CN115239110A - Navigation risk evaluation method based on improved TOPSIS method - Google Patents

Abstract

The invention discloses a navigation risk evaluation method based on an improved TOPSIS method, which is characterized in that a K-means clustering method based on minimum circle coverage is utilized to identify and divide accident-prone areas in a researched sea area, then traffic factors, ship factors and channel conditions are comprehensively considered to establish a relative navigation risk evaluation index system, index weights obtained by a hierarchical analysis method and an entropy method are combined through a game theory to obtain a combined weight of each index, finally, the relative navigation risk of each accident-prone area is evaluated by utilizing the improved TOPSIS method, the average risk is used as a risk threshold value to judge the position of each high-risk area, the accuracy of navigation risk evaluation is improved, a basis and a basis are provided for marine emergency rescue resource allocation research, a decision reference is provided for marine safety management work, and relevant managers can perform corresponding human and material resource allocation according to the calculated risk.

Description

A Navigation Risk Assessment Method Based on Improved TOPSIS Method

technical field

The invention belongs to the technical field of ship navigation risk assessment in sea areas, and in particular relates to a navigation risk assessment method based on an improved TOPSIS method.

Background technique

In recent years, with the continuous development of the economy, the throughput of my country's ports has increased rapidly, and the safety of ship navigation has been paid more and more attention. The safety of ship navigation is affected by many factors. Systematically analyze and screen to determine the key factors that can reflect the risk level, establish a risk evaluation index system, evaluate each index with the help of qualitative or quantitative models, and obtain the risk value that can reflect the overall risk level of the ship, and use this as the safety of the ship. Management provides decision support. The main navigation safety hazards in the sea area are as follows:

(1) Influence of bad weather and sea conditions

Bad weather and sea conditions such as typhoons and big waves, as one of the factors that cause marine traffic accidents, are hidden dangers that can never be eliminated. On the one hand, when sailing under bad weather and sea conditions such as strong winds and waves, the maneuverability of the ship is easily affected, especially for small ships, under such conditions, the risky sailing is particularly prone to ship capsizing accidents. On the other hand, if a ship has an accident under adverse weather and sea conditions such as strong winds and waves, its emergency rescue environment will be more difficult. It is also a great challenge to carry out emergency rescue under such harsh conditions, which will greatly reduce the efficiency of emergency rescue, greatly reduce the effect of emergency rescue, and make the consequences of the accident more serious.

(2) Navigation risk of ro-ro passenger ships and dangerous goods ships

Ro-ro passenger ships are the main means of transport for people crossing the sea. Once an accident occurs, the lives of all people on board will be threatened. Based on the basic idea of people-oriented, the area where the ro-ro passenger route is located has always been one of the focus of maritime safety supervision in the Strait. And with the increase in the scale of each port and the number of ships allocated, as the demand for inbound and outbound ports increases, the density of ships in the area where the ro-ro passenger route passes is increasing, and the probability of encountering ships will greatly increase, and the probability of collision will increase accordingly. Moreover, in addition to the passenger-rolling terminal, the strait also includes a dangerous goods terminal. Therefore, the proportion of dangerous goods ships in the strait is relatively high. Oil tankers, LNG ships and other dangerous goods ships are very likely to cause large-scale environmental pollution in the sea area after an accident. When emergency treatment is carried out, a large area of traffic control is required, which will affect the navigation efficiency of the strait and the impact on various ports. have a very negative impact on the operation. In addition, dangerous goods ships in distress are prone to explosions, fires and other accidents, which not only pose a great threat to the safety of the people on board and the ship itself, but also threaten the ships in the surrounding waters.

(3) Influence of obstructions such as reefs, sunken ships, and fishing boat activities

The water depth span of the strait is large, and ships are prone to hitting rocks and grounding accidents at points where the water depth changes sharply and reefs. In addition, there are many shoals in the waters of the strait. If there is a typhoon, the navigation mark will be displaced, which will mislead the ship's driver and cause the ship to enter the shoal area. The ship is also very likely to run aground or hit the bottom. ACCIDENT. In addition, due to the drastic undulating topography of the seabed of the strait, the complex flow pattern, and the high flow velocity in individual areas, sometimes the sunken ship cannot be dealt with in time after the ship sinks or capsizes, which will also have a certain impact on the passage of ships. The strait is the breeding ground for the aquaculture and fishing industry. However, due to the relatively small size of the fishing boats and the fact that most of the fishing boats are not equipped with the AIS system as required, other ships in the surrounding area can sometimes only use radar and visual methods to track the fishing boats. It is judged that if the fishing boat is not paid attention to, it will easily lead to a collision accident with the fishing boat, and generally speaking, the personnel driving the fishing boat are fishermen who make a living by fishing. Different from professional captains and drivers, these fishermen may not Having received professional training and lacking certain safety awareness and emergency response capabilities, in the event of an emergency, it is easier to panic, resulting in aggravating the consequences. In addition, through the analysis of the accident and danger situation in the strait, it is found that there are many accidents due to factors such as fishing nets and fishing fences. To sum up, the activities of fishing boats are also a major hidden danger to the safety of navigation in the strait.

