Does It Really Reduce Emissions? Full-Chain Life Cycle Emission and Economic Benefits Analysis of New Energy Vehicles in China

Abstract

Scientific assessment of energy conservation, emissions reduction, public health externalities, and economic costs is crucial for the sustainable development of new energy vehicles (NEVs). Despite minimal emissions during the operational phase of NEVs, the production process of energy, such as electricity and hydrogen, contributes to pollution across the full supply chain, shifting environmental and health burdens to upstream sectors and raising concerns about the overall societal benefits. To address this, we apply a full-chain life cycle assessment (FC-LCA) framework that integrates emissions from vehicle production, energy supply, and end-of-life stages, while simultaneously quantifying health-related mortality attributable to key pollutants. By incorporating upstream energy production structure and downstream industry emissions, this approach captures the complete energy supply chain and enables a systematic comparison between NEVs and conventional vehicles. We further employed and compared ARIMA, LSTM, and Bi-LSTM models to forecast future vehicle demand and defined different forecasting scenarios for China’s passenger vehicle sector. Results provide policy-relevant insights for decision-makers to make informed policy choices concerning the widespread implementation of NEVs in a sustainable manner.

1. Introduction

Urbanization and industrial upgrading are increasing the tertiary sector’s share and driving growth in transportation-related greenhouse gas (GHG) emissions, with cross-regional economic integration further intensifying transport demand, congestion, and air pollution. Against this backdrop, governments worldwide—across both developed and developing economies—are accelerating vehicle electrification, not only to address energy, climate, and environmental challenges, but also to strengthen their positions within the evolving global automotive value chain [1]. Within this global context, China presents a particularly important case: it recorded the highest total CO₂ emissions in 2024, while its transport sector ranked sixth globally in 2022 (as shown in Figure 1). Reducing transport-related emissions is therefore critical for both global climate mitigation and China’s carbon neutrality goals. Achieving effective emission control while sustaining economic growth can also provide valuable insights for energy conservation and emission reduction worldwide.

Developing new energy vehicles is a key strategy in China to reduce transport GHG emissions, address climate change, and promote green travel. As shown in Figure 2, Sales soared from 9.5 M units in 2023 to 12.9 M in 2024, maintaining the world’s top position for ten consecutive years. In 2024, new-energy vehicle sales accounted for 40.9% of China’s market, up 35.5% YoY, and made up over 70% of global sales.

Over the past 20 years, China has promoted new energy vehicles through subsidies and higher market entry barriers, leading to rapid growth in the electric vehicle market alongside a decline in petroleum-based vehicle sales. Powered by on-board energy and electric motors, electric vehicles produce no tailpipe CO₂ emissions during operation, and their large-scale deployment is expected to reduce greenhouse gas and air pollutant emissions such as CO₂, SO₂, and NO_x, while reshaping the global automotive industry. However, the environmental impact of new energy vehicles (NEVs) presents a notable paradox. Although NEVs generate no tailpipe emissions and are widely regarded as a key solution to environmental and energy challenges, their overall emission reduction benefits remain contested, with existing studies pointing to several potential limitations [3,4,5,6,7]. Due to their greater weight and rapid acceleration, NEVs may generate higher non-exhaust particulate matter than internal combustion engine vehicles (ICEVs) [8,9]. Without targeted policies, consumer preferences for larger, long-range EVs could drive an increase in PM_2.5 emissions [10]. Such transport-related PM_2.5 emissions alone are projected to cause over 10,000 premature deaths annually in the EU by 2030, highlighting the severe public health implications of road traffic [11]. Despite these concerns, the health impacts of different vehicle technologies remain insufficiently quantified, particularly in a comparative life-cycle context.

More recently, a growing body of literature has further questioned whether NEVs deliver substantial energy savings and emission reductions over their full life cycle, suggesting that these benefits may be smaller or highly dependent on upstream processes and regional conditions [12,13,14,15]. Nevertheless, it is generally accepted that EVs can effectively mitigate greenhouse gas (GHG) emissions under optimized technological and energy-system configurations. Despite this consensus, most existing studies remain centered on carbon inventories and mitigation outcomes, with relatively limited attention paid to how these emission shifts translate into public health impacts.

Empirical studies have assessed the emissions and economic effects of new energy vehicles (NEVs) from multiple angles. Xing et al. [16] show that EVs emission benefits are often overstated because EVs mainly replace relatively fuel-efficient vehicles. Sahin [17] finds that the emissions reductions from battery electric vehicles depend critically on the carbon intensity of electricity generation. Using traffic monitoring data, Shang [18] shows that significant emission reductions arise only after NEV penetration exceeds a threshold. At the macro level, Cheng [19] documents interactions between NEV promotion, transport efficiency, and regional economic growth, underscoring the importance of coordinated infrastructure development. Notably, while these studies clarify the physical emission pathways, they often stop short of quantifying the resulting health externalities, leaving a gap in understanding the comparative health burden between NEVs and ICEVs across their entire life cycles.

NEVs such as electric vehicles rely on the power grid for electricity and have a similar manufacturing process to internal combustion engine vehicles (ICEVs), resulting in significant CO₂ and pollutant emissions during both production and operation. A narrow focus on transport GHG emissions is insufficient, as upstream and downstream industries are also affected. Therefore, the existing literature mainly studies CO₂ reduction in the transport sector are neglecting pollution transferred to related industries.

This study introduces a comprehensive analytical framework that assesses NEVs within their entire biosphere, encompassing the full industry chain and life cycle. By assessing NEVs within this framework, we not only rigorously compare the GHG reduction potential of NEVs and ICEVs but also integrate a health impact assessment to compare the health burden of different vehicle types. This approach allows for a more robust comparison of the environmental-health trade-offs inherent in vehicle electrification.

Balancing environmental governance and economic implications is also crucial in promoting NEVs. A cost-benefit analysis within the FC-LCA framework is further conducted, considering user costs, pollution costs, and emission reduction benefits. The study focuses on the vehicle material cycle and energy usage cycle, comparing CO₂ and pollutant emissions between NEVs and ICEVs.

Furthermore, different scenarios, based on various driving forces, are defined for a comparative analysis of future GHG emissions. ARIMA, LSTM, and Bi-LSTM models are used to forecast vehicle demand and stock, with the best model selected for final projections. These projections and parameter settings enable an in-depth exploration of possible scenarios for individual vehicles and the Chinese passenger vehicle sector. The results contribute to a deeper understanding of the environmental, economic, and health-related aspects associated with NEVs, providing valuable insights to guide decision-makers in their pursuit of sustainable solutions.

The remainder of this paper is organized as follows: firstly, we reviewed relevant research literature; then we introduced the methodology and data sources used in this study; furthermore, we obtained the research results and analyzed them; finally, we summarized the conclusions and put forward relevant policy recommendations.

2. Literature Review

Many studies compare the environmental impacts of traditional fuel vehicles (ICEVs) and new energy vehicles (NEVs), focusing on fuel, component, and vehicle life cycles [20,21,22,23]. Most conclude that NEVs, especially battery electric vehicles (BEVs), offer significant energy consumption and GHG reduction benefits. For example, Huo et al. [24] conducted a cross-country life-cycle analysis of light-duty vehicle fleet electrification in the United States, China, and the United Kingdom, showing that electrification can reduce fleet-level life-cycle GHG emissions by more than 50% by 2050, while emphasizing that the effectiveness and policy priorities of electrification differ markedly across countries due to heterogeneity in fleet composition, electricity systems, and vehicle ownership dynamics.

In terms of environmental impacts of BEVs, a substantial body of LCA studies indicates that they play an important role in reducing greenhouse gas emissions and mitigating global warming. However, their overall benefits remain debated, as PM_2.5 and SO₂ emissions may still exceed those of ICEVs under certain conditions. Mehlig et al. [25] find that although BEVs can substantially reduce well-to-wheel CO₂ emissions, short-term improvements in local air quality are limited, with NO_x reductions mainly driven by ICEV turnover and PM_2.5 largely unaffected due to non-exhaust sources. Similarly, Pryciński et al. [26] show that BEV environmental performance strongly depends on the electricity mix, and in coal-dominated systems, upstream emissions can significantly offset electrification benefits.

An increasing number of studies have begun to quantify the health impacts of vehicle emissions. Hänninen et al. [11] assess transport-related air pollution in the EU27 under a 2030 policy scenario, showing that even with full implementation of current regulations, traffic emissions could still cause substantial premature mortality, with ultrafine particles contributing the largest burden, followed by NO₂, PM_2.5, black carbon, and SOA. Road traffic and diesel exhaust remain key sources. Furthermore, Ma et al. conducted a regional life cycle assessment in China, reporting that battery electric vehicles (BEVs) generally exhibit lower health impacts (0.005 DALY) than gasoline vehicles (GVs, 0.007 DALY). However, in coal-dependent regions such as Heilongjiang, electricity-related emissions may elevate particulate matter formation and associated respiratory risks [27].

