{l}}$ indicates the electrical load demand at time $t$. In off-grid state, $P_t^{Pur}$ is always equal to 0. $P_t^{Pur}\ge 0$ in the on-grid state, meaning that the system is allowed to purchase electricity from the large grid.

3.2.2. Hydrogen Supply and Demand Balance Constraints

The hydrogen produced by the ELE is used to meet the rural hydrogen demand and supply fuel cell power generation. The hydrogen storage tank is used to adjust the supply and demand balance. The balance constraint is shown in the following formula:

$$H_t^{\mathrm{ELZ}}-H_t^{\mathrm{FC}}-H_t^{\mathrm{HS}}=H_t^{\mathrm{l}}$$

(10)

where $H_t^{\mathrm{ELZ}}$ indicates the hydrogen produced by the ELZ at time t; $H_t^{\mathrm{FC}}$ indicates the hydrogen consumed by the FC at time t; $H_t^{\mathrm{HS}}$ indicates the charge and discharge hydrogen of the hydrogen storage tank at time t.

3.2.3. Hydrogen Energy Storage Operation Constraints

Hydrogen energy is stored as hydrogen produced by the electrolyzers (ELZs), stored in the hydrogen storage tank, and supplies rural hydrogen loads and the fuel cell to achieve full life cycle utilization of hydrogen. Reference (11) indicates the ELZ hydrogen production expression; (12) indicates the fuel cell power generation expression; (13) indicates the time series relationship of hydrogen storage tank capacity; (14), (15) indicates the upper and lower power constraints for the ELZ and the fuel cell; (16) indicates the ELZ and fuel cell cannot operate simultaneously; (17) indicates the upper and lower power constraints for the hydrogen storage tank; (18) indicates the intra-day cycle constraint of hydrogen storage [46].

$$P_t^{\mathrm{ELZ}}={\displaystyle \frac{H_t^{\mathrm{ELZ}}{L}_{{\mathrm{H}}_2}{\rho }_{{\mathrm{H}}_2}}{{\eta }_{\mathrm{ELZ}}}}$$

(11)

$$P_t^{\mathrm{FC}}={\eta }_{\mathrm{FC}}{H}_t^{\mathrm{FC}}{L}_{{\mathrm{H}}_2}{\rho }_{{\mathrm{H}}_2}$$

(12)

$$SO{C}_t^{\mathrm{HS}}=SO{C}_{t-1}^{\mathrm{HS}}+{\displaystyle \frac{H_t^{\mathrm{HS}}\Delta t}{S^{\mathrm{HS}}}}=SO{C}_{t-1}^{\mathrm{HS}}+{\displaystyle \frac{\left(H_t^{\mathrm{ELZ}}-H_t^{\mathrm{FC}}-H_t^{\mathrm{l}}\right)\Delta t}{S^{\mathrm{HS}}}}$$

(13)

$$\lambda U_t^{\mathrm{ELZ}}{S}^{\mathrm{ELZ}}\le P_t^{\mathrm{ELZ}}\le U_t^{\mathrm{ELZ}}{S}^{\mathrm{ELZ}}$$

(14)

$$0\le P_t^{\mathrm{FC}}\le U_t^{\mathrm{FC}}{S}^{\mathrm{FC}}$$

(15)

$$U_t^{\mathrm{ELZ}}+U_t^{\mathrm{FC}}\le 1,\hspace{1em}{U}_t^{\mathrm{ELZ}},U_t^{\mathrm{FC}}\in \{0,1\}$$

(16)

$${\alpha }^{\mathrm{HS},\min }\le SO{C}_t^{\mathrm{HS}}\le {\alpha }^{\mathrm{HS},\max }$$

(17)

$$SO{C}_0^{\mathrm{HS}}=SO{C}_{end}^{\mathrm{HS}}$$

(18)

where ${\eta }_{\mathrm{ELZ}/\mathrm{FC}}$ indicates the hydrogen production efficiency of the ELZ/FC, the hydrogen production efficiency of PEM is higher than that of AWE; $L_{{\mathrm{H}}_2}$ indicates the calorific value of hydrogen, which is 33 kWh/kg; ${\rho }_{{\mathrm{H}}_2}$ indicates the hydrogen density, which is 0.084 kg/m³; $SO{C}_t^{\mathrm{HS}}$ indicates the HS energy status at time t; $\lambda$ indicates the minimum load rate of the ELZ. This AWE will be higher than that of PEM. $U_t^{\mathrm{ELZ}/\mathrm{FC}}$ indicates the work state variables related to ELZ and FC at time t; ${\alpha }^{\mathrm{HS},\max /\mathrm{HS},\min }$ indicates the upper and lower limits coefficients of HS capacity.

