An emergency resource scheduling strategy generation method for sudden leakage accidents

Abstract

The application relates to the technical field of data processing, and discloses an emergency resource scheduling strategy generation method for coping with sudden leakage accidents, which comprises the following steps: acquiring accident situation data and constructing a scheduling data structure; extracting a leakage situation tightness feature and initializing a gray wolf population in a biased manner; performing gray wolf search with tightness fluctuation sensing and iteratively performing a hard constraint priority scheduling strategy; and performing a rolling scheduling application process for sudden chemical leakage accidents. The application has the beneficial effect that: a resource demand tightness factor is constructed by fusing a leakage point concentration value, a situation change intensity, a resource demand deviation degree and a reachable inventory gap degree, which can provide smooth and comprehensive leakage point tightness evaluation for emergency resource scheduling; a gray wolf population biased initialization strategy based on the tightness factor is proposed, so that the initial candidate scheme is more inclined to guarantee high-risk leakage points under the premise of meeting the inventory constraint, while necessary population diversity is maintained.

Description

Technical Field

[0001] This invention relates to the field of chemical safety data processing technology, specifically a method for generating emergency resource scheduling strategies to deal with sudden leakage accidents. Background Technology

[0002] In the event of a sudden chemical leak, it is necessary to simultaneously acquire multi-dimensional situational data such as the distribution of leak points, the inventory at rescue points, resource demand, transportation access status, and the concentration field of toxic substances. In addition, it is necessary to consider the differences in the degree of danger and resource demand structure of different leak points, the available inventory gap, and the matching and coupling relationship between leak plugging and protective materials in order to generate an emergency resource dispatch plan that meets the principles of inventory constraints and life rescue priority.

[0003] When conducting emergency resource allocation, existing technologies often determine the order of resource allocation based solely on the distance to the leak point or a single concentration value, failing to comprehensively assess multi-dimensional situational factors such as changes in the concentration field gradient, deviations in demand structure, and inventory gaps. This can easily lead to insufficient protection for high-risk leak points or misallocation of special resources. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a method for generating emergency resource scheduling strategies to cope with sudden leakage accidents. By integrating leakage point concentration values, intensity of situational changes, deviation of resource demand, and available inventory gap, a resource demand tightness factor is constructed, which can provide a smooth and comprehensive leakage point tightness assessment for emergency resource scheduling.

[0005] To achieve the above objectives, the present invention employs the following technical solution: A method for generating an emergency resource scheduling strategy to respond to sudden leakage accidents includes the following steps: S1. Obtain situational data including at least the leak point, rescue point, resource type, resource requirement, inventory, expected transportation time moment, and continuous concentration field, and construct a data structure to characterize the scheduling scheme; S2. Calculate the resource demand density factor for each leakage point based on the situation data, and inject the density factor into the resource allocation ratio and hyperspherical manifold position of the initial gray wolf population to generate an initial candidate scheduling scheme that meets the basic inventory constraints, making the scheme biased towards high-risk, high-demand and high-gap leakage points. S3. During the iteration of the Grey Wolf algorithm, the individual position update strategy is adjusted according to the dynamic fluctuation of the density factor, and the hard inventory constraint is satisfied first to generate the optimal scheduling strategy for the current iteration. S4. Apply the scheduling strategy obtained through iterative optimization to the rolling time-domain window, and repeat S1 to S3 as the accident situation data is updated to output the real-time updated emergency resource scheduling strategy.

[0006] Furthermore, S2 specifically refers to: S21. Taking the continuous concentration field, resource demand matrix, inventory matrix and expected transportation time matrix as input, output the resource demand density factor corresponding to each leakage point; S22. Using the resource demand density factor, resource demand matrix, and inventory matrix as inputs, generate an initial discrete emergency resource scheduling scheme that satisfies the basic constraints of the rescue point inventory, and map it to the location of a hyperspherical manifold. S23. Convert the hyperspherical manifold position into a discrete emergency resource scheduling scheme in non-negative integer form, and correct the upper limit of inventory, upper limit of demand, minimum guarantee of key materials and resource coupling constraints during the conversion process.

[0007] Furthermore, S3 specifically refers to: S31. Statistically analyze the fitness distribution of discrete emergency resource scheduling schemes in the current gray wolf population, calculate the normalized fitness distribution entropy, and use it to determine whether the gray wolf population is in a state of global exploration, local development, or overcrowding. S32. Based on the historical fluctuations of the normalized fitness distribution entropy and resource demand density factor, the search control factor of the gray wolf optimization algorithm is adaptively adjusted so that the search process maintains a strong exploration capability when the accident situation is unstable and accelerates convergence when the situation is stable. S33. Update the position of the hyperspherical manifold based on the gray wolf pack hierarchical coordination mechanism, and continuously modulate it using the resource demand density factor after the position update, so that the high-priority leakage point continuously obtains resource allocation tendency during the iterative search process. S34. First, evaluate the degree of violation of hard constraints for discrete emergency resource scheduling schemes, and then evaluate the total scheduling cost, maximum response time and non-critical resource gap for scheduling schemes that meet the hard constraints.

