Emission Characteristics, OFPs, and Mitigation Perspectives of VOCs from Refining Industry in China's Petrochemical Bases

China's petroleum refining industry has set off another climax and has entered a new rapid development by building seven petrochemical bases. Sector-based volatile organic compound (VOC) emissions from the refining industry in seven petrochemical-developed provinces in China were estimated for 1990–2019 and projected for 2020–2030 under the business-as-usual (BAU), new policy control (NPC), and the highest control (HC) scenarios. Furthermore, speciated VOCs and their ozone formation potentials (OFP) were estimated in 2019. Total VOC emissions from existing refineries were 541.14 Gg in 2019, of which 43.9%, 31.3%, 18.3%, and 6.6% were from fugitive, end-of-pipe, tank storage, and wastewater treatment sources, respectively. Alkanes were the most dominant compound in refineries, accounting for 55.2% of total emissions, followed by alkenes (18.9%) and aromatics (12.5%). Alkenes were the highest contributor to OFP, accounting for 59.5% of total OFP, followed by alkanes (22.3%) and aromatics (13.7%). n-Butane, ethylene, cis-2-Butene, n-Decane, and n-Pentane were the top five species with the highest emissions, accounting for approximately 50% of total emissions. Whilst, ethylene, cis-2-Butene, n-Pentane, n-Butane, and m/p-Xylene were the top five species with the highest contribution to OFPs, accounting for approximately 78% of total OFPs. In 2030, 741.03 and 165.28 Gg more VOC will be released than in 2019 under a non-control condition and the BAU scenario. It is estimated that 75.05 and 228.67 Gg of VOC from all refineries can be reduced under the NPC 2030 and the HC 2030 scenarios. To effectively reduce VOC emissions from refining industries, priority should be given to fugitive emissions by improving and upgrading the production processes and implementing enhanced leak detection and Repair system. More efficient control technologies should be invented for end-of-pipe sources. Vapor recovery systems and secondary seals have great potential for VOC emission reduction from storage tanks.