In the current research, there is no method for assessing the risk of marine navigation for the above-mentioned hidden dangers of navigation safety in the sea area; the current research on navigation risk in the water area mainly focuses on inland waters or the overall situation of a single water area. There are still some differences between sea area and inland river research, such as ship navigation rules and environment, availability of research data, etc., and the specific conditions of each sea area are different, so it should be based on the characteristics of the research object to find a suitable index system. study it with research methods.

SUMMARY OF THE INVENTION

The purpose of the present invention is to provide a navigation risk assessment method based on the improved TOPSIS method for the problems existing in the prior art.

For achieving the above object, the technical scheme adopted in the present invention is:

A navigation risk assessment method based on the improved TOPSIS method, comprising the following steps:

S1, using the K-means clustering method based on the minimum circle coverage to identify and divide the accident-prone areas in the research sea area;

S2, comprehensively consider traffic factors, ship factors and channel condition factors to establish a relative navigation risk evaluation index system;

S3, using AHP to determine the subjective weight vector W ₁ of each index;

S4, using the entropy weight method to determine the objective weight vector W ₂ of each index;

S5, adopt game theory method to combine described subjective weight vector and objective weight vector, obtain the combined weight vector W ^* of each index;

S6, use Mahalanobis distance instead of Euclidean distance and grey correlation degree to calculate the improved TOPSIS method to evaluate the relative navigable risk of accident-prone areas, and use the average risk as the risk threshold to determine the location of each high-risk area.

Specifically, step S1 includes the following steps:

S101, obtain all marine traffic accident information in the study area for a period of time, extract the latitude and longitude coordinates of each accident point, and use ArcGis software to draw a spatial distribution map of marine traffic accidents in the study area;

S102, based on the idea of minimum circle coverage, draw a number of circles on the spatial distribution map of maritime traffic accidents in the study area to cover the accident points to distinguish each clustering area, one circle represents one clustering area, and each clustering area At least two accident points are included, and isolated accident noise points are identified and eliminated according to the clustering results;

S103 , using the number of all circles as the number of clusters, using the centers of each circle as the cluster center of each clustering area, and using the K-means clustering method to obtain a final result of spatial division of areas prone to marine accidents.

Specifically, in step S2, the relative navigation risk evaluation index system includes:

Traffic factors, the traffic factors include the following indicators: ship flow (vessel/hour), average ship density (vessel) and ship density dispersion; among them, the ship flow is the conversion of standard ship flow, that is, the ship is converted according to the ship length conversion factor table After that, make statistics; the unit of statistical ship density is the grid enclosed by each longitude and each latitude on the chart;

Vessel factors, which include the following indicators: Vessel Type (%) and Vessel Speed (knots);

Channel condition factors, the channel condition factors include the following indicators: the impact of special areas, the number of special points in the channel (number) and the number of obstructions (number); the impact of the special area includes docks, fishing areas, anchorages, shoals and The overall impact of the reef; the number of special points in the channel includes the total number of channel ends, track intersections and bends.

Specifically, in step S3, the method for determining the subjective weight vector of each index by using AHP is:

According to the survey results of the intentions of experts, marine practitioners, marine safety supervisors and other personnel on the importance of selecting indicators, take the geometric mean to obtain the indicator importance judgment matrix X=(x _ij ) _n×n ; calculate the elements of each row of the judgment matrix X the geometric mean of

Among them, n is the number of indicators; x _ij represents the element of the i-th row and the j-th column in the judgment matrix;

归一化得到w_i：Will

Normalize to get w _i :

Calculate the maximum eigenroot λ _max of the judgment matrix:

Among them, w=(w ₁ ,w ₂ ,...,w _n ) ^T is the weight vector;

The CR values of single-layer sorting and total hierarchical sorting were obtained by the following formulas:

Among them, CI is the consistency index; RI is the random consistency index;

If CR<0.1, the judgment matrix is consistent, and then the subjective weight vector W ₁ =(w ₁ ,w ₂ ,...,w _n ) is determined.