Studies on BEVs often overlook benefits beyond CO₂ emissions and energy consumption. Only a small number of studies focus on economic benefits of NEVs. Existing evidence indicates that the cost competitiveness of BEVs varies markedly across regions and policy environments. For instance, Li et al. [28] show that BEVs and FCEVs are economically attractive, mainly in China’s first-tier cities, where strict traffic restrictions and environmental regulations substantially increase the implicit costs of conventional vehicles; in lower-tier cities, ICEVs remain dominant without stronger policy support. Using real-world driving cycles, Hastunç et al. [29] find that EVs can achieve a lower total cost of ownership than comparable diesel vehicles in France, with subsidies and depreciation playing a more important role than technical performance. By contrast, İnal et al. [30] report that ICEVs still have the lowest life-cycle costs in Sweden and argue that the removal of EV subsidies is likely to slow EV adoption and undermine decarbonization efforts.

LCA (Life Cycle Assessment) is a crucial tool in product environmental management, enabling quantitative analysis of environmental and resource issues across a product’s life cycle. Recent studies [31,32,33] have extended or integrated LCA approaches to assess product environmental impacts. Numerous scholars have compared the energy consumption and environmental impact of new energy vehicles versus traditional ICEVs [34,35,36]. Various practical models exist for motor vehicle life-cycle evaluation, including the GREET model from Argonne National Laboratory (USA), the EIO-LCA model from the University of Toronto, and the DfE model used by Mercedes-Benz. Notably, the WTW (Well-to-Wheel) system and the GREET model are widely utilized in life cycle evaluation studies.

Currently, most studies on new energy vehicles focus on CO₂ emissions from fuel and production cycles, with little attention to material acquisition and end-of-life recycling stages. Few include emissions from energy sources such as electricity and hydrogen in Life Cycle Assessments. There is also limited analysis of the economic benefits of emission reduction, and a need to unify environmental, economic, and health impacts for comprehensive comparisons. In particular, few studies jointly consider emissions, economic costs, and health impacts within a single integrated framework. Addressing this gap would support the development of more practical green and sustainable urban policies.

3. Full-Chain Life Cycle Assessment Model

To analyze CO₂ emission reduction and the environmental impact of new energy vehicles (NEVs) versus traditional internal combustion engine vehicles (ICEVs), this paper developed a full-chain LCA model. The vehicle life cycle is segmented into production, use, and end-of-life recycling stages, encompassing material and energy cycles [37]. The material cycle involves raw material extraction, vehicle production/assembly, and distribution. The use stage incorporates energy generation source structure and upstream emissions within the full-chain LCA framework, which integrates the vehicle’s material and energy cycles as a biosphere system. The horizontal and vertical cycle chains define the cross-chain full-chain LCA framework boundary (Figure 3). Additionally, life cycle cost (LCC) estimation is integrated to assess user costs during vehicle usage, and environmental social costs are calculated using pollutant unit costs and emissions.

3.1. Assessment Objectives and Scope

3.1.1. Assessment Objectives

This paper focuses on five types of automobiles: internal combustion engine vehicles (ICEVs), battery electric vehicles (BEVs), plug-in hybrid electric vehicles (PHEVs), hybrid electric vehicles (HEVs), and fuel cell vehicles (FCVs). Specifically:

(1)

ICEVs use an internal combustion engine to drive wheels.

(2)

BEVs operate solely on rechargeable batteries.

(3)

PHEVs share principles with BEVs but include engines.

(4)

HEVs combine traditional ICEV engines with electric motors and batteries.

(5)

FCVs use fuel cells to power electric motors.

For BEVs, ICEVs, and PHEVs, we draw on the vehicle parameter settings reported in Zhang et al. [38], which provide class-specific sedan characteristics, including but not limited to vehicle mass, fuel consumption, and power consumption (kWh/100 km). These parameters are combined with class-level sales shares obtained from the China Passenger Car Association (CPCA) to construct sales-weighted average characteristics for each vehicle type. For FCVs and HEVs, average vehicle parameters are adopted from the GREET model [39]. Detailed vehicle parameters are summarized in

Table 1. Parameters of five vehicle categories.

	ICEVs	BEVs	HEVs	PHEVs	FCVs
Energy type	Gasoline	Pure electric	Non-plug-in hybrid	Plug-in hybrid	Hydrogen fuel
Curb weight (kg)	1367.6	1647.3	1555.4	1759.6	1652.9
Fuel consumption (L/100 km)	6.0	None	5.3	5.0	None
Power consumption (kWh/100 km)	None	13.5	None	17.6	None
Battery capacity (kWh)	None	57.3	None	11.9	None
Battery type	None	Lithium-ion battery	NiMH battery	Lithium-ion battery	Lithium-ion battery
Life cycle mileage (km)	300,000	300,000	300,000	300,000	300,000
Electric driving range (km)	None	418.5	None	67.2	None
Official Price ($)	32,872	33,852	29,372	30,072	104,720

Note: Vehicle prices are proxied by the average market price for each vehicle category. Price data were obtained via web scraping from https://www.pcauto.com.cn/ (accessed on 10 March 2026). Official prices were converted into U.S. dollars using the December 2025 exchange rate (1 CNY = 0.14 USD), and the costs and prices below are all converted using this exchange rate. The life-cycle mileage is assumed to be 300,000 km, consistent with typical vehicle use in China, and sensitivity analyses on lifetime mileage are conducted in subsequent sections.

.

3.1.2. System Boundary

In the full-chain life cycle assessment of new energy vehicles and ICEVs, this study divides the life cycle of vehicles into three phases: production phase, operation phase, and end-of-life recycling phase. Energy inputs in the life-cycle inventory include electricity and major primary energy sources (e.g., coal, crude oil, and natural gas), while the emissions inventory covers key greenhouse gas and air pollutant emissions, including CO₂, CO, VOCs, NO_x, PM_2.5, SO_x, CH₄, and N₂O.

In the first phase, vehicle production includes material production, parts manufacturing, vehicle assembly, and distribution processes. Electric vehicle batteries are considered as part of vehicle assembly, and the battery production process is included in the vehicle production stage. In the second phase, the process of vehicle operation includes energy consumption and vehicle maintenance/repair processes. The final phase is the end-of-life stage of the vehicle, which includes recycling, disposal and reuse. The full-chain life cycle system boundary of vehicles is shown in Figure 4.

3.1.3. Functional Unit

In life-cycle assessment, the functional unit serves as the basis for ensuring the comparability of results across different studies. Recent literature adopts heterogeneous assumptions regarding vehicle life-cycle mileage. For example, Qiao et al. [40] assumed a 12-year lifetime with a total mileage of 200,000 km for both an A-segment LDEV and a comparable ICEV. Huang et al. [41] analyzed L-category LDPVs and assumed life-cycle vehicle-kilometers traveled (VKT) of 273,588 km for ICEVs and 193,121 km for BEVs. Wei et al. [42] focused on medium-sized LDPVs with a life-cycle VKT of 150,000 km and a vehicle age of 15 years.

Given the substantial variation in assumed life-cycle mileage across vehicle types and studies, this study defines the functional unit as a vehicle driven 300,000 km under typical Chinese driving conditions, with a vehicle lifetime of 15 years. This falls within China’s compulsory scrapping range (0–600,000 km depending on operating conditions) [43], and its robustness is tested across 50,000–550,000 km in Section 5.2. Real-world utility factors for private PHEVs in China are around 26% [44]. Accordingly, this study adopts a baseline electric driving share of 30% to reflect current usage patterns, and tests the sensitivity of emission outcomes across a range of 20–70% in Section 5.2. The charging efficiency of BEVs and PHEVs is assumed to be 95%, consistent with current average levels [45].

3.2. Inventory Analysis and Model Construction

In order to construct the full-chain life cycle assessment model, this study firstly established the mathematical models of three main stages of the vehicle life cycle, then collected the required data from the GREET database for inventory analysis, and then quantitatively calculated and analyzed the CO₂ and air pollutant emissions at each stage of the vehicle life cycle. The full life cycle mathematical model is shown in the formula:

$$E_{LCA}=E_m+E_o+E_r$$

(1)

$$P_{LCA}=P_m+P_o+P_r$$

(2)

$E_{LCA}$ and $P_{LCA}$ are energy consumption and emissions of the whole life cycle of vehicles. $E_m$ and $P_m$, $E_{\mathrm{o}}$ and $P_o$, $E_r$ and $P_r$ represent energy consumption and GHG emissions at the production, use and recycling phases respectively.

3.2.1. Vehicle Production

The vehicle production process includes material production and processing, parts manufacturing, battery and other accessories production, vehicle assembly, vehicle distribution [12]. For convenience, this study assumes that all vehicle production processes occur in China. Material production and processing refers to the entire process from ore mining and extraction to molding. Vehicle assembly includes the collection of basic components and pre-treatment techniques such as stamping, welding and painting. Vehicle distribution is the transportation process of automobiles. When estimating CO₂ emissions and other polluting gas emissions, this paper considers all energy-related processes.

(1) Material production and parts manufacturing

Due to varying emission factors among materials, the material composition of vehicles is crucial for calculating CO₂ emissions from material production. Power battery weights for new energy vehicles were obtained from the GREET database. The vehicle weight excluding the power battery were then calculated by subtracting power battery weight from total vehicle mass (

Table 2. Weight of vehicles and weight excluding power batteries.

