Large-scale crowd evacuation is an important measure guaranteeing the safety of disaster-stricken victims in typhoon relief activities. Decision-making related to antityphoon crowd evacuation must take full consideration of the destructive effect of typhoons and their secondary disasters, time urgency, and resource limitation. To give full play to limited vehicle resources, the influence of a typhoon and its secondary disasters on antityphoon evacuation are mainly manifested during the execution of evacuation tasks in this article. The shortest time spent in completing all evacuation tasks was taken as the objective. Then, a vehicle route selection model for two-phase large-scale antityphoon crowd evacuation was built under an uncertain environment, and a matrix encoding–based genetic algorithm was designed to solve the model. Under the background of Super Typhoon Meranti in 2016, the model and algorithm were applied to crowd evacuation in a typhoon in Xiamen for a simulated analysis. Results indicate that in typhoon relief activities, emergency decision makers can use the proposed method to acquire a scientific and reasonable route selection scheme for antityphoon crowd evacuation according to related typhoon disaster data.
Typhoon disasters are one of the most serious natural disasters affecting human life. Research has shown that typhoons not only bring about direct disasters such as gales, rainstorms, and storm tides. Typhoons also give rise to derivative disasters such as torrential floods, landslides, debris flows, and waterlogging, thereby causing major loss of life, personal injuries, and property losses in the disaster chain [
Typhoons that have landed in China since 2000 and losses they have caused.
Typhoon name/no. | Landing area | Maximum wind speed at the center (m/s) | Number of dead or missing persons | Number of urgently evacuated persons | Direct economic loss (RMB) |
---|---|---|---|---|---|
Yunna/200413 | Zhejiang | 45 | 186 | 468,000 | 18.121 billion |
Saomei/200608 | Zhejiang | 60 | 483 | 1,559,000 | 19.44 billion |
Haikui/201211 | Zhejiang | 48 | 6 | 2,173,000 | 2.09 billion |
Rammasun/201409 | Hainan, Guangdong, and Guangxi | 60 | 83 | 862,000 | 38.48 billion |
Maria/201808 | Fujian and Zhejiang | 58 | 0 | 616,800 | 3 billion |
Mangkhut/201822 | Guangdong, Guangxi, and Hainan | 62.8 | 6 | 1,601,000 | 5 billion |
Data sources: National Climate Center of China, China Meteorological Administration, and China Weather Typhoon Net.
A number of scholars have studied emergent traffic evacuation in sudden major disasters, such as earthquakes [
Factors such as evacuation shelter selection and evacuation route selection are mainly studied in strategic evacuation planning. Robinson and Khattak [
In evacuation simulation analysis, Williams et al. [
To sum up, a number of existing studies on large-scale antityphoon crowd evacuation have explored antityphoon evacuation behaviors, strategies, and simulations. Several studies on crowd evacuation route selection under super typhoon disasters are carried out under certain environments, but only a few studies have considered the influence of secondary disasters brought about by typhoons and disaster chains. Several decision-making method studies for antityphoon crowd evacuation vehicle routes assume that vehicles are used to evacuate people only between fixed retrieval depots and disaster-stricken points. Vehicle quantity restriction is used to express the evacuation ability constraint of retrieval depots without considering the dynamic driving process of evacuation vehicles in the evacuation process. Therefore, to avoid contradictions among typhoon destructiveness, time urgency, and resource limitation, this article considered the influence of typhoons and their secondary disasters on difficulties in evacuation vehicle transport. Given that evacuation vehicles can arrive at random retrieval depots and disaster-stricken sites to execute evacuation tasks, the shortest time spent in completing all evacuation tasks is taken as the objective. A vehicle route selection model for large-scale antityphoon crowd evacuation, which allowed evacuation vehicles to randomly select evacuation sites to execute their evacuation tasks under uncertain environments, was established. A matrix encoding–based genetic algorithm was designed to solve the model. Finally, against the background of Super Typhoon Meranti in 2016, the established model and algorithm were applied to large-scale antityphoon crowd evacuation scenarios for a simulation analysis.
As shown in Figure
Schematic diagram of two-phase antityphoon crowd evacuation.
Moreover,
Vehicles can depart from affected areas, temporary shelters for victims, and hospitals to carry out evacuation tasks, and they are allowed to go to any affected area or temporary shelter to execute their tasks.
After a vehicle completes its current evacuation task, another affected area or temporary shelter can be randomly selected to carry out the next evacuation task.