3.2.4. Battery Energy Storage Operation Constraints

Microgrids use battery storage to smooth out power supply and demand. Reference (19) indicates the upper and lower constraints for the BS charge and discharge power; (20) indicates the relationship between the BS charge and discharge power and capacity; (21) indicates the BS cannot be charged and discharged at the same time; (22) the time series relationship of the BS capacity; (23) indicates the upper and lower power constraints for the BS; (24) indicates the intra-day cycle constraint of battery energy storage.

$$\left\{\begin{array}{l}0\le P_t^{\mathrm{BS},\mathrm{c}}\le U_t^{\mathrm{BS},\mathrm{c}}{P}^{\mathrm{BS},\max }\\ 0\le P_t^{\mathrm{BS},\mathrm{d}}\le U_t^{\mathrm{BS},\mathrm{d}}{P}^{\mathrm{BS},\max }\end{array}\right.$$

(19)

$$S^{\mathrm{BS}}=k_{\mathrm{BS}}{P}^{\mathrm{BS},\max }$$

(20)

$$U_t^{\mathrm{BS},\mathrm{c}}+U_t^{\mathrm{BS},\mathrm{d}}\le 1,\hspace{1em}{U}_t^{\mathrm{BS},\mathrm{c}},U_t^{\mathrm{BS},\mathrm{d}}\in \{0,1\}$$

(21)

$$SO{C}_t^{\mathrm{BS}}=SO{C}_{t-1}^{\mathrm{BS}}+{\displaystyle \frac{\left({\eta }_{\mathrm{BS}}{P}_t^{\mathrm{BS},\mathrm{c}}-\frac{P_t^{\mathrm{BS},\mathrm{d}}}{{\eta }_{BS}}\right)\Delta t}{S^{BS}}}$$

(22)

$${\alpha }^{\mathrm{BS},\min }\le SO{C}_t^{\mathrm{BS}}\le {\alpha }^{\mathrm{BS},\max }$$

(23)

$$SO{C}_0^{\mathrm{BS}}=SO{C}_{end}^{\mathrm{BS}}$$

(24)

where $P^{\mathrm{BS},\max }$ indicates the upper limit of BS charge and discharge power; $k_{\mathrm{BS}}$ indicates the proportional relationship between BS power and capacity, in units of h. $U_t^{\mathrm{BS},\mathrm{c}/\mathrm{BS},\mathrm{d}}$ indicates the work state variables related to BS charge and discharge power; $SO{C}_t^{\mathrm{BS}}$ indicates the BS power status; ${\eta }_{BS}$ indicates the BS charge and discharge power efficiency; ${\alpha }^{\mathrm{BS},\max /\mathrm{BS},\min }$ indicates the upper and lower limits coefficients of BS capacity.

3.2.5. Photovoltaic and Wind Power Generation Operation Constraints

The following constraints must be met when WT and PV are in operation:

$$0\le P_t^{\mathrm{WT}}\le p_t^{\mathrm{WT}}{S}^{\mathrm{WT}}$$

(25)

$$0\le P_t^{\mathrm{PV}}\le p_t^{\mathrm{PV}}{S}^{\mathrm{PV}}$$

(26)

where $p_t^{\mathrm{WT}/\mathrm{PV}}$ indicates the power generated per unit capacity of WT and PV.

3.2.6. Biomass Generation Operation Constraints

Rural areas have abundant biomass resources such as straw and wood chips, which can be used to generate electricity to meet rural power supply needs [47]. Taking biomass power generation as an example, there is an equation relationship between the power generation of a biomass unit and the amount of biomass consumed:

$$P_t^{\mathrm{BMU}}=k_{\mathrm{BMU}}{N}_t^{\mathrm{bio}}$$

(27)

$$0\le P_t^{\mathrm{BMU}}\le S^{\mathrm{BMU}}$$

(28)

where $k_{\mathrm{BMU}}$ indicates the amount of electricity generated per unit of biomass consumed, in units of kg/kWh.