Specifically, in step S4, the method for determining the objective weight vector of each index by using the entropy weight method is:

Assuming that the number of objects to be evaluated is m and the number of indicators is n, the initial evaluation matrix A=(a _ij ) _m×n is obtained by assigning values to each indicator of each object, and after normalizing and normalizing it get the normalized matrix C=( _cij ) _m×n ;

Calculate the proportion f _ij of the j-th index to the i-th evaluation object according to the standard matrix C:

Calculate the information entropy value e _j contained in each evaluation index:

Calculate the entropy weight w _j of each evaluation index:

That is, the objective weight vector W ₂ =(w _j ) ^T ,j=1,2,...,n of each index determined by the entropy weight method is obtained.

Specifically, in step S5, the method for obtaining the combined weight vector of each indicator by using the game theory method is:

Perform linear combination weighting on the subjective weight vector W ₁ and the objective weight vector W ₂ ;

Based on the idea of game theory, the following combination coefficient equations are established:

Among them, a ₁ and a ₂ are the combination coefficients, that is, the proportions of the subjective weight and the objective weight in the combined weight, respectively;

The above equation system is based on the traditional game theory idea, and the obtained combination coefficient may be negative. Therefore, in order to ensure that the linear combination coefficient is positive, the combination coefficient is optimized and improved, and the following improved game theory model is obtained:

By constructing a Lagrangian function and finding partial derivatives:

where λ is the Lagrange multiplier;

By normalizing the obtained results, the combined weight coefficients obtained by the improved game theory model are as follows:

代入下式可求得组合权重向量为：Will

Substitute into the following formula to obtain the combined weight vector:

和

分别表示改进后归一化的主观权重和客观权重在组合权重中所占的比重。in,

and

respectively represent the proportion of the improved normalized subjective weight and objective weight in the combined weight.

Specifically, step S6 includes the following steps:

S601, assuming that the number of objects to be evaluated is m and the number of indicators is n, assign values to each indicator of each accident-prone area in the research sea area, and obtain an initial evaluation matrix A=(a _ij ) _m×n :

S602, determine the attribute of the index according to the influence relationship of each index on the size of the risk, and for the reverse index, use the following formula to carry out the index forwarding process, and obtain the forwardization matrix B=(b _ij ) _m×n :

S603, standardize each index according to the following formula:

为各列指标数据的均值；

is the mean of each column of indicator data;

get the normalized matrix C=( _cij ) _m×n ;

Combined with the combined weight vector W ^* , the weighted normalization matrix D=(d _ij ) _{m×n is} obtained;

Wherein, d _ij =c _ij ·w _s (1≤i≤m, 1≤j≤n, 1≤s≤n);

S604, according to the normalization matrix and the weighted normalization matrix, respectively determine the maximum risk set and the minimum risk set based on the Mahalanobis distance calculation and the gray correlation degree calculation;

Among them, C ⁺ , C ^- are the maximum risk set and the minimum risk set determined based on the Mahalanobis distance calculation and the gray correlation degree calculation according to the standardized matrix, respectively; D ⁺ , D ^- are determined according to the weighted normalization matrix and calculated based on the Mahalanobis distance. The maximum risk set and the minimum risk set calculated from the grey correlation degree;

S605, use the Mahalanobis distance calculation formula to obtain the Mahalanobis distance between each accident-prone area and the maximum risk set and the minimum risk set;

且

Σ为样本协方差矩阵；in,

and

Σ is the sample covariance matrix;

S606, according to the following formula, calculate the gray correlation coefficient between each accident-prone area and the maximum risk set and the minimum risk set with respect to each index;

Among them, ρ is the resolution;

Use the following formula to calculate the grey correlation degree between each accident-prone area and the maximum risk set and the minimum risk set;

与

S607, linearly combine the standardized Mahalanobis distance and the gray correlation degree to form a combined distance

and

Among them, α and β are the combination coefficients, and α+β=1;

S608, according to the following formula, the degree of closeness CC _i of each accident-prone area and the maximum risk set is obtained, and CC _i is taken as the relative navigable risk value of each accident-prone area; the larger the relative navigable risk value, the more accident-prone area The higher the relative navigation risk;

Among them, 0≤CC _i ≤1;

若事故易发区域的相对通航风险值大于风险阈值，则该区域为高风险区域，反之则为一般风险区域。S609, according to the relative navigable risk evaluation result of each accident-prone area, take the average value of the relative navigable risk value of all accident-prone areas as the risk threshold

If the relative navigable risk value of the accident-prone area is greater than the risk threshold, the area is a high-risk area, otherwise, it is a general-risk area.

Compared with the prior art, the beneficial effects of the present invention are: the present invention utilizes the K-means clustering method based on minimum circle coverage to identify and divide accident-prone areas in the research sea area, and then comprehensively considers traffic factors and ship factors. , channel conditions to establish a relative navigable risk evaluation index system, and combine the index weights obtained by the AHP and entropy method through game theory, and then obtain the combined weight of each index, and finally use the Mahalanobis distance to replace the Euclidean distance, gray The improved TOPSIS method for calculating the closeness of the correlation degree evaluates the relative navigation risk of each accident-prone area, and uses the average risk as the risk threshold to determine the location of each high-risk area, which improves the accuracy of the navigation risk assessment. The research on the allocation of marine emergency rescue resources provides the basis and basis, and can provide decision-making reference for marine safety management work. Relevant managers can make corresponding human and material resources according to the obtained risks.