INTRODUCTION
Over the past 40 years of reform and opening up, China has experienced rapid urbanization and industrialization.Energy consumption and industrialization have ballooned, making air pollution a severe problem in China and threatening human health (Li et al., 2016).To abate severe air pollution and protect public health, China's government implemented the toughest-ever clean air policy (Clean Air Action Plan) since 2013, targeting reducing fine particulate matter (PM2.5)nationwide.Contrarily, owing to the absence of effective control measures for volatile organic compounds (VOCs), China's anthropogenic VOCs emissions increased by 11% during 2013-2017 (Zheng et al., 2018).A recent study showed that VOC emissions from the petroleum industry in China showed a continuous increase during 2013-2019, while emissions from other industrial sources decreasing after peaking in 2016 (Simayi et al., 2022).As a result, photochemical smog pollution characterized by a high level of ozone (O3), which is produced in the air by oxidations of VOCs in the presence of nitrogen oxides (NOx) and sunlight, has become increasingly prominent in key regions such as the Beijing-Tianjin-Hebei region (BTH), the Yangtze River Delta region (YRD), and the Pearl River Delta region (PRD) (Zheng et al., 2017;Li et al., 2019b;Gong et al., 2020;Wang et al., 2022a).Therefore, China has been focusing on the atmospheric VOC pollution problem in recent years, and the Ministry of Ecology and Environment (MEE) launched a comprehensive treatment plan to reduce VOCs emissions in key industries (MEE, 2019), and mandatory industrial standards were released relating to fugitive emission sources and refineries (MEP, 2015b(MEP, , 2019)).
The petroleum refining industry is an essential contributor to VOC emissions in China and worldwide, accounting for approximately 3% and 10-30%, respectively, of the total anthropogenic VOC emissions and industrial emissions (Alyuz and Alp, 2014;Simayi et al., 2019).China's oil consumption has continued to proliferate and contributed to an approximately 60% increase in global oil refining from 1990 to 2015 (Jia et al., 2020).According to the British Petroleum Company (BP) forecast, the growth of oil consumption in 2019 was led by China, and oil will continue to occupy a vital position in the energy system in the coming decades (bp, 2020).China's government set up the "Refining and Chemical Integration" program in recent years.This program will promote building seven world-class petrochemical bases in Liaoning, Hebei, Jiangsu, Shanghai, Zhejiang, Fujian, and Guangdong provinces.Consequently, China's refining industry will usher in a new peak of refining capacity in the next few years, and this will be driven by those seven provinces.The accelerated growth of petroleum refining capacity will affect climate change, air pollution, and quality of life and may bring tremendous pressure on China's air pollution control and "carbon peaking and carbon neutrality".Thus, it is urgent to formulate and implement strict and effective emission control measures in China's petroleum refining industry.
A prerequisite for mitigating VOCs emissions is to establish a precise and comprehensive emission inventory, which can not only identify key emission sources, but also provide an important database for air quality simulation (Liu et al., 2021).In previous studies of VOCs emissions inventories from anthropogenic and industrial sources in China, VOCs emissions from the petrochemical industry have been considered as integrated sources (Qiu et al., 2014;Wu et al., 2016;Zheng et al., 2018;Li et al., 2019c;Simayi et al., 2020).However, the refining industry has large industrial installations and tens of thousands of VOC discharge routines, including refinery units, storage tanks, combustion sources, flare systems, wastewater collection and treatment, product loading/unloading processes, etc. (Liu et al., 2020).With the increase in China's VOCs emission control requirements, scholars have begun to investigate VOCs' emission characteristics from the different processes and sectors in key industries (Han et al., 2020;Cheng et al., 2022;Wang et al., 2022b).For instance, Li et al. (2019a) studied the process level VOCs emission characteristics from coking industries and established historical VOCs emission inventory for coke production.Liu et al. (2020) investigated the establishment method of process-based VOCs emission inventory for the petroleum refining industry based on the measurement of a medium-scale refinery.Although a few researchers have attached great importance to the issue of quantifying and characterizing VOCs emissions from the petroleum refining industry, there are some research limitations and shortcomings.For instance, most current studies have focused on the measurement of ambient VOC concentrations around plants, VOCs emission sources profiles from different sectors, and estimating VOCs emissions from a single refinery (Wei et al., 2016;Chen et al., 2020;Liu et al., 2020;Zhang et al., 2022).
However, lack of studies to estimate VOCs emissions and analyze future trends by taking the national or regional refining industry as a whole in China.Therefore, it is necessary to conduct a detailed analysis of historical and future changes in VOC emissions from refining processes, especially in those seven provinces that account for approximately 50% of the national crude processing capacity.Considering the above issue, this study investigated the emission changes of total VOCs from different sectors of the refining industry in the seven petrochemical bases of China from 1990 to 2019, and analyzed the speciated emission characteristics of VOCs by different emission sectors in 2019.Furthermore, the future emissions changes and reduction potentials of total VOCs by sectors and provinces were discussed under one non-control condition and three different control scenarios targeting 2025 and 2030.Finally, the feasibility of this study and policy implications were discussed.

Data Sources and Prediction of Future Activity
Petroleum refineries in the petrochemical-developed seven provinces, including Liaoning, Hebei, Jiangsu, Shanghai, Zhejiang, Fujian, and Guangdong, were selected as the research objects and located in 91 existing (old) refineries in the study area (Supplement Fig. S1).To analyze the historical and current VOC emissions performance from refineries in those seven provinces, crude oil processing capacity data was collected at provincial level for 1990-2019 from the reports and Yearbooks of Sinopec, PetroChina, and each province.The refineries planned to be constructed during 2019-2030 are regarded as new refineries.The annual crude oil processing capacity of 29 new refineries is collected at individual factory level according to the "Refining-Chemical Integration" program (Table 1).In order to predict the crude oil processing capacity of a province in a certain year, we made the following assumptions: (1) the sizes and processing capacities of 91 old refineries in the study area will not change until 2030, and (2) the processing capacity of 29 new refineries, which will be in operation until target year, will be added to existing processing capacities.