3.3. Evaluation Indicators

3.3.1. Wind and Solar Power Curtailment Rates

The wind and solar curtailment rate measures the proportion of renewable energy (such as wind and solar) that is not utilized in the power generation process. Due to various technical, economic and policy factors, some renewable energy generation capacity may not be absorbed by the power grid, resulting in wind or solar curtailment. The wind and solar curtailment rate is usually expressed as a percentage and is calculated as follows:

$$L^{\mathrm{WT}}=(1-{\displaystyle \frac{{\displaystyle \sum _{t=1}^T{P}_t^{\mathrm{WT}}}}{{\displaystyle \sum _{t=1}^T{p}_t^{\mathrm{WT}}{S}^{\mathrm{WT}}}}})\times 100\%$$

(29)

$$L^{\mathrm{PV}}=(1-{\displaystyle \frac{{\displaystyle \sum _{t=1}^T{P}_t^{\mathrm{PV}}}}{{\displaystyle \sum _{t=1}^T{p}_t^{\mathrm{PV}}{S}^{\mathrm{PV}}}}})\times 100\%$$

(30)

where $L^{\mathrm{WT}}$ and $L^{\mathrm{PV}}$ indicates the curtailment rate of wind power and solar power, respectively.

3.3.2. Self-Balancing Rate

The proportion of load that an on-grid microgrid can meet through its own distributed power supply reflects the microgrid’s power supply capacity and its degree of dependence on the main grid. When the on-grid microgrid’s internal output cannot meet the load demand, the microgrid purchases electricity from the main grid to meet the remaining load demand. The self-balancing rate represents the ratio of the power provided to the load by the microgrid to the total power demand of the load. The specific expression is as follows:

$$R={\displaystyle \frac{E_{\mathrm{self}}}{E_{\mathrm{total}}}}\times 100\%=(1-{\displaystyle \frac{{\displaystyle \sum _{t=1}^T{P}_t^{\mathrm{Pur}}}}{{\displaystyle \sum _{t=1}^T{P}_t^{\mathrm{load}}}}})\times 100\%$$

(31)

where $R$ represents the self-balancing rate; $E_{\mathrm{self}}$ is the amount of electricity provided to the load within the microgrid; $E_{\mathrm{total}}$ is the total load demand.

4. Solution Method

The K-means clustering algorithm is an iterative unsupervised learning method designed to partition a dataset into k distinct clusters, where intra-cluster similarity is maximized and inter-cluster dissimilarity is minimized [48]. Grouping is primarily based on “closeness” or “similarity”: the more similar the objects within a group, the better, and the greater the difference between groups, the better. Distance measures include Euclidean distance, Manhattan distance, and Chebyshev distance. Clustering algorithms typically use Euclidean distance as a similarity measure and the sum of squared errors (SSE) as the objective function to measure clustering quality. By minimizing the objective function, data points are divided into k clusters based on their distance from the cluster centers. In this study, it was utilized to select one typical day for wind power, photovoltaic power, and load profiles in each of the four seasons (spring, summer, autumn, and winter), resulting in four typical days for the entire year.

4.1. Mathematical Definitions

4.1.1. Euclidean Distance

The similarity between data points is measured by the Euclidean distance, which quantifies the geometric distance between two n-dimensional vectors. For two data points $x$ and $y$, the Euclidean distance $d(x,y)$ is defined as follows:

$$d(x,y)=\sqrt{{\displaystyle \sum _{i=1}^n{(x_i-y_i)}^2}}$$

(32)

where $n$ represents the dimension of the data point; $x$ and $y$ represent n-dimensional data points; $x_i$ and $y_i$ represent the values of the i-th dimension of n-dimensional data points $x$ and $y$.

4.1.2. Distance Between a Data Point and a Clustering Center

For a data point $x$ and the clustering center $c_i$ of the $i$-th cluster, their Euclidean distance $d(x,c_i)$ is calculated as follows:

$$d(x,c_i)=\sqrt{{\displaystyle \sum _{i=1}^n{(x_i-c_i)}^2}}$$

(33)

where $c_i$ represent the cluster centers of the $i$-th cluster.