Description of drawings

FIG. 1 is a schematic flowchart of a navigation risk assessment method based on the improved TOPSIS method of the present invention.

FIG. 2 is a spatial distribution diagram of marine accidents in the Haikou jurisdiction of the Qiongzhou Strait according to an embodiment of the present invention.

FIG. 3 is a schematic diagram of a preliminary division of the Qiongzhou Strait marine accident-prone areas in an embodiment of the present invention.

FIG. 4 is a schematic diagram of the final division of the Qiongzhou Strait marine accident-prone areas in an embodiment of the present invention.

Detailed ways

The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

As shown in Figure 1, the present embodiment provides a navigation risk assessment method based on the improved TOPSIS method. The present embodiment takes the sea area within the jurisdiction of the seaport on the south coast of the Qiongzhou Strait as a case to assess the navigation risk of the ship; the specific steps are as follows :

S1, use the K-means clustering method based on the minimum circle coverage to identify and divide the accident-prone areas in the research sea area, which specifically includes the following steps:

S101, according to the statistics of emergency search and rescue in the Qiongzhou Strait from 2012 to 2020 published by the Haikou Maritime Safety Administration, a total of 61 available accidents were extracted, and their latitude and longitude were imported into the ArcGIS software to obtain the spatial distribution map of maritime accidents in the Haikou area of the Qiongzhou Strait. As shown in Figure 2, the black circle in the figure is the accident point.

S102, on the basis of the spatial distribution map of marine accidents, draw a number of minimum circles that can cover the accident points, and realize the preliminary division of the marine accident-prone areas in the Qiongzhou Strait, as shown in Figure 3, from which one can be obtained. After the isolated accident points are eliminated, 11 accident-prone areas can be obtained, which are P1-P11 respectively. The location of the center of each circle is the initial cluster center of the accident-prone area.

S103, using Matlab software to input other parameters such as the number of clusters and the coordinates of the initial cluster center, perform K-means clustering on each accident coordinate point in the sea area, and obtain the final Qiongzhou Strait marine accident prone area as shown in Figure 4. The relevant information of accident-prone areas is shown in Table 1.

Table 1 Information table of accident-prone areas in Haikou, Qiongzhou Strait

From the schematic diagram of accident-prone areas, it can be seen that the accident-prone areas identified by K-means clustering covered by the smallest circle basically include Macun Port, Xiuying Port and other important port areas and passenger areas that need to be paid attention to in the jurisdiction. The waters where the rolling route is located, indicating that the method is applicable to the identification and division of accident-prone areas in the Haikou jurisdiction of the Qiongzhou Strait.

S2, comprehensively consider traffic factors, ship factors and channel condition factors to establish a relative navigation risk evaluation index system;

The relative navigation risk indicators of sea areas and their meanings are shown in Table 2 below:

Table 2 Relative navigation risk indicators in sea areas and their meanings

According to the above table, the relative navigation risk evaluation index system includes 3 first-level index factor layers and 8 second-level index factor layers, which are:

Traffic factors, the traffic factors include the following indicators: ship flow (vessel/hour), average ship density (vessel) and ship density dispersion; among them, the ship flow is the conversion of standard ship flow, that is, the ship is converted according to the ship length conversion factor table After that, make statistics; the unit of statistical ship density is the grid enclosed by each longitude and each latitude on the chart;

Vessel factors, which include the following indicators: Vessel Type (%) and Vessel Speed (knots);

Channel condition factors, the channel condition factors include the following indicators: the impact of special areas, the number of special points in the channel (number) and the number of obstructions (number); the impact of the special area includes docks, fishing areas, anchorages, shoals and The overall impact of the reef; the number of special points in the channel includes the total number of channel ends, track intersections and bends.