	ICEVs	BEVs	HEVs	PHEVs	FCVs
Curb weight (kg)	1367.6	1647.3	1555.4	1759.6	1652.9
Weight excluding power battery (kg)	1367.6	1277.9	1515.0	1642.6	1638.9
Power battery weight (kg)	--	369.4	40.4	117.0	14.0

). Based on component-level material composition shares provided by GREET, we further derived the material weight composition for each of the five vehicle types using the vehicle weight excluding the power battery (

Table 3. Vehicle material weight composition.

Weight Composition (kg)	ICEVs	BEVs	HEVs	PHEVs	FCVs
Vehicle Weight Excluding Battery Power Battery	1367.6	1277.9	1515.0	1642.6	1638.9
Steel	821.5	811.0	923.2	997.5	953.3
Cast Iron	32.1	0.0	37.5	33.2	0.1
Cast Aluminum	104.8	101.5	147.6	172.9	130.1
Wrought Aluminum	68.3	32.2	41.0	44.6	56.2
Copper	30.1	40.5	62.8	77.4	44.3
Glass	205.6	187.5	194.0	200.2	214.4
Average Plastic	54.5	49.4	53.3	54.1	63.0
Rubber	17.7	21.1	20.5	26.2	142.7
Others	32.1	0.0	37.5	33.2	0.1

).

Emissions of CO₂ and other pollutants from the material production and transformation process can be divided into direct emissions and emissions from the source process of the energy consumed. The direct emissions from this process can be obtained by multiplying the weight of the vehicle material by the emission factor of the material production and transformation process. The indirect emissions of the material production and conversion process can be calculated by combining the different energy consumption of the vehicle materials with the emission factors of the corresponding energy sources in the production process. The emission factors of CO₂ and other pollutants for the material production and transformation processes (

Table 4. Emission factors for GHG and other pollutants in material production and transformation process.

	CO2 (kg/kg)	VOC (g/kg)	CO (g/kg)	NOx (g/kg)	PM2.5 (g/kg)	SOx (g/kg)	CH4 (g/kg)	N2O (g/kg)
steel	2.26	2.60	18.84	2.44	0.69	9.23	3.77	0.02
cast iron	0.43	2.10	0.81	1.07	0.47	3.67	3.39	0.01
cast aluminum	2.87	0.41	1.38	2.78	0.56	6.88	5.34	0.05
wrought aluminum	15.21	1.52	5.47	12.61	2.21	29.38	24.99	0.31
copper	3.13	0.96	2.58	3.89	0.49	4.27	5.62	1.00
glass	1.57	0.24	1.05	2.20	0.10	0.89	3.28	0.04
average plastic	2.89	1.01	5.40	4.30	0.30	9.45	16.30	0.24
rubber	3.56	6.30	2.30	4.88	0.41	2.71	6.75	0.09
others	2.31	0.68	4.17	3.84	0.37	7.48	10.05	1.15

Note: Other materials include lead, sulfuric acid, water, brass, zinc, PVC, paint, nylon, ABS, polyurethane, etc.

) and energy consumption (

Table 5. Energy consumption of material production and transformation processes (MJ/kg).

Material	Coal	Electricity	Natural Gas	Coke	Crude Oil	Gasoline	Diesel
steel	14.858	4.702	3.200	0.052	0.000	0.000	0.000
cast iron	0.000	0.737	0.000	25.409	0.000	0.000	1.448
cast aluminum	16.288	117.426	21.413	12.954	0.000	0.000	0.543
wrought aluminum	8.659	58.823	17.856	4.626	0.000	0.000	0.221
copper	0.000	19.674	0.063	0.000	0.000	0.002	4.783
glass	0.000	2.603	14.284	0.000	0.000	0.000	0.000
average plastic	0.071	3.249	7.620	0.000	0.726	0.003	0.031
rubber	0.000	2.984	23.880	0.000	0.000	0.000	0.000
others	15.940	5.175	10.655	1.845	3.602	0.008	1.229

) were obtained from the GREET model of Argonne National Laboratory [39].

The emission factors for different energy production processes are reported in

Table 6. GHG and other pollutant emission factors in energy production processes.

Energy	CO2 (g/MJ)	VOC (mg/MJ)	CO (mg/MJ)	NOx (mg/MJ)	PM2.5 (mg/MJ)	SOx (mg/MJ)	CH4 (mg/MJ)	N2O (mg/MJ)
coal	2.33	7.15	3.34	8.45	1.22	6.48	141.00	0.04
electricity	155.60	15.83	67.38	133.24	12.52	171.68	65.39	4.21
natural gas	15.45	12.33	39.28	65.92	1.83	20.84	658.00	1.20
coke	9.25	82.49	21.53	33.81	18.44	144.10	123.20	0.13
crude oil	8.77	12.27	18.14	22.58	0.66	18.80	120.50	0.17
gasoline	15.81	25.85	13.44	20.07	1.51	5.47	106.50	0.28
diesel	11.96	7.07	11.59	17.29	1.02	4.38	101.50	0.22

. The electricity CO₂ emission factor is calculated as a weighted average based on China’s 2024 power generation mix, using emission factors from the Database of National Greenhouse Gas Emission Factors of China and generation shares from the National Energy Administration. Non-CO₂ electricity-related emission factors and those for other energy carriers are sourced from the GREET database.

(2) Battery production

BEVs and PHEVs typically use lithium-ion batteries, while nickel-metal hydride (Ni-MH) batteries are often preferred to power HEVs due to their relatively low cost. We used the GREET model to obtain the CO₂ emissions of 1 kg battery production, and then calculated the GHG emissions of the battery production process combined with the weight of the battery.

(3) Vehicle assembly

The vehicle assembly process consists of six parts: paint production and painting, heating, ventilation, air conditioning (HVAC) and lighting, material handling, heating, air compression, and welding. The detailed energy consumption of each process in vehicle assembly is reported in

Table 7. Energy consumption of parts and vehicle assembly processes.

Energy Consumption (MJ/Vehicle)	Paint Spraying	HVAC and Lighting	Material Handling	Heating	Air Compression	Welding	Lithium-Ion Battery Assembly	Lead–Acid Battery Assembly
Electricity	302	290	60	0	80	120	162	15
Natural Gas	2425	0	0	3143	0	0	95	25

[46]. Based on these data, the total energy inputs for the vehicle assembly stage amount to 1029 MJ of electricity and 5688 MJ of natural gas per vehicle.

(4) Vehicle distribution

Emissions of the vehicle distribution phase mainly come from fuel consumption during transport [12]. Distance, vehicle mass, energy consumption factor and emission factor determine the emissions due to fuel consumption. Vehicle manufacturers in China are distributed all over the country, and it is almost impossible to calculate the delivery distance of each vehicle. Considering Guangdong and Beijing are the main vehicle production cities in China, this paper chooses Guangdong as the delivery point for ICEVs, PHEVs, and HEVs, and Beijing as the delivery point for BEVs and FCVs. Vehicle demand will be influenced by regional GDP, provinces with high GDP consume more vehicles than provinces with low GDP.

Therefore, this paper employs the GDP of 31 provinces in mainland China in 2024 as the weight of distribution distance and assumes that the means of transportation in the distribution process is diesel trucks. CO₂ and other pollutant emissions of distribution process are calculated as follows:

$$E={\displaystyle \sum _{i=1}^{31}G}D{P}_i\times D\times M\times \alpha \times EF$$

(3)

$E$ denotes gas emissions ($\mathrm{kg}$), $GD{P}_i$ denotes the GDP share of province $i$, $D$ denotes distribution distance ($\mathrm{km}$), $M$ is vehicle mass ($\mathrm{kg}$), $\alpha$ is energy consumption factor ($\mathrm{kg}\cdot {\mathrm{kJ}}^{-1}{\mathrm{km}}^{-1}$), $EF$ is gas emission factor ($\mathrm{kg}\cdot {\mathrm{kJ}}^{-1}$), and the value of $\alpha$ is set as 0.6 according to Wang et al. [47].

3.2.2. Use Phase

The vehicle use phase comprises operation and maintenance. Operation consumes energy in the form of oil and electricity, while maintenance involves parts replacement and emissions from parts manufacturing. Notably, we account for both energy consumption and all production-related emissions.

(1) Vehicle operation

In the vehicle operation phase, ICEVs and BEVs run on gasoline and electricity respectively. For PHEVs, emissions during operation involve gasoline consumption emissions and electricity consumption emissions, assuming that gasoline consumption and electricity consumption mileage each account for 50% of total life cycle mileage, and that the electricity consumed is charged from charging piles. Only gasoline consumption emissions are involved in HEVs use.

$$E_{ICEV}=C_g\times LM\times WT{W}_g$$

(4)

$$E_{BEV}={\displaystyle \frac{C_e\times LM\times WT{W}_e}{e(1-r)}}$$

(5)

$$E_{PHEV}=\left({\displaystyle \frac{\alpha \times C_{ep}\times WT{W}_e}{e(1-r)}}+(1-\alpha )\times C_{gp}\times WT{W}_g\right)\times LM$$

(6)

$$E_{HEV}=0.7E_{ICEV}$$

(7)

$E_{ICEV}$, $E_{BEV}$, $E_{\mathrm{PH}EV}$, and $E_{HEV}$ are the emissions of ICEVs, BEVs, PHEVs, and HEVs in the vehicle operation phase, respectively; $C_g$, $C_{\mathrm{e}}$ are gasoline consumption per 100 km for ICEVs and electricity consumption per 100 km for BEVs; $C_{gp}$, $C_{ep}$ are gasoline consumption and electricity consumption per 100 km for PHEVs; $e$ is charging efficiency; $r$ denotes transmission and distribution losses, with a transmission loss rate of 4.37% based on data from the National Energy Administration of China [48]; $WT{W}_g$, $WT{W}_e$ represents gasoline and electricity emission factors, and $LM$ is life cycle mileage. $\alpha$ denotes the share of kilometers driven in electric mode for PHEVs, set at 30% in the baseline scenario [44]; its sensitivity is examined in a subsequent section. For HEVs, fuel consumption is assumed to be 30% lower than that of comparable ICEVs [49].