4.1.3. Sum of Squared Error

The quality of clustering is evaluated by the Sum of Squared Errors (SSE), defined as the sum of the squared distances from each data point to its corresponding clustering center. Mathematically, SSE is expressed as follows:

$$SSE={\displaystyle \sum _{i=1}^k{\displaystyle \sum _{x\in C_i}{\left|d(x,c_i)\right|}^2}}$$

(34)

where $k$ represents the number of clusters; $C_i$ represents the i-th cluster.

4.2. Implementation Steps for Typical Day Selection in Four Seasons

The K-means clustering process for selecting typical days in each seasonal dataset involves the following iterative steps:

Step 1: The measured annual wind power generation, photovoltaic power generation, and the load data from a realistic rural area in central China were divided into four subsets: spring, summer, autumn, and winter. Each subset was processed independently.

Step 2: For the dataset of a specific season, set $k=1$ (to select one typical day) and randomly initialize the cluster center.

Step 3: Calculate the Euclidean distance using Equations (32) and (33) between each daily profile in the seasonal dataset and the initial cluster center. Assign all data points to the single cluster.

Step 4: Compute the mean of all data points in the cluster and update the cluster center to this mean.

Step 5: Calculate Equation (34). If the cluster center stops changing, the SSE stabilizes, or the maximum iteration count is reached, terminate the process. Otherwise, return to Step 3.

Step 6: The converged cluster center is the typical day for the season. Repeat Steps 2–5 for all four seasons.

5. Case Study

5.1. Data Description

The unit capacity investment cost, residual value rate, and operation and maintenance rate of system equipment are shown in

Table 2. Equipment cost parameter table.

Equipment	Initial Investment Cost		Residual Value/%	Operation and Maintenance Rate/%
WT	6000 CNY/kW		5	1.5
PV	3750 CNY/kW		5	1.5
BMU	8000 CNY/kW		5	1.5
BS	1500 CNY/kWh		10	1.1
AWE	Year 2025	2500 CNY/kW	5	3
	Year 2030	2260 CNY/kW
PEM	Year 2025	7500 CNY/kW	5	3
	Year 2030	3800 CNY/kW
HS	6000 CNY/Nm3		5	3
FC	4000 CNY/kW		5	3

. The cost reduction trend of PEM is faster than that of AWE, and the cost gap between PEM and AWE will further narrow in the future. Equipment operating parameters are shown in

Table 3. Equipment operating parameter table.

Parameter	Value	Parameter	Value
AWE hydrogen production efficiency ${\eta }_{\mathrm{ELZ}}$	0.675	PEM hydrogen production efficiency ${\eta }_{\mathrm{ELZ}}$	0.85
The minimum load rate of AWE $\lambda$	0.3	The minimum load rate of PEM $\lambda$	0.05
${\eta }_{\mathrm{FC}}$	0.6	${\eta }_{\mathrm{BS}}$	0.86
${\alpha }^{\mathrm{BS},\min }$/${\alpha }^{\mathrm{BS},\max }$	0.1/0.9	${\alpha }^{\mathrm{HS},\min }$/${\alpha }^{\mathrm{HS},\max }$	0.1/0.9
$SO{C}_0^{\mathrm{BS}}$	0.5	$SO{C}_0^{\mathrm{HS}}$	0.5
$k_{\mathrm{BS}}$	2 h	$k_{\mathrm{BMU}}$	0.218 CNY/kWh
$c^{\mathrm{H}}$	29.33 CNY/kg	$c^{\mathrm{bio}}$	155.65 CNY/t

below.

This study is based on the actual annual wind and solar power and electricity load data of a rural area in central China. The annual wind, solar, and ecliptic data are shown in Appendix A Figure A1, Figure A2 and Figure A3. The K-means algorithm described in Section 4 was used to process the annual data and select typical days for each of the four seasons. The specific typical day curves of wind, solar, and electricity load are shown in Figure 2. The microgrid electricity purchase price uses peak and valley electricity prices according to local policies. The peak and valley time periods are divided differently in different seasons. Specific data are shown in Figure 3. In Figure 3, load demand exhibits a distinct bimodal pattern, with peak electricity demand concentrated primarily between 8:00–12:00 and 17:00–20:00. Wind power output is relatively flat, with no zero value over the 24-h period, but output is lower between 8:00–20:00, exhibiting an “anti-peak” characteristic. Photovoltaic power output is limited to daytime hours between 6:00–20:00, closely matching peak electricity demand.