Observe and obtain relevant indicator data through channels such as Ship News Network, Treasure Ship Network, and Haikou Maritime Safety Administration, and based on the collected data, the indicator data of various accident-prone areas are sorted and obtained, as shown in Table 3:

Table 3 Index parameters of accident-prone areas in Qiongzhou Strait

S3, using AHP to determine the subjective weight vector W ₁ of each index;

According to the survey results of the intentions of experts, marine practitioners, marine safety supervisors and other personnel on the importance of selecting indicators, take the geometric mean to obtain the indicator importance judgment matrix X=(x _ij ) _n×n ; calculate the elements of each row of the judgment matrix X the geometric mean of

Among them, n is the number of indicators; x _ij represents the element of the i-th row and the j-th column in the judgment matrix;

归一化得到w_i：Will

Normalize to get w _i :

Calculate the maximum eigenroot λ _max of the judgment matrix:

Among them, w=(w ₁ ,w ₂ ,...,w _n ) ^T is the weight vector;

The CR values of single-layer sorting and total hierarchical sorting were obtained by the following formulas:

Among them, CI is the consistency index; RI is the random consistency index;

If CR<0.1, the judgment matrix is consistent, and then the subjective weight vector W ₁ =(w ₁ ,w ₂ ,...,w _n ) is determined.

S4, using the entropy weight method to determine the objective weight vector W ₂ of each index;

Assuming that the number of objects to be evaluated is m and the number of indicators is n, the initial evaluation matrix A=(a _ij ) _m×n is obtained by assigning values to each indicator of each object, and after normalizing and normalizing it get the normalized matrix C=( _cij ) _m×n ;

Calculate the proportion f _ij of the j-th index to the i-th evaluation object according to the standard matrix C:

Calculate the information entropy value e _j contained in each evaluation index:

Calculate the entropy weight w _j of each evaluation index:

That is, the objective weight vector W ₂ =(w _j ) ^T ,j=1,2,...,n of each index determined by the entropy weight method is obtained.

S5, adopt game theory method to combine described subjective weight vector and objective weight vector, obtain the combined weight vector W ^* of each index;

Perform linear combination weighting on the subjective weight vector W ₁ and the objective weight vector W ₂ :

Based on the idea of game theory, the following combination coefficient equations are established:

Among them, a ₁ and a ₂ are the combination coefficients, that is, the proportions of the subjective weight and the objective weight in the combined weight, respectively;

The above equation system is based on the traditional game theory idea, and the obtained combination coefficient may be negative. Therefore, in order to ensure that the linear combination coefficient is positive, the combination coefficient is optimized and improved, and the following improved game theory model is obtained:

By constructing a Lagrangian function and finding partial derivatives:

where λ is the Lagrange multiplier;

By normalizing the obtained results, the combined weight coefficients obtained by the improved game theory model are as follows:

代入下式可求得组合权重向量为：Will

Substitute into the following formula to obtain the combined weight vector:

为改进后归一化的主观权重在组合权重中所占的比重，

的值为0.5423；

为改进后归一化的客观权重在组合权重中所占的比重，

的值为0.4577；in,

In order to improve the proportion of the normalized subjective weight in the combined weight,

The value of 0.5423;

In order to improve the proportion of the normalized objective weight in the combined weight,

is 0.4577;

The weights of the evaluation indicators for each accident-prone area are shown in Table 4 below:

Table 4 Weights of evaluation indicators

x₁ x₂ x₃ x₄ x₅ x₆ x₇ x₈ subjective weight 0.030 0.078 0.034 0.200 0.086 0.356 0.137 0.078 objective weight 0.065 0.089 0.234 0.154 0.043 0.180 0.107 0.129 combined weights 0.046 0.083 0.126 0.179 0.066 0.275 0.123 0.102

From Table 4, it can be seen that among the 8 indicators, the two indicators of special area impact and ship type account for a large proportion. Yes, the accidents in the waters near anchorages and wharfs account for about 20% of the total number of statistical accidents. In addition, considering the impact of wharfs, fishing areas, anchorages, reefs, and shoals as an indicator, this has also deepened to a certain extent. Importance of this metric.

S6, use the Mahalanobis distance to replace the Euclidean distance and the gray correlation degree to calculate the closeness of the improved TOPSIS method to evaluate the relative navigable risk of accident-prone areas, and use the average risk as the risk threshold to determine the location of each high-risk area;

Step S6 includes the following steps:

S601, assuming that the number of objects to be evaluated is m and the number of indicators is n, assign values to each indicator of each accident-prone area in the research sea area, and obtain an initial evaluation matrix A=(a _ij ) _m×n ;

S602, determine the attribute of the index according to the influence relationship of each index on the size of the risk, and for the reverse index, use the following formula to carry out the index forwarding process, and obtain the forwardization matrix B=(b _ij ) _m×n :

S603, standardize each index according to the following formula;

为各列指标数据的均值；

is the mean of each column of indicator data;

Combined with Table 3, the standardized matrix C=(c _ij ) _{m×n is} obtained;

Combined with the combined weight vector W ^* , the weighted normalization matrix D=(d _ij ) _{m×n is} obtained;

Wherein, d _ij =c _ij ·w _s (1≤i≤m, 1≤j≤n, 1≤s≤n);