Source Classification and Emission Calculation
Generally, the refineries contain 12 specific emission sources (Supplement Fig. S2), according to "Checking Guideline for VOCs Pollution Sources in the Petrochemical Industry" recommended by the Ministry of Environmental Protection (MEP) of China (MEP, 2015a).Sources classification of refinery VOC emissions in this study followed the guidelines of MEP and related studies of Liu et al. (2020).These 12 kinds of emission sources were classified into four categories in this study to calculate the sector-based VOCs emissions, including fugitive sources (including accidental emissions, leakage from pumps and valves, start-up and shut down, sampling, cooling towers, and process unit leakage), end-of-pipe sources (including combustion sources, flare system, and exhaust funnel), storage tank sources (tanks breath and loading/unloading process), and wastewater treatment (wastewater collection and treatment) sources (Supplement Text.S1).
The different sectors' VOCs emissions were calculated by multiplying the crude oil processing capacity and sector-based EFs by Eq. (1).To determine the sector-based EFs of refining industry, a systematic survey was conducted on VOC emissions control technologies in China's refineries, and the removal efficiency of each control technology was given in Table S1.These technologies' application rate differs from region to region in China, ranging from 20% to 50%, and the penetration rate in key regions is higher than in other regions (Gu, 2020).The highest application rate (50%) was assumed for those seven provinces in this study because they are almost located in key regions, for further details please refer to Supplement Text S2.
where p is the province (n = 7), s is the emission sectors (end-of-pipe, fugitive, wastewater treatment, and storage tank), t is the abatement technology, y is the target year; E(VOC): total VOC emissions from refinery processes in the study area; A: activity data; ef: unabated emission factors; η: removal efficiency; X: actual application rate of abatement technology.In Eq. ( 1), crude oil processing capacity was considered as the activity data, and the sectorbased unabated EFs were collected from earlier local studies and AP-42 databases (Table 2).Then, the abated EFs were calculated by removal efficiency and application rate of different control technologies as shown in Eq. (2).
where A-EF and U-EF represent abated EFs and unabated EFs from a certain source, respectively; η and X are the removal efficiency and penetration rate of the corresponding sources.

Estimation of Speciated Emissions and its OFPs
In this study, the source spectrum data measured in different emission sectors in the refining industry in China were collected (Wei et al., 2014b;Mo et al., 2015;Zheng et al., 2018;Lv et al., 2021), and the source spectrum data of each VOCs component in four different emission sectors and whole refining industry were obtained by giving the same weight to the source spectrum measured by each researcher, as shown in Table 3.The speciated VOC emissions were calculated by multiplying the total VOCs emissions of each sector and the corresponding weight percentages in an integrated VOC source profile, as formulated in Eq. (3).
where Ei,s is the emissions of species i from emission-sector s; Es is the total VOCs emissions from emission-sector s; fi,s is the weighted percentage of species i from emission-sector s.
Ozone formation potentials (OFP) of VOCs is used to evaluate the maximum contribution of VOCs species to O3 formation under the optimal reaction conditions.The contribution of different VOC species to form ambient O3 was calculated by multiplying the maximum incremental  reactivity (MIR) by the emissions of the corresponding VOC species, as formulated in Eq. ( 4).
where OFPi,s is the ozone formation potentials of VOC species i in emission-sector s; Ei,s is the emissions of species i from emission-sector s; MIRi is the maximum incremental reactivity of species i.