At present, the promotion of hydrogen fuel cell vehicles and the construction of supporting hydrogen refueling stations are being accelerated worldwide [49]. Therefore, the hydrogen load demand in this article is defined as the hydrogen supply demand of hydrogen refueling stations. The daily hydrogen load of hydrogen refueling stations is in a “V” shape. Assuming that the hydrogen refueling station can normally supply the hydrogen demand of three hydrogen fuel buses per day, a hydrogen fuel bus travels 200 km per day and consumes 5.5 kg of hydrogen per 100 km. The specific hydrogen load is shown in Figure 4.

The proposed rural hydrogen–electricity microgrid system was simulated using MATLAB 2024b (MathWorks, Natick, MA, USA), a widely used numerical computing and simulation platform in power system research. Official information and technical details of MATLAB 2024b can be accessed via the MathWorks website: https://www.mathworks.com/products/matlab.html (accessed on 1 September 2024).

5.2. Technical and Economic Analysis of Energy Storage in Off-Grid Operation Mode

In order to verify the economic feasibility of hydrogen energy storage technology in rural microgrids, we set up the following case verification, where the planning time is 2025, and the ELZ type is selected as AWE in an off-grid state:

Case 1: energy storage only considers BS;

Case 2: The energy storage method only considers HS and takes into account the trading income of hydrogen in the investment cost.

Case 3: The energy storage method considers electric-hydrogen hybrid storage and takes into account the trading income of hydrogen in the investment cost.

The results are shown in

Table 4. Planning results of different energy storage technologies in off-grid conditions.

Target	LCOE/(CNY/kWh)	WT/kW	PV/kW	BMU/kW	ELZ/kW	HS/m3	FC/kW	BS/kWh	${\mathit{L}}^{\mathit{W}\mathit{T}}$	${\mathit{L}}^{\mathit{P}\mathit{V}}$
Case 1	0.4219	467	250	128	0	0	0	132	4.2%	0
Case 2	0.2851	702	376	120	104	290	0	0	7.55%	2.87%
Case 3	0.2824	672	403	125	102	250	0	66	6.38%	0.31%

. In Table 4, the LCOE of Case 2 is 0.1368 CNY/kWh lower than that of Case 1. Due to the large amount of electricity required to produce hydrogen to meet the hydrogen load, Case 2 requires increased WT and PV capacity. As shown in Table 1, hydrogen storage construction costs are higher than those of electrical storage. The lower LCOE for hydrogen storage is due to the rural microgrid’s profit from hydrogen sales, which offsets the equipment costs. This demonstrates that hydrogen storage offers superior economic benefits compared to electrical storage in meeting rural load demands. Case 3’s LCOE decreases by 0.0027 CNY/kWh compared to Case 2. This is because the ELZ has a minimum operating range and lacks the flexibility to adjust compared to electrical storage, necessitating the deployment of more WT with stable output rather than intermittent PV. The addition of electrical storage compensates for the lack of hydrogen storage’s adjustable capacity and addresses the PV absorption issue. System planning reduces high-cost WT (6000 CNY/kW) in favor of low-cost, highly load-matched PV (3750 CNY/kW), reducing total investment costs. Figure 5 shows the electricity balance diagram using electric-hydrogen hybrid energy storage. Photovoltaic and wind power complement each other, and biomass power generation fills the gap in electricity. When the pressure to ensure power supply is high in summer and autumn, it is necessary to increase biomass power generation. Figure 6 shows the hydrogen supply and demand balance diagram. Under the mitigating effect of the hydrogen storage tank, the supply and demand of hydrogen are balanced.

In Table 4, the planning results of Case 2 and Case 3 do not configure FC. The reason is that the conversion efficiency of electricity-hydrogen-electricity is only 40.5% (67.5% × 60%), which is too low compared to the conversion efficiency of BS of about 74% (86% × 86%). The direct sale of hydrogen produced by ELZ is more economical. Therefore, the current rural microgrid project does not need to configure FC, and the energy storage method is selected as an electricity-hydrogen hybrid energy storage.

5.3. Comparative Analysis of the Economic Efficiency of On-Grid and Off-Grid Operation

The essential difference between on-grid and off-grid microgrids is whether they rely on the main grid. To compare the differences between the two models, we will now plan rural microgrids in both on-grid and off-grid states, where the planning time is 2025, and the ELZ type is selected as AWE. Based on the conclusions in Section 4.1, FC will not be deployed in the future.