Vehicles in affected areas and temporary shelters for victims give priority to carrying out evacuation tasks in Phase I.
Vehicles like hospital ambulances only transport victims who need medical rescue and give priority to executing evacuation tasks in Phase II.
Influences of typhoon disasters and their secondary disasters, like debris flow, landslide, and waterlogging, on evacuation tasks are mainly manifested by damaged roads that are used for evacuation. By-pass, rush repair, and other measures should be adopted for inaccessible roads.
Driving time for vehicles to and from affected areas, temporary shelters for victims, and hospitals is influenced by a typhoon and its secondary disasters depending on the damage degree of this route. A high degree of damage requires a long driving time.
Departure is defined as the journey of transporting victims from affected areas to temporary shelters or hospitals. It also refers to the journey of transporting people in need of medical rescue from temporary shelters to hospitals.
Return is defined as the journey to affected areas or temporary shelters to transport victims and the journey to affected areas from temporary shelters to execute evacuation tasks.
Road impedance coefficient expresses the influence of a typhoon and its secondary disasters on running speed of evacuation vehicles, and its value is in direct proportion to the degree of road damage. The formula of road impedance coefficient from point
Symbols used in this article are shown in Table
Symbols.
|
||||
|
Set of affected areas, |
|
Set of temporary shelters, |
|
|
Set of medical rescue points, |
|
Set of vehicles, |
|
|
Set of points passed by the |
|||
|
||||
|
||||
|
Capacity of the |
|
Quantity of people to be evacuated from the affected area |
|
|
Quantity of victims who can be accommodated by temporary shelter |
|
Quantity of injured people who can be accommodated by hospital point |
|
|
Total departure time of the |
|
Total return time of the |
|
|
Average running speed of the |
|
Route distance from point |
|
|
Proportion of casualties needing medical rescue at the affected area |
|
Time needed for the |
|
|
||||
|
||||
|
Road damage rate | |||
|
Road damage rate on the route from point |
|||
|
||||
|
||||
|
Number of people (who do not need medical rescue from hospitals) who are transported by the |
|||
|
Number of people (who need medical rescue from hospitals) who are transported by the |
|||
|
Proportion occupied by the number of people (who do not need medical rescue from hospitals) who are transported by the |
|||
|
Proportion occupied by the number of people (who need medical rescue from hospitals) who are transported by the |
|||
|
Evacuation scheme for the |
|||
|
Driving route selection plan for the |
The minimum time needed to complete all evacuation tasks, specifically minimizing the time used by vehicles to execute evacuation tasks, is taken as the objective. The route decision-making model M for large-scale antityphoon crowd evacuation vehicles is as follows:
In 1959, Dantzig and Ramser [
In full consideration of two pragmatic encoding rules proposed by Dejong [
Chromosome structure.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
To improve the algorithm efficiency, assuming that in the departure journey, none of the victims transported by vehicles need medical assistance in hospitals, the value function of the proportion occupied by the number of people transported by the
The number of chromosome rows in the matrix is the number of vehicles, and the number
Two-column crossing method is used, and operation steps are as follows. Genetic codes from column Constraint equations ( Otherwise, the feasible operation will be executed.
Row mutation operation method is used, and the operation steps are as follows. Genetic codes of two rows are randomly generated to replace genetic codes of row Constraint equations ( If not, feasible operation will be executed.
Expected value method and optimal individual retention strategies are adopted.
Both subindividuals generated by crossover and mutation operations may generate unfeasible solutions. Under this circumstance, an initial genetic individual is randomly generated to replace this unfeasible individual.
The study results of Jong (1975) indicate a good circumstance when population size is the same as or twice the size of the chromosome length. In full consideration of population diversity and computational efficiency, population size is selected as twice the chromosome length in this article.