During the vehicle use phase, gasoline WTW emissions are calculated by accounting for both fuel production and combustion processes. The well-to-wheel emission factor of gasoline is obtained from the GREET database [39]. To derive electricity-related emission factors, this study accounts for China’s electricity generation mix and calculates average emission factors as weighted sums of source-specific emission factors and their corresponding generation shares in 2024, as reported in

Table 8. Electricity mix and emission factors of different power sources in China.

Power Source	Proportion in China Grid	CO2 (g/kWh)	VOC (mg/kWh)	CO (mg/kWh)	NOx (mg/kWh)	PM2.5 (mg/kWh)	SOx (mg/kWh)	CH4 (mg/kWh)	N2O (mg/kWh)
Coal	57.8%	924.0	86.1	315.1	762.4	71.3	1009.0	315.1	23.3
Natural Gas	3.2%	450.3	75.2	254.4	332.9	23.0	86.7	1117.4	7.2
Nuclear	4.4%	6.5	0.4	2.1	3.6	0.2	2.1	4.5	0.2
Hydro	14.2%	14.1	2.5	23.8	13.7	1.2	15.1	20.6	0.1
Biomass	2.1%	40.4	40.5	1249.2	761.6	73.0	663.3	149.9	60.5
Wind	9.9%	32.4	6.2	37.2	10.6	2.0	20.5	14.7	0.2
Solar	8.3%	52.0	33.5	220.1	64.1	11.7	138.7	105.0	1.8
Average		559.1	56.8	241.9	475.8	44.8	615.7	234.3	15.1

Note: A transmission loss rate of 4.37% and the shares of power generation sources are based on data from the National Energy Administration of China and the Low-Carbon Power database. CO₂ emission factors are sourced from the National Greenhouse Gas Emission Factor Database of China, while other pollutant emission factors are taken from the GREET database [39].

. CO₂ emission factors are obtained from the Database of National Greenhouse Gas Emission Factors of China, while generation shares are based on data published by the National Energy Administration of China. Emission factors for non-CO₂ pollutants are adopted from the GREET database.

CO₂ emissions of FCVs during vehicle operation are zero, and all pollutant emissions come from the WTT (Well-to-Tank) phase of hydrogen fuel. This paper considered five sources of hydrogen production: electrolytic water, coke oven gas, biomass, coal and natural gas, and assume that the representative model of FCVs consumes 1.04 kg of hydrogen per 100 km [50]. Then we obtained the emissions of hydrogen fuel in its WTT phase according to the emissions of per-unit hydrogen fuel from the GREET database, as shown in

Table 9. Emissions of 1 kg hydrogen fuel in its WTT phase.

Hydrogen Source	CO2 (kg)	VOC (g)	CO (g)	NOx (g)	PM2.5 (g)	SOx (g)	CH4 (g)	N2O (g)
Coke Oven Gas	3.65	4.95	1.99	1.72	2.98	6.97	22.88	0.03
Biomass	1.36	0.54	2.31	3.86	0.21	2.40	2.45	0.88
Coal	1.96	1.76	1.09	1.23	0.25	3.01	29.02	0.01
Natural Gas	8.46	1.56	4.06	4.86	0.23	1.46	23.83	0.20
Electrolytic Water	27.72	3.09	15.49	30.88	2.45	28.53	47.73	0.78

. In China, fossil fuel-based hydrogen production still dominates the hydrogen supply. In 2024, coal-based hydrogen accounted for 56.7% of total hydrogen production [51]. Therefore, this study adopts coal-based hydrogen as the baseline hydrogen source and considers the impacts of alternative hydrogen production pathways in subsequent scenario analyses.

(2) Maintenance and repair

In the maintenance and repair phase, this paper considers the replacement frequency of liquids, tires, and other parts of different types of vehicles. The use mileages of powertrain coolant, transmission oil, brake fluid, and windshield oil are 30,000 km, 60,000 km, 20,000 km and 12,500 km respectively, the replacement cycle of lead–acid batteries is assumed to be 3 years, and the mileage of tires is about 80,000 km [43]. This study assumes that the mileage of the vehicle life cycle is 300,000 km. The number of replacements of parts such as liquids and tires during the life cycle and the corresponding CO₂ and other pollutant emission factors for one replacement are shown in

Table 10. Number of replacements and emissions of components (per kg).

Parts	Number of Replacements	CO2 (kg)	VOC (g)	CO (g)	NOx (g)	PM2.5 (g)	SOx (g)	CH4 (g)	N2O (g)
Powertrain coolant	10	0.504	0.275	0.561	0.757	0.038	0.259	2.594	0.013
Transmission fluid	5	3.102	1.227	1.043	4.122	0.367	7.531	4.669	0.061
Brake fluid	15	3.102	1.227	1.043	4.122	0.367	7.531	4.669	0.061
Windshield fluid	24	0.181	500.2	0.246	0.293	0.016	0.169	1.669	0.003
Lead–acid battery	5	0.656	0.384	0.401	0.913	0.503	8.502	2.631	0.012
Tire	3	3.128	5.057	7.818	4.071	0.500	4.871	5.782	0.068

.

3.2.3. End-of-Life Phase

The end-of-life (EOL) phase of vehicles includes disassembly, shredding, and the battery recovery process. Electricity is the main energy source in the EOL phase. Therefore, at this stage, the emission of electricity consumption is mainly considered, and it’s determined by the electricity consumption factor (the electricity consumed by unit mass of vehicle), vehicle mass and electricity emission factor. The calculation formula is as follows:

$$E_{vp}=M_v\times \alpha \times E{F}_{vp}$$

(8)

where $E_{vp}$ indicates CO₂ or other pollutant emissions ($\mathrm{kg}$), $M_v$ indicates vehicle mass ($\mathrm{kg}$), $\alpha$ and ${EF}_{vp}$ indicate electricity consumption factor ($\mathrm{kJ}/\mathrm{kg}$) and emission factor ($\mathrm{kg}/\mathrm{kJ}$), respectively. $\alpha$ is set to 370 [47].

In this study, the CO₂ emission factor for lithium iron phosphate (LFP) batteries recovered via hydrometallurgical processes is set to 2.8 kg CO₂ per kg of battery, and additional emissions from natural gas consumption during wet recovery (0.034 m³ per kWh) are also included [52]. Based on these emission factors, together with battery weight and capacity data, CO₂ emissions from the power battery recovery stage are estimated for BEVs, PHEVs, and FCVs. For HEVs, battery recovery emissions are estimated using nickel–metal hydride cathode recovery emission factors from the GREET database and HEV battery weights.

3.3. Life Cycle Cost Model

Life cycle cost (LCC) refers to all costs incurred during the effective use of a product. This paper compares and analyzes the economic costs of vehicles throughout their life cycle from the perspective of users. Assuming that consumers purchase a vehicle in 2021, the discounted value of costs incurred during the full life cycle of this vehicle is calculated by constructing a full life cycle cost model with the following formula:

$$TC=C_p+C_o-SV$$

(9)

$TC$ represents the discounted value of the total life cycle cost of the vehicle; $C_p$ is purchase cost, which refers to the sum of one-time expenses involved in the purchase of the vehicle; $C_{\mathrm{o}}$ stands for operation cost, which is the discounted value of the sum of energy costs, maintenance costs, taxes, insurance, and other expenses incurred during the use of the vehicle; $SV$ represents the discounted salvage value of the vehicle after scrapping.

1. Acquisition cost $C_p$

Acquisition cost refers to all the expenses that consumers must pay in order to purchase a car, including the manufacturer’s suggested retail price, purchase tax, and licensing fee. The calculation formula is as follows:

$$\begin{array}{c}{C}_p=SRP+PT+LF\\ =SRP+{\displaystyle \frac{SRP}{1+17\%}}\times r+LF\end{array}$$

(10)

The suggested retail price (SRP) is the fee paid to the manufacturer. The purchase tax (Purchase Tax, PT) is related to the price of the vehicle, and is usually calculated according to a certain percentage after deducting value-added tax (17%) from the purchase cost of the vehicle.