Emission Mitigation Scenarios
Based on primary measures employed to control VOCs emissions in the refining industry in China and the specific situation of new refineries, we established three emission mitigation scenarios and one non-control scenario to compare VOC emission changes in the context of "Carbon Peaking".These three control scenarios include the business-as-usual scenario (BAU), the new control policies scenario (NPC), and the highest control scenario (HC).Each scenario considers the existing VOCs emission control policies/standards in refineries, new policies/standards which might be implemented in the future, and the penetration rates of different control technologies under the different control policies as described below, and for further details on emission reduction scenarios please refer to Supplement Table S2.
(1) The non-control condition assumes that there will be no new control technologies or control policies/standards in the refining industry for all seven provinces until 2030.Both old and new refineries will follow the same control measures applied before the baseline year of 2019.The non-control condition aims to provide a baseline picture of how the VOC emissions will evolve without any new policy interventions.
(2) The BAU scenario is a basic scenario and follows the existing control policies before 2019, also considers new policies/standards which will be implemented after 2019, such as "VOCs emission control from key industries", "Standard for fugitive emissions of volatile organic compounds, GB37822-2019", VOCs control policies in "13 th five-year plan" and "14 th five-year plan".In this scenario, control measures and technologies will be implemented at a slow pace and with a lower application rate until 2030 compared to the NPC and the HC scenarios.1.0 Note: species included in the table were detected at levels greater than 2% in at least one profile of four sectors; The MIR value of other species is assumed to be 1 g O3 g VOC -1 .
(3) The NPC scenario assumes that the national requirements for VOCs emissions reduction will be gradually stringent during the "14 th five-year plan" and "15 th five-year plan".For example, the highest emission standards (special emission limits) will be implemented in all refineries, the proportion of floating roof tanks will increase, and old/outdated storage tanks will be largely eliminated.Besides, the quarterly and monthly Leak Detection and Repair (LDAR) technology will be implemented in local and state-owned refineries, respectively, and the definition of VOC leakage will be reduced from 5000 µmol mol -1 to 2500 µmol mol -1 .
(4) The HC scenario is an ideal case to control VOC emissions in refineries.This scenario assumes that the best control technologies are assumed to be readily available for abatement, new controlling policies will continually be released and all available technologies will be applied with the highest penetration rate ever recorded.Besides, monthly LDAR will be implemented in both the local and state-owned refineries, and the definition of VOC leakage will be reduced from 2500 µmol mol -1 in the NPC to 1000 µmol mol -1 .All of the old fixed and external floating roof tanks will be replaced with pressure tanks and internal floating tanks or will be eliminated.Technologies such as gas recovery systems and secondary sealing will be widely adopted in all refinery tanks.All wastewater treatment plants will be covered and the fugitive emissions will turn into organized emissions, then will be discharged into the atmosphere after treatment by end-of-pipe emission control technologies.

Total VOCs Emission Characteristics
The refining capacity of each province and its VOC emissions by different sectors in 2019 were plotted as in Fig. 1.The refineries in seven provinces produced 541.14 Gg of VOCs in 2019, with 297.33 Tg of crude oil processing capacity.Refineries in Liaoning had the largest VOCs emission (180.03Gg) with accounting for 33% of total VOC emissions, followed by Guangdong (20%) and Jiangsu (16%), while Hebei (5%) had the least VOC emissions.From the perspective of different emission sectors, fugitive sources have always been the largest VOC emitters in refineries and accounted for more than 40% of total emissions, followed by end-of-pipe sources (31.3%),tank sources (18.3%), and wastewater treatment sources (6.6%).Yen and Horng (2009) and Wei et al. (2014b) conducted on-site observations on oil refineries in Taiwan and Beijing, respectively.The results showed that unorganized emissions accounted for 40-50% of the VOC emissions from refineries, followed by end-of-pipe and tank sources, which is similar to the emission inventory results of this study.
Variation of crude oil processing capacity and VOC emissions from refineries in seven provinces during 1990-2019 is represented in Fig. 2. Total VOC emissions in seven provinces increased from 106.60 Gg to 541.14 Gg during 1990-2019, with an annual average rate of 5.6%.On the contrary, due to the Clean Air Act's implementation in 1970 and more stringent regulations on refineries emissions after 1990 in the US, VOCs emissions from refineries in the USA decreased from 225 Gg in 1990 to 70 Gg in 2015 (Nelson and Nguyen, 2015; U.S. EPA, 2019).VOC emissions from the refining industry in these seven provinces of China in 2015 (441.19Gg) were six times higher than that of the USA, while their crude oil processing capacity was only 30% of that of the US.
VOCs emissions in Liaoning were the highest owing to the large processing capacity of refineries.Emissions increased from 38.79 Gg in 1990 to 179.66 Gg in 2019, with an average annual growth rate of 5.4%.The refineries in Guangdong and Jiangsu have large VOCs emission increments during 1990-2019 and emissions increased by 87.95 Gg and 73.27 Gg, with the growth rate of 6.1% and 6.6%, respectively.Fujian's emissions were the least in 1990 among the seven provinces, but the emissions increased from 3.62 Gg to 40.04 Gg with the highest average annual growth rate of 8.9%.The number of refineries in Hebei (15 refineries) was five times the number of refineries in Shanghai (3 refineries).However, due to the large scale and huge refining capacity of Shanghai's