S604, according to the normalization matrix and the weighted normalization matrix, respectively determine the maximum risk set and the minimum risk set used for the Mahalanobis distance calculation and the gray correlation degree calculation;

Among them, C ⁺ , C ^- are the maximum risk set and the minimum risk set determined based on the Mahalanobis distance calculation and the gray correlation degree calculation according to the standardized matrix, respectively; D ⁺ , D ^- are determined according to the weighted normalization matrix and calculated based on the Mahalanobis distance. The maximum risk set and the minimum risk set calculated from the grey correlation degree;

The maximum risk set and minimum risk set of the standardized matrix and the weighted norm matrix are shown in Table 5 below:

Table 5 Maximum risk set and minimum risk set of standardized matrix and weighted norm matrix

x₁ x₂ x₃ x₄ x₅ x₆ x₇ x₈ C⁺ 1.688 2.939 4.264 4.192 1.287 4.681 2.444 3.385 C^- 0.291 0.419 0.000 0.000 0.680 0.000 0.000 0.000 D⁺ 0.078 0.245 0.537 0.749 0.085 1.289 0.301 0.344 D^- 0.013 0.035 0.000 0.000 0.045 0.000 0.000 0.000

S605, use the Mahalanobis distance calculation formula to obtain the Mahalanobis distance between each accident-prone area and the maximum risk set and the minimum risk set;

且

Σ为样本协方差矩阵；in,

and

Σ is the sample covariance matrix;

S606, according to the following formula, calculate the gray correlation coefficient between each accident-prone area and the maximum risk set and the minimum risk set with respect to each index;

Among them, ρ is the resolution, and in this embodiment, the resolution ρ=0.5;

Use the following formula to calculate the grey correlation degree between each accident-prone area and the maximum risk set and the minimum risk set;

与

S607, linearly combine the standardized Mahalanobis distance and the gray correlation degree to form a combined distance

and

Among them, α and β are the combination coefficients respectively, and α+β=1, in this embodiment, α=β=0.5.

S608, according to the following formula, the degree of closeness CC _i of each accident-prone area and the maximum risk set is obtained, and CC _i is taken as the relative navigable risk value of each accident-prone area; the larger the relative navigable risk value, the more accident-prone area The higher the relative navigation risk;

Among them, 0≤CC _i ≤1;

若事故易发区域的相对通航风险值大于风险阈值，则该区域为高风险区域，反之则为一般风险区域；S609, according to the relative navigable risk evaluation result of each accident-prone area, take the average value of the relative navigable risk value of all accident-prone areas as the risk threshold

If the relative navigable risk value of the accident-prone area is greater than the risk threshold, the area is a high-risk area, otherwise, it is a general-risk area;

The standardized Mahalanobis distance, grey correlation degree and relative navigable risk assessment calculation results are shown in Table 6 below:

Table 6 Calculation results of relative navigation risk assessment

P₁ P₂ P₃ P₄ P₅ P₆ P₇ P₈ P₉ P₁₀ P₁₁ d(C⁺) 0.951 0.674 1.033 1.370 1.069 0.616 1.072 1.115 1.074 0.986 1.040 d(C^-) 1.003 1.711 0.841 0.761 0.801 2.027 0.749 0.669 0.714 0.915 0.809 r⁺ 0.914 1.002 1.015 1.164 1.007 1.218 0.949 0.949 0.946 0.919 0.916 ^- 1.078 0.936 0.968 0.853 0.979 0.827 1.045 1.061 1.059 1.085 1.108 H⁺ 1.917 2.712 1.856 1.925 1.808 3.245 1.698 1.617 1.660 1.835 1.725 H^- 2.029 1.610 2.001 2.223 2.049 1.443 2.117 2.177 2.133 2.071 2.148 CC 0.486 0.628 0.481 0.464 0.469 0.692 0.445 0.426 0.438 0.470 0.445