The maximum number of iterations is set as
On September 10, 2016, Meranti was generated on the surface of the Northwest Pacific Ocean. Meranti landed in Taiwan and Xiamen (Fujian Province) in China at a super typhoon scale on September 15 with a maximum wind power of 52 m/s, inflicting major losses in Xiamen. Typhoon Meranti caused 28 deaths, 49 injuries, and 18 people to go missing after its landfall in mainland China. A total of 655,500 people were urgently evacuated in nine municipal-level cities and Pingtan comprehensive experimental zone and 86 counties (cities and districts) in the Fujian Province. Figure
Schematic antityphoon evacuation graph during Super Typhoon Meranti. Note: The map was derived from Wenzhou Typhoon Net (
Each evacuation vehicle departing from hospitals could accommodate 10 people at most, and each vehicle departing from other points could accommodate 50 people at most. The proportion of typhoon-stricken victims who needed accept medical assistance in hospitals was 10%, the road impedance coefficient of vehicles departing from hospitals was 1.5. The road impedance coefficient of vehicles departing from affected areas and temporary shelters for victims was 1, namely,
The road impedance coefficient follows uniform distribution from 0 to 1, namely:
Quantities of victims and vehicle resources in affected areas are shown in Table
Numbers of to-be-evacuated victims and quantities of vehicle resources in affected areas.
DD | PB | BH | XK | |
---|---|---|---|---|
Number of to-be-evacuated victims | 600 | 300 | 400 | 400 |
Number of vehicles | 2 | 1 | 1 | 1 |
Total number of victims and quantities of vehicle resources that can be accommodated by temporary shelters.
XS | XA | XUT | |
---|---|---|---|
Total number of victims who can be accommodated | 600 | 900 | 500 |
Number of vehicles | 1 | 2 | 1 |
Total number of patients and quantities of vehicle resources that can be accommodated by hospitals.
N2HX | N5HX | |
---|---|---|
Total number of patients who can be accommodated | 100 | 100 |
Number of vehicles | 2 | 1 |
Average vehicle running time and distance among affected areas, temporary shelters for victims, and hospitals under normal circumstances (min/km).
XS | XA | XUT | N2HX | N5HX | |
---|---|---|---|---|---|
DD | 26/18 | 21/13 | 47/39 | 45/33 | 30/22 |
PB | 25/14 | 19/9 | 43/30 | 40/30 | 26/19 |
BH | 28/17 | 17/11 | 35/25 | 35/25 | 23/17 |
XK | 36/25 | 22/14 | 20/12 | 20/14 | 21/10 |
XS | 0 | 18/10 | 42/36 | 38/32 | 25/15 |
XA | 19/12 | 0 | 29/22 | 27/21 | 13/7 |
XUT | 44/31 | 31/21 | 0 | 16/7.5 | 35/21 |
By using the proposed model and algorithm, and using MATLAB R2012a as the solver engine, we obtain the evacuation plan. The entire procedure took 12 minutes. We also recorded the best individuals of each generation. As shown in Figure
Total evacuation time during 2,000 generations.
Through 2,000 times of iteration, the optimal evacuation strategy with the total evacuation time of 376 minutes is obtained, and the concrete routes of 12 vehicles are shown in Table
Concrete routes of vehicles (time unit: min).
Route | Time | |
---|---|---|
DD1 | DD ⟶ N.5HX ⟶ DD ⟶ XS ⟶ DD ⟶ XS | 306 |
DD2 | DD ⟶ XS ⟶ DD ⟶ XS ⟶ DD ⟶ XS | 290 |
PB1 | PB ⟶ XA ⟶ PB ⟶ XA ⟶ PB ⟶ XA ⟶ PB ⟶ XA | 306 |
BH1 | BH ⟶ XA ⟶ BH ⟶ XA ⟶ BH ⟶ XA ⟶ BH ⟶ XA ⟶ BH ⟶ XA | 356 |
XK1 | XK ⟶ XUT ⟶ XK ⟶ XUT ⟶ XK ⟶ XUT ⟶ XK ⟶ XUT | 320 |
XS1 | XS ⟶ DD ⟶ XS ⟶ DD ⟶ XS ⟶ DD ⟶ XS | 342 |
XA1 | XA ⟶ PB ⟶ XA ⟶ BH ⟶ XA ⟶ BH ⟶ XA ⟶ BH ⟶ XA ⟶ N.5HX | 356 |
XA2 | XA ⟶ DD ⟶ XA ⟶ DD ⟶ XA ⟶ DD ⟶ XA ⟶ DD ⟶ XA | 376 |
XUT1 | XUT ⟶ XK ⟶ XUT ⟶ XK ⟶ XUT ⟶ XK ⟶ XUT ⟶ XK | 360 |
N.2HX1 | N.2HX ⟶ XUT ⟶ N.2HX ⟶ XUT ⟶ N.2HX ⟶ XUT | 258 |
N.2HX2 | N.2HX ⟶ XUT ⟶ N.2HX | 106 |
N.5HX1 | N.5HX ⟶ XA ⟶ N.5HX ⟶ XA ⟶ N.5HX ⟶ XA ⟶ N.5HX | 264 |
Number of people evacuated in affected areas, temporary shelters for victims, and hospitals.