Speciated Emissions and its OFPs
To identify the key species with large emissions and high contributions to O3 formation, this study estimated speciated VOCs emissions from four sectors in the refining industry based on the source profiles of local measurements (Wei et al., 2014a;Lv et al., 2021;Zhang et al., 2022), then estimated OFPs of each species in 2019.Results showed that alkanes accounted for approximately 60% of total VOCs emitted by refineries with emissions of 326.93 Gg in 2019, of which 19.1%, 39.8%, 33.6%, and 7.6% come from the tank, fugitive, end-of-pipe, and wastewater treatment sources, respectively (Table 4).The emissions of alkenes and aromatics accounted for 19.6% and 13.8% of total emissions, and end-of-pipe sources were the largest sources of alkenes (47.9%), while fugitive sources were the largest sources for aromatics (64.8%).
For the individual species, n-Butane has the largest emissions (87.53 Gg) accounting for 16.18% of total emissions, followed by ethylene (49.01 Gg), cis-2-butene (47.58 Gg), n-decane (47.43 Gg), and n-pentane (29.63 Gg), accounting for 9.06%, 8.79%, 8.76%, and 5.48% of total emissions, respectively.In these five species, n-butane, ethylene, and n-decane are mainly from end-of-pipe sources, and cis-2-butene and n-pentane are from fugitive sources.In addition, aromatics and OVOCs such as benzene, MTBE, toluene, and m/p-xylene also had a considerably high emissions, each of them accounting for more than 2% of total emissions.Therefore, priority should be given to these species to control total VOCs emissions in China.
In 2019, the VOCs emitted from the refining industry in the seven petrochemical bases had 1524.05Gg of ozone generation potentials, in which the alkenes had the largest contribution accounting for 59.5% of total OFPs, followed by alkanes (22.3%) and aromatics (13.7%), and the OVOCs, halocarbons, and other species only contributed 4.5% of total OFPs.Due to the high reactivity of alkene species, their OFP is equivalent to nine times of the emissions and three times higher than the OFP of alkanes, while their emissions are half of those of alkanes.As seen in Fig. 3, the magnitude of VOCs species emissions and their OFP values differ due to the large differences  in MIR values of various species.For example, alkene species such as ethylene and cis-2-Butene have the highest OFP with lower emissions than those n-Butane and n-Decane due to their very high MIR values, with an OFP of 469.73 and 411.99 Gg, respectively.Contrarily, alkane species such as n-Decane and Isopentane, which have higher emissions, and most halogenated hydrocarbons have smaller OFP due to their lower MIR values.In conclusion, species such as ethylene, cis-2-Butene, n-Pentane, n-Butane, m/p-Xylene, and toluene are the largest contributors to OFP from the refining industry accounting approximately 80% of total OFPs, thus to control ozone pollution in the study area these species must be controlled as a priority

