A Risk Evaluation Method of Coastal Oil Depots for Heavy Rainfall Vulnerability Assessment

Abstract

Oil depots in the coastal areas of China are prone to disasters caused by heavy rain due to the monsoon climate. Studies focusing on heavy rainfall vulnerability in coastal oil depots are limited. Therefore, we evaluated the safety of oil depots based on four factors in this study: personnel, equipment and facility, environment, and resilience. Complex networks, analytic hierarchy processes, and information entropy theory were used to establish an evaluation index system including four first-level indicators, nine second-level indicators, and 40 third-level indicators. Scores of 40 evaluation indicators were taken as the input, a vulnerability level of oil depots affected by heavy rain was gained as the output, and results were presented visually (different warning levels distinguished by color) to help oil depot enterprises improve their safety performance under extreme weather conditions.

1. Introduction

Coastal oil depots are built with reliance on ports and distributed in many parts of the world [1]. Currently, China has nine large-scale crude oil depots with total reserves of more than 60 million tons. Owing to the frequent extreme climatic conditions in coastal areas, including extremely heavy rainfall and typhoons, oil depots are prone to considerable damage [2]. Heavy rainfall can further lead to disasters, such as atmospheric storage tank depression, road waterlogging, workplace waterlogging, atmospheric pressure tank deflation, and chemical park flooding, which influence normal production and cause casualties and economic losses [3]. For the disaster management of coastal oil depots, it is important to research their vulnerability to intense storms and heavy rainfall while considering coastal oil depot exposure.

Most previous studies have focused on common accidents such as fires and explosions [4] and hope to reduce their effects. Shi et al. [5] proposed a fuzzy fault tree assessment based on analytic hierarchy processes (AHP) for steel oil storage tanks, but the evaluation method, which is easily affected by experts’ experiences, is unable to include as many accident causes as possible. Similarly, Chen et al. [6] developed an interpretive structural model and system dynamics for large crude oil depots, which depends on the subjective judgment of experts and has four indexes, including human, equipment, environment, and management. However, resilience should be regarded as an important part of the evaluation system and should be considered. Ding et al. [7] proposed a model considering fire and explosion in a synergistic effect, which serves as a quantitative risk assessment tool to evaluate equipment vulnerability under a spatial-temporal synergy of heat and pressure. Liu et al. [8] studied the aftermath of a tank leakage accident by computational fluid dynamic simulation, which plays a huge role in accident prediction. Except for equipment vulnerability, other factors should be taken into account in simulations, such as management, human vulnerability, environment vulnerability, and so on.

Research on other aspects of oil tanks has been published in recent years, such as studies considering the combined loadings of the fragmenting impact and pool-fire caused by domino accidents, analyzing the prevention of fire-induced domino effects based on a fuzzy Petri et (FPN) [9], constructing the automation of emergency response for petroleum oil storage terminals [10], identifying the most vulnerable installations of process plants subject to domino effects [11], and using a comprehensive risk assessment to analyze the vulnerability of the oil and gas sector to climate change and extreme weather events.

Vulnerability studies in other industries also provide a great and valuable reference, such as studies on pipelines and LNG. Wang et al. [12] studied the vulnerability of gas transmission capabilities of natural gas pipeline networks; by multiplying the component importance and risk values, the vulnerability of the component was obtained. Rossi et al. [13] implemented a quantitative risk assessment (QRA) for evaluation of the failure probability and frequency of piping involved in flooding. Hu et al. [14] analyzed marine LNG offloading systems’ dynamic resilience, considering weather-related hazards based on IRML, weather factors affecting human performance, maintenance measures, and system resilience.

With the development of technology, some advanced technologies are used in the evaluation system, such as cloud model theory and algorithms. Karthik et al. [15] developed a comprehensive procedure for risk assessment of the industry’s sites, using geospatial tools to evaluate and map personnel vulnerability in Kerala, India. Additionally, social media is more and more commonly used as a data source to track disaster events and assess the consequences caused by natural hazards, such as heavy rain and floods [16,17]. The methods have higher requirements for historical data and are more sensitive to the quality of data processing, but are applied with difficulty to those fields with little data.