When purchasing a vehicle with a displacement greater than 1.6 L, the purchase tax is paid at a rate of 10%; when the displacement is 1.6 L and below, the purchase tax rate is 7.5%. License Fee (LF) is usually calculated based on the one-stop service fee provided by the merchant. Owners of new energy vehicles can benefit from the vehicle purchase tax exemption policy, so the purchase tax of new energy vehicles is 0. In addition, new energy vehicles are exempt from vehicle and vessel tax and consumption tax, and the government still supports road permits and license plate indicators.

2. Operation cost $C_o$

Operation cost refers to the necessary expenses, such as energy, tax, and insurance paid, by consumers to ensure the normal operation of the vehicle, including energy cost, maintenance cost, tax and insurance fees, and other costs such as parking fees, road tolls, etc. Other costs will not differ significantly due to different vehicle types, so these costs will not be considered in this study. The calculation formula is as follows:

$$C_o={\displaystyle {\displaystyle \sum _{i=1}^n}\frac{EC+TIF}{{(1+k)}^{i-1}}}+{\displaystyle {\displaystyle \sum _{i=1}^n}\frac{M{C}_i}{{(1+k)}^i}}$$

(11)

where $EC$ denotes the annual energy cost; $TIF$ denotes the annual vehicle use tax and insurance cost; $M{C}_i$ denotes the maintenance cost in year $i$; and $k$ denotes the discount rate, which is set to 6% in this paper.

(1) Energy cost

$$EC={\displaystyle \frac{C\times P\times T}{e}}$$

(12)

where $C$ indicates energy consumption per kilometer traveled (electricity or gasoline consumption); $P$ indicates the price per unit of energy (electricity or gasoline price); $T$ indicates the annual mileage; $e$ indicates the charging efficiency for BEVs and PHEVs, and it takes a value of 1 for ICEVs and HEVs. The power consumption and fuel consumption mileage of PHEVs each account for 50% of the life cycle mileage, and the energy cost is also calculated according to the proportion of power and gasoline consumed. The 2024 annual average price of China’s No. 92 gasoline is $0.94/L, based on data from the Eastmoney database (https://data.eastmoney.com/cjsj/oil_default.html; accessed on 24 March 2026). Specifically, the reported average price is 8981.1 CNY/ton, which is converted using the 2024 exchange rate (1 CNY = 0.14 USD) and the gasoline density from GREET (1 L = 0.74 kg)). For BEVs, this paper obtains the number of charging times based on the full life cycle mileage and pure electric range of BEVs. The single charging cost is $0.21/kWh for DC charging, including electricity and service fees, based on data from the State Grid, with a charging efficiency of 93%. On the basis of the above, we can calculate the charging cost of BEVs during their full life cycle.

For FCVs, $C$ represents the hydrogen fuel consumption of a hydrogen fuel vehicle driving 1 km, $P$ represents the price of a unit of hydrogen fuel, and $T$ denotes the annual mileage of FCVs. The energy consumption of FCVs is about 1 kg per 100 km. According to the calibration price of hydrogen fuel, the basic travel cost per 100 km is about $8.4/kg. Considering that there is a certain subsidy for hydrogen energy, the price of hydrogen after subsidy is about $5.6/kg.

(2) Maintenance cost

Maintenance costs are what must be spent to keep the vehicle running properly, including the cost of replacing the battery and other maintenance costs.

$${MC}_i={BRC}_i+{OMC}_i$$

(13)

where, ${MC}_i$ denotes maintenance cost in year $i$; ${BRC}_i$ denotes battery replacement cost in year $i$; ${OMC}_i$ denotes other maintenance cost in year $i$. Power battery is one of the key components of electric vehicles. With the use of electric vehicles and the continuous charging and discharging of batteries, the batteries will gradually degrade. We assume that the battery life of BEVs is 150,000 km. Other maintenance costs mainly include maintenance costs, parts replacement, repair costs, etc. In addition, the cost of daily maintenance and repair will gradually increase as the car’s parts age over time. The calculation formula is:

$$\begin{array}{c}{OMC}_i=C_i+{NC}_i\\ =C_i+10\%\times {\displaystyle \sum _{j=1}^{i-1}{C}_j}\end{array}$$

(14)

$C_i$ represents the fixed maintenance cost in year $i$; $N{C}_i$ represents the maintenance cost that increases with the age of the vehicle. Fixed maintenance cost $C_i$ is related to the vehicle grade, and the higher the vehicle grade, the higher the vehicle maintenance cost. This paper assumes that the fixed maintenance costs of BEVs and FCVs are $30.8 and $70 respectively, and the fixed maintenance costs of ICEVs, HEVs, and PHEVs are $140.

(3) Taxes and insurance premiums

Taxes and insurance fees mainly include vehicle and vessel tax, compulsory liability insurance for motor vehicle traffic accidents and commercial insurance fees. To encourage the promotion of new energy vehicles, China has exempted new energy vehicles from vehicle and vessel tax.

$$TIF=V+S+CI$$

(15)

where $TIF$ denotes annual vehicle use tax and insurance premium; $V$ denotes annual vehicle and vessel tax; $S$ denotes compulsory traffic insurance premium; $CI$ denotes commercial insurance premium.

According to the average situation in China, this paper assumes that traffic compulsory traffic insurance premium is $133, commercial insurance is $224, including $140 for vehicle damage insurance and $84 for third-party liability insurance, and the vehicle and vessel tax of ICEVs and HEVs is $58.8 and $29.4 respectively.

3. Disposal cost $SV$

After cars have been used for a certain number of years, auto parts can be recycled, but a systematic recycling mechanism has not been formed in China. Therefore, this paper assumes that the vehicle is resold at the end of its life cycle, and the residual value is obtained through the annual depreciation rate $r$. The service life of the vehicle is 15 years, and the double declining balance method is used in calculating depreciation. The residual value of the vehicle at resale is as follows:

$$SV={\displaystyle \frac{{(1-r)}^n\times SRP+C_B}{{(1+k)}^n}}$$

(16)

where, $SV$ is the residual value of vehicle disposal; $SRP$ is the manufacturer’s suggested retail price; $C_B$ is the battery recycling price, which is $1120 according to Zhou et al. [53] when calculating the disposal cost of BEVs, PHEVs, and FCVs, and 0 when calculating the cost of ICEVs and HEVs; $r$ is the annual depreciation rate, which is set to 20%; $k$ indicates the discount rate, which is uniformly set at 6% in this paper.

3.4. Environmental Pollution Cost Model

We calculate the total social cost of environmental pollution with the help of unit cost and corresponding emissions of pollutants. The calculation formula is as follows:

$$C_p={\displaystyle \sum e_i}{d}_i$$

(17)

where $C_p$ is the total cost of pollutant emissions; $e_i$ is the emission of pollutants; and $d_i$ is the unit emission cost of the corresponding pollutant. The unit emission cost of pollutants is calculated as follows:

$$d_i={\displaystyle \frac{d_i\prime }{1000}}\times {\displaystyle \frac{CP{I}_{2022}}{CP{I}_{2000}}}\times R$$

(18)

where $d_i$ and $d_i\prime$ are the unit emission cost of a pollutant in China in 2022 and the unit emission cost of the corresponding pollutant in Europe in 2000 [54]. And R represents the corresponding annual average exchange rate in 2000. China’s consumer price index (CPI) for 2024 and 2000 is calculated based on the price in 1978. By converting $d_i\prime$ into $d_i$ through Formula (18), we can get the unit cost of pollution emissions in China in 2024, as shown in

Table 11. Calculation of the unit emission cost of pollutants (data source: [54]).

Pollutants	Unit Cost in Europe in 2000 di′ (€/t)	Unit Cost in China in 2024 (¥/kg)	Unit Cost in China in 2024 ($/kg) di
CO2	21	0.26	0.04
CH4	480	5.99	0.84
NO2	6200	77.32	10.83
NOx	6700	83.56	11.70
SOx	6500	81.07	11.35
CO	62	0.77	0.11
VOC	870	10.85	1.52
PM2.5	63,000	785.72	110.00
PM10	7400	92.29	12.92

.

4. Results

4.1. Emissions Outcome

The results of CO₂ emissions and emission reduction rates for the five vehicle types are presented in Figure 5. The full-chain life-cycle CO₂ emissions of an internal combustion engine vehicle (ICEV) are estimated at 63.9 t CO₂, compared with 35.6 t, 48.1 t, and 53.2 t CO₂ for battery electric vehicles (BEVs), hybrid electric vehicles (HEVs), and plug-in hybrid electric vehicles (PHEVs), respectively. Correspondingly, BEVs achieve an emission reduction rate of 44.3% relative to ICEVs, while the reduction rates for HEVs and PHEVs are more modest, at 24.8% and 16.8%, respectively. For fuel cell vehicles (FCVs), when coal-based hydrogen is used as the fuel source, total emissions are the lowest among all vehicle types, with an emission reduction rate reaching 75.2%.

Figure 6 presents the full-chain lifecycle pollutant emissions for five vehicle types, with values normalized to ICEV levels. The results show that ICEVs generate the highest levels of VOCs, CO, and N₂O. In contrast, CH₄ emissions peak in FCVs, while BEVs and PHEVs yield higher PM_2.5 and SO_X than ICEVs. Specifically, CO emissions from ICEVs are approximately seven times those of BEVs, and CH₄ emissions from FCVs are more than double the levels observed for BEVs.