Total mitigation potentials
The VOC emissions from refineries in the seven provinces for 2025 and 2030 under the three control scenarios were calculated from the premise and considerations presented.VOC emissions from all refineries for the baseline year and projected emissions for the three future scenarios were summarized in Fig. 4. If no mitigation actions are assumed for all sectors of refineries (Non-control condition), VOC emissions from the seven provinces would increase significantly from 541.14 Gg in 2019 to 1283.27Gg in 2030.In the BUA scenario, due to the large numbers of newly established refineries and their huge processing capacity in the future, VOC emissions from refineries will sharply increase from 541.14 Gg in 2019 to 741.12 Gg in 2025, with an annual average increase rate of 6%.However, by 2025, 345.19 Gg of VOCs would be eliminated compared to the Non-control condition.In 2030, VOC emissions will slightly decrease compared to 2025, but will be higher than that in 2019.Compared with BAU, the VOC emissions in 2025 will still be higher than that in 2019 in the NPC scenario, despite the new measures and advanced control technologies will be taken, but the growth rate of emissions will slow down.In 2030, VOC emissions will decrease to 467.19 Gg, which is 13.8% lower than in 2019, indicating that the emission reduction in the NPC scenario 2030 is greater than the emissions increment from growth in the processing capacity of refineries.However, strict multiple control measures will effectively mitigate the VOC emissions despite increasing processing capacity.Thus, VOC emissions from refineries will sharply decrease to 387.42 Gg and 312.39 Gg in 2025 and 2030, respectively, under the HC scenario.The HC scenario will eliminate 42.2% of VOCs in 2030 compared to the baseline year and reduce 393.95Gg and 153.62 Gg of more VOCs than the BAU 2030 and the NPC 2030.

Mitigation potentials by province and emission sector
Fig. 5 illustrated the emission performance of each province under the three control scenarios.As discussed above, VOC emissions will peak in 2025 under the BUA scenario, and most of the increase will mainly originate from Hebei and Zhejiang, accounting for 30% and 26% of the total emission increase in 2025, followed by Guangdong and Jiangsu.Due to the slow growth rate of crude oil processing capacity in Liaoning during 2020-2025, refinery VOC emissions will be lower than in 2019 in this scenario, owing to new emission standards and control measures.However, with the construction of a large number of refineries in Liaoning after 2025, VOC emissions will also gradually increase and emissions in 2030 will be higher than that in previous years despite the stringent control measures in previous years.In 2030, the emission in five provinces except Hebei and Liaoning will decrease compared to 2025, but emissions in all seven provinces will still be higher than in 2019.
Under the NPC scenario, refineries in Hebei and Zhejiang in 2025 will generate 47.93 Gg and 31.04Gg of more VOCs relative to 2019, but Liaoning and Fujian will eliminate 57.56 and 5.15 Gg of VOCs, respectively.In 2030, except for Hebei and Zhejiang, the other five provinces will have emission reductions with different degrees than in 2019, and total emissions will be 75.05Gg lower than in 2019.Although Hebei will adopt more strict control measures, the VOCs emissions increment caused by the huge growth of processing capacity will be larger than the emission reduction of NPC control scenarios, and emissions in 2030 will be 50.65 Gg higher than in 2019.Under the HC scenario, total VOC emissions in 2025 and 2030 will be 154.82Gg and 229.85 Gg lower, respectively, than in 2019.The largest reduction potential will come from Liaoning, followed by Guangdong and Jiangsu, and these three provinces are responsible for 38.3%, 21.4%, and 16.5% of the total reduction.Emissions in Hebei will continue to increase even under the HC scenario due to the substantial increment of processing capacity, and VOC emissions in 2030 will be 91.5% higher than in 2019.
From the perspective of different emission sectors (Fig. 6), wastewater treatment processes are expected to have a particularly low VOC mitigation potential relative to other sources.Under the BAU scenario, emissions from all sectors in 2025 and 2030 will be higher than the baseline year, but lower than in non-control conditions.Under the NPC scenario, emissions from tank and wastewater treatment sources will have an increasing trend until 2025, while emissions from end-of-pipe and fugitive sources will decrease.In 2030, VOC emissions from wastewater treatment will still be higher than in 2019, while the other three sources will show decreasing trend, and the fugitive sources had the largest reduction potentials, followed by end-of-pipe and tank sources.Under the HC scenario, in addition to large emission reductions from fugitive and end-of-pipe sources, tank sources will also have considerable emission reduction potential due to the large replacement of floating roof tanks with domed fixed roof tanks and installing vapor recovery systems.Reduction from fugitive emissions will become the largest emission reduction source in 2030 under the HC scenario and will eliminate 97.13 Gg of VOCs compared to 2019.Tank and end-of-pipe sources will also have considerable reduction potentials, reducing 60.31 Gg and 63.93 Gg of VOCs, respectively, compared to 2019.
VOC emission control in China started relatively late compared to developed countries.Since 2010, China's government has been concerned about VOCs emission pollution to tackle photochemical smog pollution (Zheng et al., 2018).However, due to unclear emission characteristics of VOCs and the weak control measures, there is a lack of experience in both the screening of emission reduction objects and the selection of emission reduction technologies in China (Ye and Chen, 2017).This study and previous studies all showed that refining industry have great contribution on China's anthropogenic VOCs emissions and ambient ozone formation (Liu et al., 2020;Lv et al., 2021).Based on results of this study, if strict control measures are not taken, VOCs emissions