Rainy weather and flooding with particles will damage oil tanks and cause oil spill accidents due to the potential energy of the water or long-term immersion. Previous studies provided various methods for analyzing the vulnerability of natural hazards, such as the quantitative risk assessment method, a weighted comprehensive evaluation, analytic hierarchy processes, etc. However, the scenarios are different as petrochemical companies gradually migrate to islands, many oil depots are placed in coastal areas and islands, and the heavy rainfall brought by the subtropical monsoon climate should be regarded as a significant natural hazard. Thus, it is important to establish an evaluation system to reduce the impact of heavy rain on coastal oil depots. The present study intended to take advantage of previous studies and carry out a quantitative risk assessment, supporting the oil depot industry to strengthen its emergency management system capability.

This paper is structured as follows. Section 2 presents the vulnerability of heavy rainfall to coastal oil depots with the damage caused and the factors affecting it. In Section 3, the evaluation index system for assessing heavy rainfall vulnerability in coastal oil depots in detail is introduced, which includes four first-level indicators, nine second-level indicators, and 40 third-level indicators. The indicators were selected using a complex network combined with expert opinions, and indicator weight was generated using multiple methods such as analytic hierarchy process (AHP), and information entropy theory. Through the evaluation index system, a heavy rainfall disaster assessment was constructed for a coastal oil depot.

2. Heavy Rainfall Vulnerability

2.1. Damage Caused by Heavy Rainfall to Coastal Oil Depots

The vulnerability analyses of coastal oil depots should be based on specific heavy rainfall disasters, which are often accompanied by sudden cooling. The disasters caused in oil depots following heavy rainfall as shown in Figure 1 include (1) short-circuiting of instruments due to water, (2) clogging of tank’s breathing valve, (3) overflowing of the cesspool, (4) road waterlogging, (5) pump room flooding, (6) fire dike flooding, (7) storage tank material expanding and contracting with temperature variation, (8) flange expansion and contraction due to temperature, (9) pipe-line crystal precipitation, and (10) temperature abnormality of high-temperature devices.

2.2. Factors Affecting Heavy Rainfall Vulnerability in Coastal Oil Depots

This study considers personnel, equipment and facilities, and environmental vulnerabilities as the different subsystems of heavy rainfall exposure vulnerability in coastal oil depots based on analyses of the damage caused [18]. This system also presents the ability to repair damaged equipment and resume production, which is represented by resilience, as shown in Figure 2.

(1)

Personnel vulnerability

Personnel, who also constitute a part of the disaster victims, play an important role in the disaster response capacity of oil storage enterprises. The physical health, operational proficiency, and safety awareness of the staff in the oil depot affect their safety in daily work, as well as the disaster resistance and mitigation capability of the oil depot. The factors affecting personnel vulnerability include age structure, working years, health status, education, psychological factors, operational proficiency, and exposure [19,20].

(2)

Equipment and facility vulnerabilities

Equipment and facilities in coastal oil depots are the main victims of heavy rainfall, and vulnerability is affected by exposure, sensitivity, and management coping abilities. The factors affecting equipment and facility vulnerabilities include equipment and facility exposure, function importance, asset ability, disaster resistance, degree of material danger, material quantity, and energy.

(3)

Environmental vulnerability

The oil depot environment includes the natural and the surrounding environments. In addition, the oil depot environment also regulates a company’s reputation and sociocultural environment. On the one hand, the environment has a role to play in deciding whether heavy rainfall will cause disasters. On the other hand, it is vulnerable to accidental pollution from the oil storage facility. The factors affecting the natural environment include slope, water environment, and soil environment [21].

(4)

Resilience

Resilience is the ability to quickly restore a damaged body to its original state after a disaster, which enables the facility to resume production and reduce economic losses. The factors affecting resilience include the disaster recovery plan, loss assessment, bearing capacity, and social assistance [22].