The results are shown in

Table 5. Planning results in grid-connected and off-grid states.

Target	LCOE/(CNY/kWh)	WT/kW	PV/kW	BMU/kW	ELZ/kW	HS/m3	BS/kWh	${\mathit{L}}^{\mathit{W}\mathit{T}}$	${\mathit{L}}^{\mathit{P}\mathit{V}}$	$\mathit{R}$
Off-grid	0.2824	672	403	125	102	250	66	6.38%	0.31%	100%
On-grid	0.2744	610	428	120	103	176	88	4.47%	0.16%	94.43%

. The LCOE in grid-connected mode is 0.008 CNY/kWh lower than in off-grid mode, with $L^{WT}$ decreasing by 1.91% and $L^{PV}$ decreasing by 0.15%. This is because with the support of the larger grid, reducing the construction capacity of WT, BMU, and HS can still meet the system’s load demand, reducing equipment investment costs. The increase in PV capacity is due to the fact that in grid-connected mode, the disadvantage of PV’s intermittent output can be compensated for by purchasing electricity. PV, with its lower cost and better output-to-load match, has an advantage over WT. The increase in BS capacity is due to the fact that in off-grid mode, battery energy storage is only used to maintain a balance between electricity supply and demand, while in grid-connected mode, it can be used as a tool for peak–valley arbitrage. Figure 7 shows the power supply and demand balance diagram of the microgrid in the grid-connected state. Compared with the off-grid state, the grid-connected microgrid can more efficiently utilize wind and solar resources. At the same time, it can purchase electricity to supply the power load when the electricity price is low, and can also store it in the BS and discharge it during peak power consumption. Figure 8 shows a hydrogen supply and demand balance diagram.

Therefore, when grid connection is available, microgrids should preferably choose an on-grid mode.

5.4. Comparative Analysis of AWE and PEM at Different Time Points

The current mainstream hydrogen production methods are AWE and PEM. Compared to AWE, PEM has a lower load-shedding operating range and higher hydrogen production efficiency, but its equipment costs are currently higher, three times that of AWE. There is room for further cost reductions in PEM equipment, potentially down to 3800 ¥/kWh by 2030. To select the optimal hydrogen production method, we planned rural microgrids in grid-connected states using AWE and PEM in 2025 and 2030. To ensure equal generation resources for comparison, we fixed the capacities of WT, PV, and BMU, with the specific capacities uniformly set to the 2025 AWE plan.

The results are shown in

Table 6. Planning results for hydrogen production using hybrid electric–hydrogen energy storage at different time stages, selecting AWE and PEM, respectively.

Target	Year	LCOE/(CNY/kWh)	WT/kW	PV/kW	BMU/kW	ELZ/kW	HS/m3	BS/kWh	${\mathit{L}}^{\mathit{W}\mathit{T}}$	${\mathit{L}}^{\mathit{P}\mathit{V}}$	$\mathit{R}$
AWE	2025	0.2744	610	428	120	103	176	88	4.47%	0.16%	94.43%
	2030	0.2730	610	428	120	103	176	88	4.47%	0.16%	94.43%
PEM	2025	0.2916	610	428	120	79	94	157	7.24%	1.62%	95.48%
	2030	0.2726	610	428	120	85	88	171	7.25%	0.92%	95.43%

. In 2025, PEM’s LCOE will be 0.0172 CNY/kWh higher than AWE, primarily due to the significantly higher construction costs of PEM equipment compared to AWE. It can be seen that when using PEM as a hydrogen production method, the construction capacity of PEM and HS is lower. This is due to PEM’s higher hydrogen production efficiency and lower minimum load rate, making regulation more flexible and eliminating the need to increase hydrogen storage capacity to ensure a stable hydrogen load. The increase in BS and wind and solar curtailment rates when using PEM is due to PEM’s smaller construction capacity. While flexible, its overall ability to handle fluctuations is insufficient, leading to the system prioritizing cheaper electrochemical energy storage to smooth fluctuations. The “implicit costs” of curtailing wind and solar power are lower than the “explicit costs” of expanding PEM or energy storage. PEM’s higher self-balancing rate is due to its low load rate, which allows for it to quickly respond to real-time changes in renewable energy and load. Furthermore, PEM’s high efficiency reduces losses in converting electricity to hydrogen, reducing the need for external power purchases.