[0008] S35. Select the leader wolf based on the hard constraint priority evaluation results, iteratively update the gray wolf population, and output the optimal emergency resource scheduling strategy after the termination condition is met.

[0009] Furthermore, S21 specifically refers to: Reading the first in a continuous concentration field The local concentration values ​​at the spatial coordinates of each leak point are used to obtain the concentration values ​​at the leak points. With the first Using the spatial coordinates of each leak point as the center, the gradient amplitude of the concentration field is calculated in a continuous concentration field to obtain the intensity of the leak situation change. According to the The resource requirement vectors of all leak points within the scheduling time window are used to calculate the i-th... Deviation in resource requirements at each leak point; Based on the supplies at the relief point and the estimated transport time, calculate the first... The achievable inventory gap at each leakage point; Normalize the concentration value at the leakage point, the intensity of the change in leakage status, the deviation of resource demand, and the available inventory gap, respectively, to obtain the concentration risk weight, diffusion change weight, demand deviation weight, and inventory gap weight. By integrating the concentration risk weight, diffusion change weight, demand deviation weight, and inventory gap weight, we obtain the first... Resource demand density factor for each leakage point.

[0010] Furthermore, S22 specifically refers to: Regarding the first The gray wolf generates an initial discrete emergency resource scheduling scheme; For each rescue point and each resource category, construct the probability of leak point allocation based on the resource demand density factor and the demand quantity; Based on the quantity of emergency resources of type I at the i-th rescue point and the probability of allocation of leak points, generate the initial quantity of resources corresponding to each leak point; Perform demand ceiling correction on the initial discrete emergency resource scheduling scheme to avoid allocating resources that significantly exceed the demand to a leak point during the initialization phase; Construct an extended density weight vector based on the resource demand density factor; The initial discrete emergency resource scheduling scheme is transformed into the first... The initial hyperspherical manifold position of the gray wolf; Repeat the above steps until the generated size is [size missing]. The initial gray wolf population.

[0011] Furthermore, S23 specifically refers to: Construct an inverse scaling vector based on the inventory quantity at the rescue point and the required quantity at the leak point; The first The position of the gray wolf's hyperspherical manifold is multiplied element-wise with the inverse scaling vector to obtain a continuous resource quantity vector. Integerization is performed on the continuous resource quantity vector to obtain the initial discrete resource quantity; The initial discrete resource quantity is projected onto the inventory limit to ensure that the total amount of each type of emergency resource mobilized from each rescue point does not exceed the corresponding inventory quantity. Apply a minimum guarantee projection for critical materials to the initial discrete resource quantity to ensure a minimum satisfaction rate of life-saving materials for severely injured leak points; Perform resource coupling constraint projection on the initial discrete resource quantity to ensure that the basic operation ratio relationship is met among the resources that need to be used in combination. The resource quantity after completing the projection of the upper limit of inventory, the minimum guarantee of key materials, and the resource coupling constraint is expanded into a discrete emergency resource scheduling scheme.

[0012] Furthermore, S31 specifically refers to: In the In the nth iteration, for the th Calculate the temporary fitness value for the discrete emergency resource scheduling scheme corresponding to the gray wolf; Based on the size of the gray wolf population Determine the number of fitness partitions ; Count the number of gray wolves in each fitness interval to obtain the number of intervals; fitness distribution entropy Divide by the maximum possible entropy to obtain the normalized fitness distribution entropy.

[0013] Furthermore, S32 specifically refers to: Record recent The normalized fitness distribution entropy of each iteration forms a local entropy sequence; Calculate entropy fluctuation values ​​based on local entropy sequences; Based on the The resource demand density factor between the current scheduling time window and the previous scheduling time window is used to calculate the resource demand density factor of the current scheduling time window. The fluctuation value of resource demand density of each scheduling time window relative to the previous scheduling time window; The search retention index is calculated based on the current normalized fitness distribution entropy, entropy fluctuation value, and resource demand tightness fluctuation value. The search control factor is calculated based on the search retention index.

[0014] Furthermore, S33 specifically refers to: Screening based on the current evaluation results of the gray wolf population Wolf, wolves and Wolf; The wolf represents the leader wolf, which has the best overall performance in the current population. The wolf represents the alpha wolf, which has the second-best overall performance in the current population. The wolf represents the leader wolf, which has the third best overall performance in the current population; For the The first gray wolf in In the next iteration, the hyperspherical manifold position is updated using the grey wolf position to obtain candidate continuous positions; Construct a density modulation vector based on the resource demand density factor; The candidate continuous positions are multiplied element-wise by the compactness modulation vector and then reprojected onto the unit hypersphere to obtain the updated hyperspherical manifold positions. The updated hyperspherical manifold configuration is input into the discrete scheduling back projection process to obtain the first... The first gray wolf in Discrete emergency resource scheduling scheme in the next iteration.