3. Evaluation Index System

The factors affecting heavy rainfall vulnerability in coastal oil depots analyzed in Section 2 cannot be directly considered as an evaluation index. Some of the pre-indicators of influencing factors are relatively imprecise and need to be refined into multiple indicators. Through research on complex network theory, an evaluation index is selected and obtained by applying AHP and entropy theory.

3.1. Index Identified

The pre-evaluation is refined, and the following operations are performed based on the analyses of vulnerability factors.

(1)

Establishing an adjacency matrix of vulnerability factors

An adjacency matrix as shown in Figure 3 was established using Microsoft Spreadsheet, with different nodes indicating different evaluation indexes determined by applying AHP and entropy theory. The adjacency matrix is the data that characterizes the relationship between a set of vertices. In the adjacency matrix, two connected nodes imply that they are related and affect each other. The relationship between nodes is represented by 1 and 0. One indicates that the nodes affect each other and a line is formed between them. In addition, zero means that the two nodes do not influence each other and do not form a line between them [23]. For example, Node2#, the physical examination pass rate, influences Node4#, operational proficiency (a high physical examination pass rate has a positive influence on an employee’s operational proficiency); thus, they are connected with a line in the network and given the number 1.

The node order and corresponding evaluation index are shown in

Table 1. Node order and corresponding evaluation index.

Node Order	Evaluation Index	Node Order	Evaluation Index
1	Personnel exposure density	25	Equipment assets
2	The physical examination pass rate	26	Lifeline system score
3	Psychological quality	27	Material properties
4	Operational proficiency	28	Radioactive contamination management
5	A high degree of education	29	Solid waste management
6	Safety assessment pass rate	30	Storage tank resilience
7	The average length of service	31	Emergency plan
8	Maximum working life	32	Emergency organization and responsibility
9	Age structure	33	Emergency supplies
10	Organizational structure completeness	34	Emergency response
11	The implementation rate of safety education	35	Environmental exposure density
12	Safety management perfection	36	Vegetation coverage
13	Equipment exposure	37	Vegetation types
14	Average remaining pipeline life	38	Soil runoff rate
15	Replacement frequency of valve accessories	39	Average slope
16	Breather valve qualification rate	40	Distance from water
17	Separate clean and dirty rainwater	41	Altitude
18	Meter waterproof rate	42	Corporate reputation building
19	Instrument accuracy	43	Degree of soil corrosion
20	Initial rainwater treatment capacity	44	Disaster recovery plan
21	Clean rainwater treatment capacity	45	Damage assessment
22	Hazard resistance of accident pool	46	Economic bearing capacity
23	Standard discharge of pollutants	47	Fire rescue force
24	High-temperature working area blocking rate

.

(2)

Establishing a complex network

According to the node relation data provided by the adjacency matrix, we established a complex network using MATLAB R2018a (Figure 4) and obtained the distribution diagram of the node degree (Figure 5).

(3)

Final approval

It can be observed from the complex network and node distribution graph that there are many connections between surrounding nodes in part of the network, such as the node group around Node 4 and the node groups around Nodes 10 and 20. Calculating the node’s degree centrality can confirm whether a node is important and, furthermore, expert opinions can be combined to obtain an appropriate evaluation index for heavy rainfall vulnerability assessment in coastal oil depots. Degree centrality states that the more nodes near a node, the greater the influence of the node, and the more important it is [24]. A degree-centered method was used for the comparison of different nodes.

$$\begin{gathered} D{C}_i=\frac{K_i}{N-1} \\ D{C}_i=\frac{K_i}{N-1} \end{gathered}$$

(1)

where $D{C}_i$ represents degree centrality, $K_i$ represents the degree of node i, N is the number of nodes in the complex network, and N − 1 is the maximum possible degree value of the node [25,26]. The degree centrality of nodes is shown in

Table 2. Degree centrality of nodes.