4.2. Cost Comparison

The vehicle selling price is taken into account when calculating the full life cycle cost of vehicles, but the price fluctuates greatly and is affected by various factors. Therefore, the selling price is excluded from the full life cycle economic cost of vehicles for comparison, as shown in Figure 7. Compared with ICEVs, BEVs, HEVs, PHEVs, and FCVs, all exhibit lower life-cycle costs, with FCVs being the most cost-effective, followed by PHEVs. A similar pattern is observed for pollution-related costs in Figure 8, where all alternative vehicle types incur lower pollution costs than ICEVs; among them, FCVs have the lowest cost, followed by BEVs.

4.3. Health Impacts

In this section, we evaluate and compare the public health impacts of different vehicle types based on the PM_2.5 emission results presented in Section 4.1. We first report the baseline mortality attributable to transportation-sector air pollution in China in

Table 12. Baseline mortality attributable to transportation-sector air pollution in China.

Parameter	Value	Source
Transportation-sector PM2.5-attributable deaths (2015, thousand/year) and share of total mortality burden (2015, %)	218.3 (10%)	[56]
Ambient PM2.5-attributable deaths, China (2021, thousands/year)	1860	[55]
Transportation-sector PM2.5-attributable deaths as a share of total PM2.5-attributable mortality (2010–2049, %)	14–23%	[57]
Estimated transportation PM2.5 deaths (2021, thousands/year)	223.2	This study
NOx-mediated fraction in traffic PM2.5 (nitrate)	55%	[58]
Transportation NOx-mediated PM2.5 deaths (thousands/year)	122.8	This study

. Transportation PM_2.5-attributable deaths in 2021 are estimated by applying a sectoral contribution share to total ambient PM_2.5-attributable deaths in China (1860 thousand/year) [55]. The transportation share is set at 12%, a conservative midpoint between 10% reported for 2015 [56] and 14–23% for 2010–2049 [57], yielding 223.2 thousand deaths per year (1860 × 12%). Assuming a 55% NO_x-mediated (nitrate) fraction in traffic-related PM_2.5 [58], NO_x-mediated deaths are estimated at 122.8 thousand per year (223.2 × 55%).

Building on this baseline, we derive vehicle-specific PM_2.5 emission contribution ratios from Figure 6 and apply these ratios to the 2021 transportation-sector PM_2.5- and NO_x-attributable deaths reported in Table 12 to estimate attributable mortality for each vehicle type.

Table 13. Public health impact assessment by vehicle type.

	ICEVs	BEVs	HEVs	PHEVs	FCVs
a. PM2.5 emission ratio (from Figure 6)
Relative to ICEV	1.00	1.09	0.88	1.12	0.70
b. Attributable deaths (thousands/year)
PM2.5-caused	46.56	50.97	40.93	51.94	32.81
NOx-mediated	25.61	28.04	22.52	28.58	18.05
c. Health benefit vs. ICEVs
Avoidable PM2.5 deaths (thousands/year)	—	−4.41	5.63	−5.38	13.75
Change (%)	—	−9.48%	12.09%	−11.56%	29.53%

Note: Panels a–c illustrate the stepwise calculation procedure in order. Attributable deaths are allocated proportionally based on each vehicle type’s PM_2.5 emission ratio (from Figure 6) relative to the sum of ratios across all five types.

shows clear heterogeneity across vehicle types. BEVs and PHEVs exhibit higher PM_2.5-attributable mortality than ICEVs (50.97 and 51.94 vs. 46.56 thousand/year), while HEVs and FCVs reduce impacts (40.93 and 32.81 thousand/year). A similar pattern holds for NO_x-mediated deaths. Relative to ICEVs, only HEVs and FCVs deliver net health benefits, avoiding 5.63 (12.09%) and 13.75 (29.53%) thousand deaths per year, respectively. In contrast, BEVs and PHEVs increase attributable mortality. These results highlight the importance of upstream emissions, indicating that the health benefits of electrification depend critically on power sector decarbonization.

5. Prediction and Scenario Analysis

Building on the vehicle-level life-cycle emissions results presented in Section 4, we extend the analysis to the sectoral level through scenario projection. Section 5.1 applies machine learning models to project the stock of different powertrain types in China by 2030, while Section 5.2 conducts sensitivity analyses on key LCA parameters. Based on these results, Section 5.3 constructs scenario matrices for the passenger vehicle sector to assess the emission reduction effects of increasing NEV penetration under future development pathways. These projections are exploratory rather than causal, illustrating conditional outcomes under specified assumptions rather than estimating policy impacts.

5.1. Forecasting Methods and Results

In addition to individual vehicle analysis, assessing industry-wide emissions is crucial. Socioeconomic shifts influence private vehicle demand, and climate policies like carbon peaking and neutrality affect ownership rates. Therefore, developing forecasting models is essential for analyzing future vehicle trends and scenario precision. We construct and compare prediction models using statistical (ARIMA) [59] and deep learning (LSTM, Bi-LSTM) approaches on a vehicle dataset [60]. Model performance is evaluated using MAE, RMSE, and R-squared metrics [61].

In LSTM and Bi-LSTM models, the number of epochs is experimentally determined to avoid overfitting and maintain predictive accuracy. We utilize 50 neurons, batch sizes of 1, 8, and 16, and epochs of 50, 100, and 200. Optimizers ‘adam’ and ‘rmsprop’ are employed to improve model fitting. Parameter optimization is conducted using a grid-search method [62], and the optimized parameters listed in

Table 14. Parameters of prediction models and their values.

Method	Parameters	Values
LSTM/Bi-LSTM	Layers	4
	Learning rate	0.001
	Optimizer	Adam/Rmsprop
	Batch size	1/8/16
	Epochs	50/100/200
	Time step	2
ARIMA	(p,d,q)	d = 1, p and q depend on data, use AIC and BIC to choose the best parameters.

are used for model construction, fitting, and prediction. All parameter optimizations were performed through trial and error.

This study evaluates the effectiveness of LSTM and Bi-LSTM models at different learning levels by adjusting the training ratios to 50%, 60%, 70%, and 80%, respectively. The optimal model is expected to yield a Mean Absolute Error (MAE) of zero, as mathematically expressed in Equation (19):

$$MAE={\displaystyle \frac{1}{M}}{\displaystyle \sum _{i=1}^M\left|C_{\mathrm{i}}-{\widehat{C}}_{\mathrm{i}}\right|}$$

(19)

Here, $C$ denotes the actual value and $\widehat{C}$ for estimated value.

The Root Mean Square Error (RMSE) is more sensitive to large errors or outliers, as a single significant prediction error can substantially impact the overall RMSE value. The RMSE is formally defined in Equation (20) as:

$$RMSE=\sqrt{{\displaystyle \frac{1}{M}}{\displaystyle \sum _{i=1}^M{\left(C_i-{\widehat{C}}_i\right)}^2}}$$

(20)

To illustrate the variance between the dependent and independent variables, the R-squared value is presented in Equation (21) as follows:

$$R\text{-}squared=1-{\displaystyle \frac{{\displaystyle \sum _{i=1}^M{\left({\widehat{C}}_i-C_i\right)}^2}}{{\displaystyle \sum _{i=1}^M{\left(\overline{C_i}-C_i\right)}^2}}}$$

(21)

Prediction errors are compared across different prediction techniques using bar charts for various performance metrics. Figure 9, Figure 10 and Figure 11 show the R-squared, RMSE, and MAE for car stock, BEV stock, and PHEV sales at different training percentages. Figure 12 provides the corresponding performance metrics for FCV sales.

The Bi-LSTM model achieves the highest R-squared value and the lowest RMSE and MAE for car stock prediction, outperforming the other two models (as shown in Figure 10, Figure 11 and Figure 12). For BEVs, the LSTM model exhibits consistently superior predictive performance. Although the Bi-LSTM model shows performance comparable to that of the LSTM model at the 50%, 60%, and 70% training ratios, its performance deteriorates substantially at the 80% training ratio. For PHEVs, the LSTM model demonstrates the best overall performance, although it is slightly inferior to the Bi-LSTM model at the 50% training ratio. Regarding FCVs, the LSTM model slightly outperforms the Bi-LSTM model in terms of RMSE and MAE and exhibits more stable R-squared values. In contrast, the Bi-LSTM model yields poor R-squared performance at higher training ratios.

Based on this analysis, the most appropriate model is selected for each time series and trained using historical data up to 2024 to forecast trends from 2025 to 2030. Specifically, Bi-LSTM models are applied to predict total vehicle stock, LSTM models are used for BEV stock, FCV sales, and PHEV sales. The corresponding forecast results are illustrated in Figure 13, Figure 14 and Figure 15.

We compute the average values under various training percentages after forecasting the future conditions of four sequences using deep learning models (LSTM and Bi-LSTM). Over the past several years, we have observed an increasing trend in the proportion of new energy vehicles, driven by technological advancements and policy support.