Future reduction perspectives
According to the results of this study, to effectively reduce VOC emissions from the refining industry in China, special emphasis should be placed on VOC emission management in these petrochemically developed seven provinces.By 2030, new refineries' crude oil processing capacity will account for 60% of the total processing capacity in these seven provinces.Therefore, new refineries' emission control work should be managed well before the operation, and new refineries should implement the "special emission limits" in the GB 37822-2019 andGB 31570-2015.Due to the fugitive emissions control requiring a significant financial and human resources investment, the small local refineries with poor management may not implement the fugitive emission standards in the GB 37822-2019 without strict supervision to save their costs.The government should strengthen the supervision of control of fugitive emissions.Meanwhile, the end-of-pipe and tank emissions control measures also should be strengthened, and the removal efficiency of end-of-pipe control technology at both new and old refineries in these seven provinces must meet the special emission limits (≥ 97%) required by GB 31570-2015.Emission control devices must be regularly maintained and replaced, and the VOC concentration in the exhaust gas from the end-of-pipe must be regularly or continuously observed to know whether the concentration meets the emission standards.
The tank sources also have great VOCs reduction potentials by eliminating or replacing old tanks, installing floating roofs on fixed-roof tanks and domed fixed roofs on external floating roof tanks, installing double stage vapor recovery system, adopting a double seal between the floating plate and the tank wall, and adopting high-efficiency sealing methods such as liquid inlay in the primary seals.The new refiners will use the inner floating roof tanks and the best available control technologies before the operation.Thus, controlling tank emissions in old refineries is a major problem in tank sources because the elimination or replacement of old storage tanks requires a large amount of capital investment.
In summary, the emissions control policies and measures proposed by this study are feasible for both old and new refineries, and old refineries require a large investment and powerful management to reduce VOC emissions.

Uncertainty analysis
Although the assumptions in this study are based on the actual condition of refineries, current and future control policies, and measures that will be implemented in the future, there are still several uncertainties and limitations in the assumptions.First, when predicting the future activity data, the total numbers and processing capacities of 91 existing refineries were assumed to remain as of 2019.The policies of eliminating backward productivity and excess capacity in the future may lead to changes in the crude oil processing capacity.However, this kind of change will not be large and will not impact the results, owing to a large increase in those seven provinces' processing capacity after 2019.Second, the vintage, size, and management systems are different for each refinery.Therefore, the emission rates and control methods of each refinery are quite different.For this condition, the entire refinery was considered a single refinery, and the same application rate of control technologies was assumed for all refineries.To tackle this kind of issue, further work should be carried out to collect more accurate data on facility conditions and control technologies.Third, in addition to the storage of oil products in refinery tanks, the storage of oil products in terminals and transportation from terminals to the refineries is also a very important VOCs source.Unfortunately, the activity data and EFs for storage and transportation in terminals are still unclear in China.Thus, this study only considered the storage and loading/unloading of oil products in refineries and recommended conducting more research on VOCs emissions from terminals and ports.