Node order	1	2	3	4	5	6
Degree centrality	0.000	0.217	0.217	0.304	0.065	0.174
Node order	7	8	9	10	11	12
Degree centrality	0.130	0.130	0.022	0.261	0.152	0.261
Node order	13	14	15	16	17	18
Degree centrality	0.000	0.087	0.087	0.022	0.065	0.065
Node order	19	20	21	22	23	24
Degree centrality	0.065	0.326	0.304	0.087	0.130	0.022
Node order	25	26	27	28	29	30
Degree centrality	0.000	0.109	0.022	0.022	0.022	0.196
Node order	31	32	33	34	35	36
Degree centrality	0.130	0.174	0.152	0.217	0.000	0.065
Node order	37	38	39	40	41	42
Degree centrality	0.022	0.043	0.043	0.065	0.022	0.022
Node order	43	44	45	46	47
Degree centrality	0.109	0.130	0.065	0.087	0.087

.

It can be seen from the above Table 2 that the degree centrality of some nodes is low or even close to zero. Nodes with a degree centrality lower than 0.022 considered with very low influence. Simultaneously, key nodes such as nodes 4, 10, 20, and 21 are reserved. However, due to the defects of the complex network itself, the test results sometimes conform to the complex network theory but are not quite realistic. Therefore, it is necessary to ask experts for final approval. By combining expert opinions and the selection of the complex network, the necessary nodes were maintained and the final evaluation index of exposure vulnerability was obtained, as shown in

Table 3. Evaluation index and weight of heavy rainfall vulnerability in coastal oil depots.

First-Level Indicators	Second-Level Indicators	Third-Level Indicators
Personnel vulnerability index A (0.295)	Personnel exposure A1 (0.163)	Personnel exposure density A11 (1)
	Personnel sensitivity A2 (0.297)	Physical examination pass rate A21 (0.066)
		Psychological quality A22 (0.068)
		Operational proficiency A23 (0.183)
		A high degree of education A24 (0.106)
		Safety assessment pass rate A25 (0.187)
		The average length of service A26 (0.288)
		Maximum working life A27 (0.102)
	Personnel management response A3 (0.540)	Organizational structure completeness A31 (0.297)
		The implementation rate of safety education A32 (0.163)
		Safety management perfection A33 (0.540)
Equipment facility vulnerability index B (0.356)	Equipment facility exposure B1 (0.170)	Equipment exposure B11 (0.50)
		Equipment assets B12 (0.50)
	Equipment facility sensitivity B2 (0.443)	Replacement frequency of valve accessories B2 (0.056)
		Breather valve qualification rate B22 (0.061)
		Separate clean and dirty rainwater B23 (0.141)
		Meter waterproof rate B24 (0.039)
		Instrument accuracy B25 (0.039)
		Initial rainwater treatment capacity B26 (0.128)
		Clean rainwater treatment capacity B27 (0.128)
		Hazard resistance of accident pool B28 (0.066)
		Standard discharge of pollutants B29 (0.116)
		Average remaining pipeline life B210 (0.042)
		Lifeline system score B211 (0.046)
		Storage tank resilience B212 (0.138)
	Equipment facility management response capability B3 (0.387)	Emergency plan B31 (0.25)
		Emergency organization and responsibility B32 (0.25)
		Emergency supplies B33 (0.25)
		Emergency response B34 (0.25)
Environmental vulnerability indicator C (0.251)	Environmental exposure C1 (0.250)	Environmental exposure density C11 (1)
	Environmental sensitivity C2 (0.750)	Vegetation coverage C21 (0.092)
		Soil runoff rate C22 (0.179)
		Average slope C23 (0.197)
		Distance from water C24 (0.229)
		Corporate reputation building C25 (0.196)
		Degree of soil corrosion C26 (0.107)
Resilience D (0.098)	Resilience D (1)	Disaster recovery plan D1 (0.161)
		Damage assessment D2 (0.103)
		Economic bearing capacity D3 (0.368)
		Fire rescue force D4 (0.368)

.