5.2. Scenario Analysis of Individual Vehicles

Section 5.2 presents a sensitivity analysis to examine the robustness of the results under key modeling assumptions. We test variations in energy and resource mixes, technical parameters, and usage patterns, including electricity mix, hydrogen pathways, vehicle energy consumption, vehicle weight, lifetime mileage, and the PHEV electric driving share. This analysis helps ensure the robustness of the vehicle-level emission results and provides a foundation for the sector-wide emission analysis presented in Section 5.3.

The baseline lifetime mileage of 300,000 km follows China’s compulsory vehicle end-of-life standards and typical private vehicle usage patterns [43]. This section’s second part demonstrates that BEVs maintain superior emission performance across all tested mileages from 50,000 km to 550,000 km, with the advantage becoming more pronounced at higher mileages due to the higher operational efficiency of electric powertrains. Even at the lower bound, BEVs still achieve meaningful reductions relative to ICEVs, indicating that the 300,000 km baseline does not artificially inflate the environmental case for electrification.

For PHEVs, given that real-world utility factors for private vehicles in China are around 26% [44], this study adopts 30% as the baseline assumption for the LCA analysis. The analysis in Part 6 of this section examines a range of 20–70%, indicating that the environmental performance of PHEVs is highly sensitive to consumer charging behavior, with direct implications for charging infrastructure deployment policies.

(a)

Electricity generation mix

In this section, we analyze vehicle emissions under different electricity generation mix scenarios in China. Projections of per-capita electricity generation by source for 2030, 2050, and 2060 are developed based on anticipated changes in installed capacity and 2024 baseline levels, from which the corresponding shares of electricity generation are derived [63]. It is assumed that low-carbon power sources (e.g., nuclear, wind, and solar) are assumed to expand over time, while coal-fired power declines but is not entirely phased out. To maintain power system reliability, the 2060 scenario adopts a low-carbon generation pathway with CCUS, under which a residual 3% of coal-fired capacity is retained and equipped with CCUS to provide stable and dispatchable electricity [64]. Based on these assumptions, four future power-structure scenarios are constructed (

Table 15. Scenario settings of power generation mixes.

Power Source	Base Scenario (2024)	Scenario 2 (2030)	Scenario 3 (2050)	Scenario 4 (2060)
Coal	57.8%	47.8%	10.2%	3.0%
Natural Gas	3.2%	3.4%	4.5%	4.4%
Nuclear	4.4%	5.8%	7.9%	10.0%
Hydro	14.2%	14.8%	14.7%	15.2%
Biomass	2.1%	2.3%	3.5%	3.7%
Wind	9.9%	12.8%	26.1%	29.9%
Solar	8.3%	13.2%	33.1%	33.7%

Note: The table presents the shares of per-capita electricity generation by source in China. The 2024 data are sourced from the National Energy Administration of China, while the figures for 2030, 2050, and 2060 are projections [63].

).

The full life cycle CO₂ emissions of vehicles are calculated in different electricity structure scenarios, and the results are shown in Figure 16. With the increase in the proportion of clean energy such as wind power and solar energy, CO₂ emissions decline across all vehicle types, while FCVs consistently exhibit the lowest emissions.

(b)

Life cycle mileage

Based on China’s “Compulsory Vehicle End-of-life Standard,” vehicle life cycle mileage varies between 0 and 600,000 km, depending on operating conditions, and influences full life cycle CO₂ and pollutant emissions. This study analyzed six life cycle mileages: 50,000 km, 150,000 km, 250,000 km, 350,000 km, 450,000 km, and 550,000 km. Figure 17 and Figure 18 present CO₂ emissions per kilometer and the corresponding reduction rates relative to ICEVs under different life-cycle mileage scenarios, revealing a clear ranking across vehicle types. ICEVs exhibit the highest emissions per kilometer, followed by PHEVs, HEVs, and BEVs, while FCVs show the lowest emissions.

(c)

Electricity consumption per 100 km

Electricity consumption per 100 km has a significant impact on the CO₂ emissions of BEVs and PHEVs during their operating phase, and varies depending on the vehicle model, season, and driving conditions during actual use. We changed the power consumption per 100 km of BEVs and PHEVs for scenario analysis. The CO₂ emission results of BEVs and PHEVs and the CO₂ emission reduction rate relative to ICEVs are shown in Figure 19. With the increase in electricity consumption per 100 km, the life-cycle CO₂ emissions of BEVs and PHEVs gradually increase, and the CO₂ emission reduction rate gradually decreases. BEVs consistently outperform PHEVs in terms of carbon emission reduction until electricity consumption reaches 30 kWh/100 km.

(d)

Vehicle weight

Lightweighting is a key strategy for energy saving and emission reduction in global automotive trends. By setting a unified weight for five vehicle types and adjusting it, we investigated the impact of weight on CO₂ emission reductions. Figure 20 and Figure 21 present the CO₂ emissions and reduction effects. Results indicate that as vehicle weight increases, the emission reduction effect of FCVs, PHEVs, HEVs, and BEVs decreases, with FCVs consistently outperforming other vehicles in reduction potential.

(e)

Hydrogen production method

China’s current hydrogen production methods encompass electrolytic water, natural gas reforming, coal gasification, coke oven gas extraction, and biomass gasification. This study examines the impact of five hydrogen production methods—electrolysis of water, coke oven gas, biomass, coal, and natural gas—on the life cycle CO₂ emissions of FCVs. Figure 22 presents the CO₂ emissions of FCVs using these methods and their reduction effects compared to ICEVs. In the baseline scenario, the hydrogen fuel for FCVs is assumed to be produced from coal. When electrolysis-based hydrogen is used instead, the emission reduction performance of FCVs declines substantially under China’s current power mix, with the reduction rate turning negative. By contrast, biomass-based hydrogen delivers the greatest mitigation effect, reducing FCV emissions by 78.2%.

(f)

Share of electric-mode mileages in PHEVs

Empirical evidence indicates that the real-world electric driving share (utility factor, UF) of PHEVs is substantially lower than type-approval values and varies widely across regions and use types. In particular, China exhibits relatively low real-world UF levels, at around 26% for private vehicles [44]. Based on this context, this study adopts a baseline assumption of 30% for the electric driving share to reflect China’s current usage patterns. A broader range of 20–70% is further considered in the scenario analysis to capture both conservative real-world behavior and potential improvements under favorable conditions. Figure 23 illustrates the emission outcomes of PHEVs under different shares of electric driving. As the electric driving share increases, CO₂ emissions gradually decline, accompanied by a corresponding rise in the emission reduction rate. When the electric driving share reaches 70%, the reduction rate of PHEVs increases to 26.4%, exceeding that of HEVs (24.8%) but remaining below that of BEVs (44.3%) and far lower than that of FCVs using coal-based hydrogen (75.2%).

5.3. Scenario Analysis for China’s Passenger Car Sector

In Section 5.1, forecasts indicate a higher proportion of new energy vehicles in 2030 compared to 2024, suggesting a decreased ICEV share in China. We explore three ambition scenarios—high, reference, and low—with higher ambition indicating greater NEV proportions. We also assess the need for traditional vehicle improvements beyond current NEV share increases, considering three corresponding improvement scenarios. Details of these scenarios are provided in

Table 16. Scenario matrix.

Scenarios		Baseline (2024)	2030 High	2030 Reference	2030 Low	2030 High and Improvement	2030 Reference and Improvement	2030 Low and Improvement
Higher NEV share	BEVs share (%)	6.26	12.02	10.93	9.84	12.02	10.93	9.84
	PHEVs share (%)	2.22	10.59	9.63	8.67	10.59	9.63	8.67
	FCVs share (%)	0.01	0.014	0.012	0.011	0.014	0.012	0.011
	ICEVs share (%)	91.1	59.40	66.00	72.60	59.40	66.00	72.60
	HEVs share (%)	0.41	17.97	13.43	8.88	17.97	13.43	8.88
Performance improvement of conventional vehicles	Gasoline emission factor (kg/L)	2.36	No improvement			1.49	1.66	1.83
	Gasoline consumption of ICEVs (L/100 km)	6.00	No improvement			3.87	4.30	4.73
	Gasoline consumption of HEVs (L/100 km)	5.30	No improvement			3.42	3.80	4.18
Total passenger car stock (million units)		353.00	432.21	480.23	528.25	432.21	480.23	528.25
Reference		MPS; CAAM; ANL [39]	[65,66]			[65,66,67]

Notes: Baseline vehicle ownership data for 2024 were obtained from the Ministry of Public Security of China (MPS) and the China Association of Automobile Manufacturers (CAAM) to project total vehicle stock and the shares of BEVs, PHEVs, and FCVs in 2030. ICEV stock shares in 2030 primarily follow forecasts from CICC Research [66], whose BEV–PHEV–HEV–ICEV classification is consistent with this study. Adjustment factors for average fuel consumption in 2030 relative to 2025 are sourced from the Ministry of Industry and Information Technology (MIIT) [65] and applied to estimate ICEV and HEV fuel consumption. Gasoline emission factors for 2024 are taken from ANL [39], and 2030 values are estimated based on improvement targets published by the European Commission [67]. In the high and low scenarios, the 2030 shares are assumed to be ±10% relative to the reference scenario.

. The scenario assumptions are aligned with existing policy targets and plausible development pathways. Specifically, NEV penetration, electricity mix transitions, and improvements in conventional vehicle performance are based on policy objectives from the Ministry of Industry and Information Technology of China and the European Commission, and calibrated using recent data for China (2024).