CONCLUSION
China's refining industry is developing rapidly and will develop even faster after 2020, which will bring tremendous pressure on VOC emission reduction and ozone pollution control.91 old and 29 new petroleum refineries were investigated in the petrochemically developed seven provinces in China during 1990-2030.The VOC emissions from four emission sectors in the old refineries were estimated for 1990-2019 and speciated VOCs emission characteristics and their OFPs in 2019 were analyzed.Three sector-based emission control scenarios were established for 2020-2030 and effective emission control measures were recommended for each sector based on scenario results.
The annual VOC emissions from refineries in the seven provinces increased by order of magnitude from 106.60 Gg to 541.14 Gg during 1990-2019.Liaoning had the highest emissions, accounting for 33% of total emissions, followed by Guangdong, Jiangsu, Zhejiang, Shanghai, Fujian, and Hebei in 2019.From the different emissions sectors, fugitive emissions contributed more than 40% of total emissions, followed by end-of-pipe, tank, and wastewater treatment sources.
The refining industry in seven provinces generates 1524.05Gg of ozone with 541 Gg of VOCs emissions.Alkanes were the dominant chemical species in refineries accounting for 60% of total VOCs emissions, followed by alkenes (17.3%), aromatics (12.2%), and OVOC + halocarbons (10.5%).However, alkenes had the highest contribution to OFP accounting 59.5% of total OFPs, followed by alkenes (22.3%) and aromatics (13.7%).Alkanes are mainly emitted from end-of-pipe sources, while alkenes, aromatics, and OVOCs are mainly emitted from fugitive and tank sources.The n-butane, ethylene, cis-2-butene, n-decane, and n-pentane were the top five species with the highest emissions in refineries, accounting for approximately 50% of total VOC emissions.Ethylene, cis-2-butene, n-pentane, n-butane, and m/p-xylene are the top 5 species with large OFP in the refining industry, and complete control of these species would reduce OFP by about 78%.
Under the BAU scenario, emissions in 2030 will still be 30.5% higher than in 2019, while the NPC and the HC scenarios will reduce 13.8% and 42.2% of VOC emissions compared to 2019.Liaoning and Guangdong will have the largest VOC emission mitigation potentials, while emissions in Hebei and Zhejiang will continue to increase even under the Highest control scenario.Both fugitive and tank storage emissions have great potential for VOC emission reduction, but needed huge investment in capital and technologies.End-of-pipe emissions also have considerable emission reductions due to inefficient control measures and lower penetration rates of technologies in the old refineries.The highest pressure to reduce VOCs emissions in the future mainly comes from old refineries, but emissions in new refineries are also not negligible owing huge increase in processing capacity.
Fi-Tanks stands for Fixed roof tank; Ex-Tank stands for External floating roof tank; In-Tanks stands for Inner floating roof tank.a Data from (Qiu et al., 2014); b Data from U.S. EPA (2009); c Data from local measurement Li (2017); d Data from native research Qiang (2010).

Fig. 1 .
Fig. 1.Oil processing capacity of each province and corresponding VOC emissions by sector in 2019.

Fig. 2 .
Fig. 2. Historical changes in crude oil processing capacity and VOC emissions of refineries in seven provinces during 1990-2019.

Fig. 4 .
Fig. 4. Historical VOC emissions from refining industries in seven provinces during 1990-2019 and projected emissions from 2020 to 2030 under three control scenarios and non-control conditions.

Fig. 5 .
Fig. 5. Changes in refining VOC emissions from different provinces during 2020-2030 relative to 2019 under the three control scenarios.

Fig. 6 .
Fig. 6.Changes in refining VOC emissions from different sectors under the three control scenarios during 2020-2030 relative to 2019.

Table 1 .
Refinery information for future construction (the unit of changing and adding processing capacity is Tg yr -1 ).

Table 2 .
Default and abatement emission factors under the three control scenarios.

Table 3 .
Key VOCs species in different emission sectors in the refining industry and their MIR.

Table 4 .
Source profiles and emissions of VOCs (four categories) in the different sectors of refineries.