According to the above table, a total of 11 accident-prone areas were identified, and the relative navigation risk of each area was 0.486, 0.628, 0.481, 0.464, 0.469, 0.692, 0.445, 0.426, 0.438, 0.470, and 0.445, respectively. According to the relative navigable risk value, the accident-prone areas are ranked as follows: P ₆ , P ₂ , P ₁ , P ₃ , P ₁₀ , P ₅ , P ₄ , P ₁₁ , P ₇ , P ₉ , P ₈ . According to the calculation, the risk threshold value of the research water area is 0.495. By comparing the relative navigation risk value of each area with the risk threshold value, it can be concluded that there are two high-risk areas, namely P ₆ and P ₂ . Among them, P ₂ and P ₆ are both located in the port area and there are a large number of nearby anchorages. There are close ro-ro passenger ships, high ship density, relatively concentrated distribution of ships, and the influence of anchorages, resulting in high index values. Risk Value ranks top. P ₄ is also located in port waters, but because the waters are relatively open and less affected by special areas such as anchorages and shoals, the risk level in this area is lower than that of P ₆ and P ₂ waters. On the other hand, P ₁ is close to the reporting line at the west entrance of the Qiongzhou Strait, with a high flow of ships, and there are multiple track intersections, and there are shallow areas and the influence of fishing boats in the surrounding area. Therefore, the risk of P ₁ is also at a high level. P ₃ and P ₅ are located near the warning area, and are at the intersection of the east-west ro-ro passenger ships and the north-south ship tracks. The daily flow of ships is large, and the proportion of passenger ships is relatively high, so the risk is at an upper-middle level. Compared with other areas, P ₇ and P ₈ are less affected by special points and special areas. Although the index values of ship flow and average ship speed are high, their relative risks are low. P ₉ , P ₁₀ , and P ₁₁ are all located near the middle waterway of Qiongzhou Strait. There is a shoal and track intersection area near P ₁₀ , while P ₁₁ is located at the east mouth of Qiongzhou Strait, which is greatly affected by wind and waves, and the buoy is easily displaced during typhoon seasons , and ships need to enter and exit the Qiongzhou Strait waterway through P _11. Therefore, P ₁₀ and P ₁₁ are more risky than P ₉ .

Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, and substitutions can be made in these embodiments without departing from the principle and spirit of the invention and modifications, the scope of the present invention is defined by the appended claims and their equivalents.

Claims (7)

1. a navigation risk assessment method based on improving TOPSIS method, is characterized in that, comprises the following steps: S1, using the K-means clustering method based on the minimum circle coverage to identify and divide the accident-prone areas in the research sea area; S2, comprehensively consider traffic factors, ship factors and channel condition factors to establish a relative navigation risk evaluation index system; S3, using AHP to determine the subjective weight vector W ₁ of each index; S4, using the entropy weight method to determine the objective weight vector W ₂ of each index; S5, adopt game theory method to combine described subjective weight vector and objective weight vector, obtain the combined weight vector W ^* of each index; S6, use Mahalanobis distance instead of Euclidean distance and grey correlation degree to calculate the improved TOPSIS method to evaluate the relative navigable risk of accident-prone areas, and use the average risk as the risk threshold to determine the location of each high-risk area. 2. a kind of navigation risk assessment method based on improving TOPSIS method according to claim 1 is characterized in that, step S1 comprises the following steps: S101, obtain all marine traffic accident information in the study area for a period of time, extract the latitude and longitude coordinates of each accident point, and use ArcGis software to draw a spatial distribution map of marine traffic accidents in the study area; S102, based on the idea of minimum circle coverage, draw a number of circles on the spatial distribution map of maritime traffic accidents in the study area to cover the accident points to distinguish each clustering area, one circle represents one clustering area, and each clustering area At least two accident points are included, and isolated accident noise points are identified and eliminated according to the clustering results; S103 , using the number of all circles as the number of clusters, using the centers of each circle as the cluster center of each clustering area, and using the K-means clustering method to obtain a final result of spatial division of areas prone to marine accidents. 3. a kind of navigation risk assessment method based on improving TOPSIS method according to claim 1, is characterized in that, in step S2, described relative navigation risk assessment index system comprises: Traffic factors, which include the following indicators: ship flow, average ship density, and ship density dispersion; Vessel factors including the following indicators: vessel type and vessel speed; Channel condition factors, the channel condition factors include the following indicators: the impact of special areas, the number of special points in the channel and the number of obstructions; the impact of the special area includes the overall impact of docks, fishing areas, anchorages, shoals and reefs; The above-mentioned number of special points of the channel includes the total number of the end of the channel, the intersection of the track and the bend. 4. a kind of navigation risk assessment method based on improving TOPSIS method according to claim 1, is characterized in that, in step S3, utilizes AHP to determine the method for the subjective weight vector of each index is: According to the results of the intentional investigation on the importance of the selected indicators, the indicator importance judgment matrix X=(x _ij ) _{n×n is} obtained; the geometric mean of the elements in each row of the judgment matrix X is calculated.