In the 2030 scenarios, we assume a 10% increase in NEV (BEV, PHEV, FCV) stock share for high ambition and a 10% decrease for low-target, with converse changes for conventional vehicles. Improvement scenarios assume reduced emission factors and gasoline consumption for ICEVs and HEVs compared to 2024. By calculating total life cycle emissions for all passenger vehicles in China, we evaluate the carbon reduction potential of China’s NEV sector growth and its impact on energy transition.

Figure 24 illustrates total life-cycle CO₂ emissions under improved conventional vehicle performance scenarios. Total CO₂ emissions from passenger vehicles are projected to increase in 2030 due to the expansion of the vehicle fleet, while a higher NEV share partially mitigates this growth. Further reductions are achieved through performance improvements in conventional vehicles, such as lower fuel consumption and emission factors. Notably, the 2030 high scenario, which features a higher NEV penetration than the 2030 reference scenario, achieves only limited emission reductions. In contrast, the improvement scenario, which combines NEV expansion with enhanced conventional vehicle performance, delivers substantially larger mitigation effects and can even offset the additional emissions associated with fleet growth. Overall, promoting NEV adoption alongside improving conventional vehicle efficiency is essential for achieving carbon peaking and neutrality targets and supporting effective climate policy implementation.

6. Conclusions and Policy Implications

Based on life cycle assessment theory, this paper conducted a comprehensive analysis of ICEVs, BEVs, HEVs, PHEVs, and FCVs, considering upstream energy source emissions. Scenario analysis was performed by adjusting key emission factors. Additionally, life-cycle economic costs, pollutant costs, and public health impacts were systematically assessed. Future emissions for China’s passenger vehicle sector were also projected. The following main conclusions are drawn:

(a)

In terms of CO₂ emissions, BEVs, HEVs, PHEVs, and FCVs generally outperform ICEVs. However, differences emerge across other pollutants: PHEVs emit higher NO_x and PM_2.5 than ICEVs, BEVs are associated with higher SO_x, and FCVs exhibit relatively higher CH₄ emissions. For FCVs, emission performance depends strongly on the hydrogen pathway. Coal-based hydrogen achieves a 75.2% reduction, while electrolysis under China’s current power mix can lead to negative reductions; in contrast, biomass-based hydrogen delivers the highest reduction at 78.2%.

(b)

In terms of life-cycle economic costs, BEVs, PHEVs, FCVs, and HEVs all exhibit lower costs than ICEVs. Among them, FCVs have both the lowest economic and pollution costs, whereas PHEVs, despite their relatively low economic cost, incur comparatively high pollution costs.

(c)

Under the current energy mix, only HEVs and FCVs yield net health benefits compared with ICEVs, avoiding 5.63 (12.09%) and 13.75 (29.53%) thousand deaths per year, respectively. In contrast, BEVs and PHEVs show higher PM_2.5-related mortality than ICEVs (50.97 and 51.94 vs. 46.56 thousand/year), with similar results for NO_x, suggesting that upstream power emissions offset part of the health benefits of electrification. It is important to note that these findings reflect distinct dimensions: while BEVs reduce CO₂ emissions substantially, their PM_2.5-related health impacts are shaped by the current coal-dominated electricity mix and higher vehicle weight, not by the electrification technology itself.

(d)

FCVs using coal-based hydrogen exhibit the highest carbon reduction rates and the lowest pollution costs over the full life cycle under the current power grid structure. Across scenarios varying in clean energy share, lifetime mileage, electricity consumption and vehicle weight, they consistently achieve the greatest CO₂ reductions, underscoring their superior mitigation potential under this energy pathway.

(e)

Under both the 2030 reference and 2030 high NEV share scenarios, carbon emissions remain higher than the 2024 baseline, indicating that NEV expansion alone cannot offset the growth in vehicle demand. Although a higher NEV share can alleviate the emission increase to some extent, absolute reductions are only achieved under the 2030 high NEV share with technological improvement scenario, where emissions fall below the 2024 level.

(f)

Increasing the proportion of NEVs in the passenger vehicle sector reduces full life-cycle CO₂ emissions for society, beyond the transport sector, even with current grid power generation. Improvements in power generation technology, energy mix, and cleaner energy use can further decrease these emissions and are essential to reversing the current net health deficit associated with BEV and PHEV adoption.

Energy structures vary by region due to resource endowments and technology. Despite China’s suboptimal grid mix, this study reveals that new energy vehicles (BEVs, PHEVs, FCVs) outperform ICEVs in environmental impact over their full life cycle. Emissions from BEVs and PHEVs are influenced by direct impacts and factors such as electricity generation, battery life cycle, and material recycling/disposal. Even with the current grid, BEVs and PHEVs exhibit better environmental performance than ICEVs in terms of greenhouse gases, though their immediate public health burden remains higher due to concentrated upstream pollution.

China’s power generation remains coal-dominant, with coal-fired power accounting for 57.8% of total generation in 2024. Given its coal-rich resources, coal power will remain crucial for China’s reliable energy supply. To improve this, China should integrate renewable energy with hydrogen storage, enhance research on power electronics for hydrogen production from wind and PV, and optimize hydrogen efficiency. Notably, even under the current power structure, BEVs and PHEVs achieve lower CO₂ emissions than ICEVs when upstream and downstream pollution transfers are considered.

While ecological governance is essential, economic considerations must also be taken into account. Compared with BEVs, PHEVs offer notable economic advantages and are therefore often a preferred choice at present. However, this comes at the cost of higher pollution impacts and health risks, as PHEVs generate the highest PM_2.5-attributable mortality (51.94 thousand/year) among all studied vehicle types. In contrast, choosing BEVs enables consumers to more effectively reduce both GHG emissions and air pollutants, allowing for more environmentally aligned decisions based on individual circumstances and preferences.

The emission performance of FCVs relative to ICEVs is highly sensitive to the hydrogen production pathway. In the case of coal-based hydrogen, which remains a major supply route in China, FCVs can achieve lower CO₂ emissions and pollution-related costs than other vehicle types under the modeled scenarios. However, this advantage is not universal and varies significantly across alternative hydrogen pathways. In particular, when hydrogen is produced via water electrolysis, the benefit diminishes. Given China’s coal-dominated power mix, the electricity-intensive production process leads to substantial upstream emissions, despite zero tailpipe emissions, FCVs may generate higher life-cycle GHG emissions and air pollutants than ICEVs.

Economically, BEVs, HEVs, PHEVs, and FCVs currently outperform ICEVs in terms of cost and pollution. China has a cost advantage in new energy vehicles due to its promotion and technological progress. Globally, especially in nascent markets, incentive policies [68] are crucial to foster growth and innovation in new energy vehicles. Encouraging research, development, and deployment of these technologies reduces pollution, promotes innovation, creates jobs, and drives economic growth in the clean energy sector [69,70]. Therefore, international cooperation is necessary to accelerate new energy vehicle adoption and achieve a sustainable transportation future.

From a broader transition perspective, BEVs are widely viewed as a key pathway for long-term decarbonization due to the elimination of tailpipe CO₂ emissions. However, this study highlights a key caveat: under the current energy mix, BEVs and PHEVs may impose higher health externalities than ICEVs (about 4.41–5.38 thousand additional deaths annually), reflecting a “pollution transfer” from tailpipe to power generation and non-exhaust sources. This suggests that the health benefits of electrification are highly dependent on electricity decarbonization, while BEVs remain crucial for long-term climate mitigation.

Despite an anticipated shift towards NEVs and a decline in ICEVs share, China’s total passenger vehicle life-cycle CO₂ emissions in 2030 are projected to surpass 2024 levels due to an overall increase in vehicle stock. However, concurrent performance improvements in traditional vehicles, such as reduced fuel consumption and CO₂ emissions from ICEVs and HEVs, alongside NEVs growth, could potentially lower total emissions below 2024 levels. This underscores the importance of enhancing traditional vehicle performance alongside NEVs expansion.

Integrating local health externalities into the decarbonization discourse is imperative, as the transition to NEVs involves a trade-off between global climate mitigation and localized particulate matter pollution. Our findings confirm that transport-related emissions contribute to a substantial mortality burden, with specific vehicle technologies shifting the weight between exhaust and non-exhaust or upstream sources. By applying vehicle-specific emission contribution ratios, this study clarifies that while electrification removes tailpipe CO₂, the increased upstream emissions and non-exhaust particulate matter from heavier NEVs currently impose significant cardiovascular and respiratory risks. Even for FCVs, optimizing the PTW phase through vehicle lightweighting remains essential to reduce non-exhaust externalities. As NEV penetration rises, a dual-track policy framework is required: alongside improving ICEV efficiency, regulators should adopt targeted measures addressing key drivers of non-exhaust emissions—particularly vehicle mass and braking dynamics—to mitigate local health impacts.

This study is limited by its relatively narrow system boundary. Future research could expand the analysis to include charging infrastructure, grid reinforcement, battery recycling uncertainties, and the evolving electricity mix, as well as extend to other countries or energy systems to assess the generalizability of the findings.