XS | XA | XUT | N2HX | N5HX | |
---|---|---|---|---|---|
DD | 350 | 200 | 0 | 0 | 50 |
PB | 0 | 300 | 0 | 0 | 0 |
BH | 0 | 400 | 0 | 0 | 0 |
XK | 0 | 0 | 400 | 0 | 0 |
XS | — | — | — | 0 | 0 |
XA | — | — | — | 0 | 80 |
XUT | — | — | — | 40 | 0 |
Existing literature has developed the scheduling model by minimizing the total evacuation time [
By using the same data from Section
Concrete routes of vehicles (time unit: min).
Route | Time | |
---|---|---|
DD1 | DD ⟶ XS ⟶ DD ⟶ XS ⟶ DD ⟶ XS ⟶ DD ⟶ XS ⟶ DD ⟶ XS ⟶ DD ⟶ XS | 572 |
DD2 | DD ⟶ XA ⟶ DD ⟶ XA ⟶ DD ⟶ XA ⟶ DD ⟶ XA ⟶ DD ⟶ XS ⟶ DD ⟶ XS | 492 |
PB1 | PB ⟶ XA ⟶ PB ⟶ XA ⟶ PB ⟶ XA ⟶ PB ⟶ XA ⟶ PB ⟶ XA ⟶ PB ⟶ XA | 456 |
BH1 | BH ⟶ XA ⟶ BH ⟶ XA ⟶ BH ⟶ XA ⟶ BH ⟶ XA ⟶ BH ⟶ XA ⟶ BH ⟶ XA ⟶ BH ⟶ XA ⟶ BH ⟶ XA | 544 |
XK1 | XK ⟶ XUT ⟶ XK ⟶ XUT ⟶ XK ⟶ XUT ⟶ XK ⟶ XUT ⟶ XK ⟶ XUT ⟶ XK ⟶ XUT ⟶ XK ⟶ XUT | 672 |
XS1 | XS ⟶ N.5HX | 50 |
XA1 | XA ⟶ N.5HX | 26 |
XA2 | — | 0 |
XUT1 | XUT ⟶ N.2HX | 32 |
N.2HX1 | — | 0 |
N.2HX2 | — | 0 |
N.5HX1 | — | 0 |
Comparison of the results of two tests (time unit: min).
Total evacuation time | Time needed to complete all evacuation tasks | |
---|---|---|
Results with equation ( |
3,640 | 376 |
Results with equation ( |
2,844 | 672 |
Super typhoons are one of the most common disasters that affect human life. Nearly every year, people die of typhoon disasters. To maximize limited vehicle resources within a limited time period, road impedance coefficient was introduced in this article to embody the influences of typhoons and their secondary disasters on evacuation tasks. The vehicle road impedance coefficient was used to embody the abilities of different vehicles in executing evacuation tasks. Evacuation vehicles were allowed to select any affected area or temporary shelter in the task execution process. To minimize the time needed by evacuation vehicles to complete evacuation tasks, a route selection model for antityphoon crowd evacuation vehicles was built, and a corresponding genetic algorithm was designed to solve it. Moreover, a large-scale crowd evacuation simulation of Super Typhoon Meranti indicated that the proposed method can provide emergency decision makers with a scientific and reasonable route selection scheme for antityphoon crowd evacuation vehicles.
The influence of typhoons and their secondary disasters are important factors in the execution of antityphoon crowd evacuation tasks, and their influence were also embodied by vehicle running time in this article. However, to-be-evacuated crowd structure, geological information, and geographic information in a disaster-stricken area and evacuation cost are also important factors influencing evacuation decision-making. Other related factors can be taken into consideration in future research to improve decision-making methods.
The data used to support the findings of this study are available from the corresponding author upon request.
The authors declare that they have no conflicts of interest.
This work was supported by the Zhejiang Provincial Philosophy and Social Science Program of China (grant no. 16ZJQN025YB), the National Natural Science Foundation of China (NSFC) (grant nos. 71601146, 71872131, and 71603237), the Zhejiang Provincial Philosophy and Social Science Program of China (grant no. 18NDJC198YB), and Zhejiang Provincial Natural Science Foundation of China (grant nos. LQ16G010005 and LY19G030004).