In 2030, PEM’s LCOE will decrease to 0.2726, 0.0004 CNY/kWh lower than AWE’s. This is primarily due to PEM’s significant cost reduction potential. From 2025 to 2030, PEM costs will decrease by 3700 CNY/kW, far exceeding AWE’s 240 CNY/kW. Furthermore, as PEM costs decrease, the capacity of PEM and BS increases, while the capacity of HS and curtailment decreases. This is because, as the PEM’s cost disadvantage narrows, expanding PEM capacity can directly absorb more fixed excess renewable energy power, reducing curtailment. Its low load factor allows for real-time adjustment of hydrogen production rates to directly match hydrogen load consumption, reducing the need for hydrogen tank buffers and encouraging the system to deploy more low-cost electrochemical energy storage to smooth fluctuations.

Therefore, at the current stage, due to the high cost of PEM, AWE becomes the preferred method of hydrogen production; however, in 2030, the cost of PEM will drop significantly. At the same time, due to the high efficiency and low minimum operating restrictions of PEM, it is more suitable for the scenario of rural microgrid renewable energy fluctuations.

5.5. Sensitivity Analysis

5.5.1. Hydrogen Price Analysis

With a planning timeframe of 2025, an AWE-type ELZ, and a hybrid electric-hydrogen energy storage system in an on-grid state, the LCOE under different hydrogen prices is shown in Figure 9. Figure 9 shows that the LCOE of the rural electric–hydrogen microgrid decreases as the hydrogen price increases, confirming that hydrogen sales are a direct and stable source of income, and that the hydrogen price is a key lever for improving system economics.

5.5.2. Electrolyzer Price Analysis

The LCOE under different ELZ (PEM and AWE) investment costs is shown in Figure 10, assuming the energy storage type is a hybrid electric-hydrogen energy storage system and the system is connected to the grid. Figure 10 shows that the LCOE of rural electric-hydrogen microgrids increases with the investment cost of ELZ. Furthermore, at the same investment cost, the LCOE of ELZs with PEM is lower than those with AWE, indicating the applicability of PEM in rural electric-hydrogen microgrids. As the cost gap between PEM and AWE narrows in the future, PEM will become a better option for hydrogen production in rural electric–hydrogen microgrids.

5.5.3. Discount Rate Analysis

With a planning timeframe of 2025, an AWE-type ELZ, and a hybrid electric–hydrogen energy storage system in an on-grid state, the LCOE under different discount rates is shown in Figure 11. Figure 11 shows that the LCOE of rural electric–hydrogen microgrids increases with the increase of the discount rate. This trend arises because a higher discount rate amplifies the annualized cost of initial investments, as future cash flows are discounted more heavily. These findings highlight the importance of securing low-cost financing for renewable energy projects. Applying for low-interest loans supported by renewable energy project policies can significantly reduce the LCOE by lowering the discount rate by 2–3 percentage points, thereby significantly improving its competitiveness relative to traditional energy sources.

6. Conclusions and Discussions

To promote the development of rural hydrogen energy and reduce the investment cost of rural hydrogen microgrids, this paper evaluates the economic performance of technical selection for rural hydrogen microgrid systems. The main conclusions are as follows:

(1)

In the off-grid operation mode, the economic performance of selecting electric-hydrogen hybrid energy storage as the energy storage method is the best. The technical economic performance of configuring fuel cells for power generation is too poor and is not considered in the rural microgrid scenario.

(2)

Compared with the off-grid mode, the grid-connected mode can effectively reduce the equipment construction capacity, save floor space, and have better technical economic performance. It can effectively reduce the wind and solar power curtailment rate. Rural microgrids should try to choose the grid-connected mode.

(3)

The current PEM cost is too high, and AWE is the preferred hydrogen production method. However, after the cost of PEM is significantly reduced in the future, PEM will become the preferred hydrogen production method for rural microgrids due to its high hydrogen production efficiency and low minimum load limit.

This paper currently investigates the economic feasibility comparison of rural hydrogen microgrids, without considering other environmental benefits and simplifies the modeling of ELZ, BMU, etc., into linear models. More refined modeling would further improve the accuracy of the results. More refined multi-seasonal modeling will be fully considered in our future research.