Among them, n is the number of indicators; x _ij represents the element of the i-th row and the j-th column in the judgment matrix; Will

Normalize to get w _i :

Calculate the maximum eigenroot λ _max of the judgment matrix:

Among them, w=(w ₁ ,w ₂ ,...,w _n ) ^T is the weight vector; Use the following formulas to obtain the CR values of single-level sorting and total hierarchical sorting:

Among them, CI is the consistency index; RI is the random consistency index; CR is the consistency ratio; If CR<0.1, the judgment matrix is consistent, and then the subjective weight vector W ₁ =(w ₁ ,w ₂ ,...,w _n ) is determined. 5. a kind of navigation risk assessment method based on improving TOPSIS method according to claim 1, is characterized in that, in step S4, utilize entropy weight method to determine the method for the objective weight vector of each index is: Assuming that the number of objects to be evaluated is m and the number of indicators is n, the initial evaluation matrix A=(a _ij ) _m×n is obtained by assigning values to each indicator of each object, and after normalizing and normalizing it get the normalized matrix C=( _cij ) _m×n ; Among them, a _ij and c _ij represent the elements of the i-th row and the j-th column in the initial evaluation matrix and the standardized evaluation matrix, respectively; Calculate the proportion f _ij of the j-th index to the i-th evaluation object according to the standard matrix:

Calculate the information entropy value e _j contained in each evaluation index:

Calculate the entropy weight w _j of each evaluation index:

That is, the objective weight vector W ₂ =(w _j ) ^T ,j=1,2,...,n of each index determined by the entropy weight method is obtained. 6. a kind of navigation risk assessment method based on improving TOPSIS method according to claim 1 is characterized in that, in step S5, the method that adopts game theory method to obtain the combined weight vector of each index is: Perform linear combination weighting on the subjective weight vector W ₁ and the objective weight vector W ₂ ; Based on the idea of game theory, the following combination coefficient equations are established:

Among them, a ₁ and a ₂ are the combination coefficients, that is, the proportions of the subjective weight and the objective weight in the combined weight, respectively; In order to ensure that the linear combination coefficient is positive, the combination coefficient is optimized and improved, and the following improved game theory model is obtained:

By constructing a Lagrangian function and finding partial derivatives:

where λ is the Lagrange multiplier; By normalizing the obtained results, the combined weight coefficients obtained by the improved game theory model are as follows:

Will

Substitute into the following formula to obtain the combined weight vector W ^* ={w ₁ ,w ₂ ,...w _s }, s=1,2,...,n is:

in,

and

respectively represent the proportion of the improved normalized subjective weight and objective weight in the combined weight. 7. a kind of navigation risk assessment method based on improving TOPSIS method according to claim 1 is characterized in that, step S6 comprises the following steps: S601, assuming that the number of objects to be evaluated is m and the number of indicators is n, assign values to each indicator of each accident-prone area in the research sea area, and obtain an initial evaluation matrix A=(a _ij ) _m×n : S602: Determine the attribute of the index according to the influence relationship of each index on the size of the risk. For the reverse index, use the following formula to carry out the index forwarding process to obtain the forwardization matrix B=(b _ij ) _m×n :

S603, standardize each index according to the following formula:

is the mean of each column of indicator data; get the normalized matrix C=( _cij ) _m×n ; Combined with the combined weight vector W ^* , the weighted normalization matrix D=(d _ij ) _{m×n is} obtained; Wherein, d _ij =c _ij ·w _s (1≤i≤m, 1≤j≤n, 1≤s≤n); Among them, ws represents the weight of the _s -th indicator; S604, according to the normalization matrix and the weighted normalization matrix, respectively determine the maximum risk set and the minimum risk set based on the Mahalanobis distance calculation and the gray correlation degree calculation;

Among them, C ⁺ and C ^- are respectively the maximum risk set and the minimum risk set calculated based on Mahalanobis distance calculation and grey correlation degree according to the standardized matrix; D ⁺ and D ^- are determined according to the weighted normalization matrix and calculated based on Mahalanobis distance respectively. The maximum risk set and the minimum risk set calculated from the grey correlation degree; S605, use the Mahalanobis distance calculation formula to obtain the Mahalanobis distance between each accident-prone area and the maximum risk set and the minimum risk set;

in,

and

Σ is the sample covariance matrix; S606, according to the following formula, calculate the gray correlation coefficient between each accident-prone area and the maximum risk set and the minimum risk set with respect to each index;

Among them, ρ is the resolution; Use the following formula to calculate the grey correlation degree between each accident-prone area and the maximum risk set and the minimum risk set;

S607, linearly combining the standardized Mahalanobis distance and the gray correlation degree to form combined distances H _i ⁺ and H _i ⁻ ;

Among them, α and β are the combination coefficients, and α+β=1; S608, according to the following formula, the degree of closeness CC _i between each accident-prone area and the maximum risk set is obtained, and CC _i is taken as the relative navigable risk value of each accident-prone area; the larger the relative navigable risk value, the more accident-prone area The higher the relative navigation risk;

Among them, 0≤CC _i ≤1; S609, according to the relative navigable risk evaluation result of each accident-prone area, take the average value of the relative navigable risk value of all accident-prone areas as the risk threshold

If the relative navigable risk value of the accident-prone area is greater than the risk threshold, the area is a high-risk area, otherwise, it is a general-risk area.

Images