“Personnel vulnerability-A” is one of the four first-level indicators; it is composed of three second-level indicators, which are “Personnel exposure-A1”, “Personnel sensitivity-A2”, and “Personnel management response-A3”. “Personnel exposure-A1” has only one third-level indicator, “Personnel exposure density-A11” while A2 has seven third-level indicators. The number in brackets is the weight of that indicator under the upper-level indicator.

3.2. Weight Definition

The AHP method was used to obtain the judgment matrix from the experts’ scores. The standard score is from Satty’s proportional scale. From multiple weight value schemes, we calculated and compared the information entropy of several weight value schemes and found the best scheme with the least amount of unknown information E [27].

(1)

Establish the judgment matrix of the evaluation index

Establishing a judgment matrix to evaluate the index of the personnel, equipment and facilities, environmental vulnerabilities, and resilience according to the structure of the index system is as follows:

$$A=\left[\begin{array}{c}{a}_{11}\\ \dots \\ a_{n1}\end{array}\begin{array}{c}{a}_{12}\\ \dots \\ a_{n2}\end{array}\begin{array}{c}\dots \\ \dots \\ \dots \end{array}\begin{array}{c}{a}_{1n}\\ \dots \\ a_{nn}\end{array}\right]$$

(2)

where A is the judgment matrix, $a_{ij}$ is the index i’s importance degree compared to index j, n represents the order of the matrix and the importance degree is as per Satty’s scale method. The index’s significance ratio is 1.

(2)

Perform consistency check

The judgment matrix can be used only after passing the consistency check and when the comparison test coefficient CR is less than 0.1, the matrix passes the consistency check. Otherwise, a new matrix must be established [28]. The consistency index of the matrix CI can be calculated as follows:

$$CI=\frac{{\lambda }_{\max }\left(A\right)-n}{n-1}$$

(3)

After obtaining the RI value through the table query, the comparison test coefficient can be calculated as follows:

$$CR=CI/RI$$

(4)

(3)

Generate weight scheme

The total information of the index measure $I_a$ was obtained through the composition of vulnerability assessment indicators. Similarly, through a series of control measures and the implementation of engineering measures, the uncertainty was reduced before the implementation of measures was obtained, and the amount reduced was represented as $I_b$ [29]. Risk is the potential before the implementation of regulatory measures or risk management and is subject to the inevitable constraint information amount E under certain constraints. The essence of risk is to display and map an intrinsic measure of risk. When $I_a$ is certain, the larger the value of E, the smaller the value of $I_b$ the fewer are the functions of risk control and management measures to the risk itself, and the higher is the risk [30,31].

$$I_a=\log \left(n\right)$$

(5)

$$I_b={\displaystyle \sum }_{i=1}^n-p_i\log \left(p_i\right)$$

(6)

$$E=I_a-I_b$$

(7)

where n represents the number of evaluation indexes and $p_i$ represents the node’s weight.

The maximum Eigen, weight, CR, and E values were obtained using MATLAB R2018a. For example, the first-level indicator weight scheme (

Table 4. First-level indicator weight scheme 1.

M	A	B	C	D	CR	E	Weighted Value
A	1.000	1.000	1.000	3.000	0.022	0.121	0.295
B	1.000	1.000	2.000	3.000			0.356
C	1.000	0.500	1.000	3.000			0.251
D	0.333	0.333	0.333	1.000			0.098

,

Table 5. First-level indicator weight scheme 2.

M	A	B	C	D	CR	E	Weighted Value
A	1.000	0.500	3.000	3.000	0.045	0.209	0.311
B	2.000	1.000	3.000	3.000			0.439
C	0.333	0.333	1.000	2.000			0.146
D	0.333	0.333	0.500	1.000			0.104

and

Table 6. First-level indicator weight scheme 3.

M	A	B	C	D	CR	E	Weighted Value
A	1.000	0.333	2.000	3.000	0.053	0.268	0.242
B	3.000	1.000	3.000	4.000			0.502
C	0.500	0.333	1.000	3.000			0.172
D	0.333	0.250	0.333	1.000			0.084

) chooses the scheme with the smallest E.

The weight of each index was calculated according to the above steps to obtain the weighting scheme (Table 3).

3.3. Comprehensive Evaluation

A weighted comprehensive evaluation is one of the most used methods in a comprehensive evaluation. We consider this to be a comprehensive evaluation of heavy rainfall vulnerability in coastal oil depots.

$$F_i=D_i/X$$

(8)

$$L={\displaystyle \sum }_{n=1}^n{q}_i\times F_i$$

(9)

where $D_i$ is the actual score of the third-level indicators, X is the full score, $F_i$ is the status value and the value range is [0, 1], n is the number of indicators in a layer, $q_i$ is the weight value of the corresponding index, and L is the score status value of the layer index. The final score is obtained in the same manner. For example, the A2 index, after calculation, scored 0.81 (

Table 7. Score table of A2 and its subordinate indexes.

Second-Level Indicators	Weighted Value	Third-Level Indicators	Weighted Value	Full Score	Actual Score	Status
A2	0.297	A21	0.066	10	9.000	0.900
		A22	0.068	10	7.000	0.700
		A23	0.183	10	8.000	0.800
		A24	0.106	10	6.000	0.600
		A25	0.187	10	9.000	0.900
		A26	0.288	10	8.000	0.800
		A27	0.102	10	9.000	0.900

).

3.4. Classification of Vulnerability

The classification of vulnerability is based on the obtained vulnerability and the value range of vulnerability is [0, 1]. The specific analyses are presented in

Table 8. Heavy rainfall vulnerability classification table.

Range	Exposure Vulnerability Degree	Pre-Warning Level	Identification
[0, 0.3]	Exposure’s disaster resistance is poor. There are big defections of disaster-resistant ability in all four aspects of personnel, equipment and facilities, environmental vulnerabilities, and resilience. The exposure is highly vulnerable.	4	Red
[0.3, 0.6]	Exposure’s disaster resistance is general. There are big defections of disaster-resistant ability in some of the four aspects of personnel, equipment and facilities, environmental vulnerabilities, and resilience. The exposure is more vulnerable.	3	Orange
[0.6, 0.9]	Exposure’s disaster resistance is good. There are no defections of disaster-resistant ability in all four aspects of personnel, equipment and facilities, environmental vulnerabilities, and resilience. The exposure is less vulnerable.	2	Yellow
[0.9, 1]	Exposure’s disaster resistance is excellent. All four aspects of personnel, equipment and facilities, environmental vulnerabilities, and resilience work very well. The exposure is very less vulnerable.	1	Blue

.

4. Conclusions

Research on the risk management of coastal oil depots during heavy rainfall disasters is still relatively weak. In this study, we constructed an emergency management system suitable for coastal oil depots by studying their vulnerability to heavy rainfall events. The damage caused to coastal oil depots by heavy rainfall was analyzed to determine the factors affecting the vulnerability of disaster-bearing bodies.

An evaluation index of heavy rainfall vulnerability in coastal oil depots was established using a questionnaire survey, expert scoring, and complex network theory combined with information entropy theory. Using the comprehensive evaluation method with three levels of indicators, it was found that “safety management perfection-A33” holds the highest weight in the evaluation system, which means the oil depot operating corporation should this give a high priority and carry out continuous improvement activities. A final score of risk assessment was gained and could be presented visually with its prewarning level to an oil depot corporation.

Considering that most of the coastal oil depots are located in industrial parks or remote island areas, connecting with outside emergency rescue responders is crucial in responding to natural disasters. With the wide application of mobile social networks in different industries, social network technology (WeChat is more commonly used in China) can be applied to send a message including precipitation trends, safety warning levels, etc., to the workgroup, composed of the emergency management department of the oil depot company, local rescue teams, and local government, to obtain a fast transmission of information and more effective cooperation under extreme weather conditions. When a disaster occurs, emergency procedures can be triggered in a timely manner to minimize the damage and impact caused by a natural hazard.