[0015] Furthermore, S34 specifically refers to: For the discrete emergency resource scheduling scheme to be evaluated Calculate the degree of hard constraint violation; For a discrete emergency resource scheduling scheme with a hard constraint violation degree of 0, calculate the total scheduling cost; For a discrete emergency resource scheduling scheme with a hard constraint violation degree of 0, calculate the maximum response time; For a discrete emergency resource scheduling scheme with a hard constraint violation rate of 0, calculate the non-critical resource gap rate.

[0016] Furthermore, 35 specifically refers to: After each iteration, a set of scheduling schemes that satisfy the hard constraints is selected based on the degree of hard constraint violation. Received by filtering Wolf, wolves and The wolf performs search control factor adjustment, hyperspherical manifold position update, density modulation, and discrete scheduling back projection to obtain the discrete emergency resource scheduling scheme for the next iteration, and continues to perform hard constraint priority evaluation; Select the optimal emergency resource scheduling strategy from the final set of non-dominated feasible candidates; The optimal emergency resource allocation strategy is reduced to a three-dimensional scheduling table. Compared with the prior art, the beneficial effects of the present invention are as follows: This invention proposes an emergency resource scheduling strategy generation method for responding to sudden leakage accidents, which is innovative compared to existing technologies in the following aspects: 1. This invention constructs a resource demand tightness factor by integrating the leakage point concentration value, the intensity of situational change, the deviation of resource demand, and the available inventory gap, which can provide a smooth and comprehensive leakage point tightness assessment for emergency resource scheduling.

[0017] 2. This invention proposes a biased initialization strategy for gray wolf population based on density factor, which makes the initial candidate scheme more inclined to protect high-risk leakage points while maintaining necessary population diversity under the premise of meeting inventory constraints.

[0018] 3. In the iterative search process, this invention introduces a nonlinear convergence factor driven by both fitness distribution entropy and density fluctuation. It adaptively adjusts global exploration and local development according to the population dispersion and accident situation changes, maintaining strong exploration capability when the accident situation is unstable and accelerating convergence when the situation is stable.

[0019] 4. This invention establishes a two-layer scheduling scheme evaluation mechanism that prioritizes hard constraints, ensuring that safety constraints such as life rescue, inventory limits, and resource coupling take precedence over cost and timeliness objectives. Furthermore, it applies continuous density modulation during location updates to maintain continuous protection for high-priority leakage points. Attached Figure Description

[0020] Figure 1 This is a flowchart of the present invention; Figure 2 This is a heatmap of the expected transportation time matrix; Figure 3 It is a map showing the continuous concentration field in the accident area and the distribution of leak points and rescue points; Figure 4 It is a heatmap of the initial allocation probability weights of various resources to each leak point; Figure 5 This is a curve showing the convergence of the total scheduling cost. Figure 6 This is the maximum response time convergence curve; Figure 7 This is a convergence curve of the non-critical resource shortage rate; Figure 8 This is the resource allocation table for rescue point 1 in the optimal scheduling scheme. Detailed Implementation

[0021] The present invention will be further illustrated below with reference to specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. Furthermore, it should be understood that after reading the teachings of this invention, those skilled in the art can make various alterations or modifications to the invention, and these equivalent forms also fall within the scope defined in this application.

[0022] like Figure 1 As shown, this invention proposes a method for generating emergency resource scheduling strategies to cope with sudden leakage accidents, the specific contents of which are as follows: S1. Accident Situation Data Acquisition and Scheduling Data Structure Construction Emergency resource allocation for sudden chemical leaks needs to consider the spread of toxic substances, resource demand at the leak point, inventory capacity at rescue points, transportation access status, and the coupling relationship of critical materials. If resources are allocated solely based on the distance to the leak point or a single concentration value, problems such as insufficient protection for high-risk leak points, misallocation of special resources, over-limit inventory at rescue points, and excessively long response times may occur.

[0023] This step involves uniformly constructing data objects such as leak point, rescue point, resource type, inventory, demand, transportation time, and concentration field within each scheduling time window. This provides input for subsequent resource demand density calculations and optimization of discrete emergency resource scheduling schemes. The specific steps are as follows: 1> Industrial data acquisition, in the first Within a scheduling time window, obtain the set of leak points, the set of rescue points, and the set of resource categories. Among them, This represents the scheduling time window index, used to characterize the stages of accident situation changes over time; This represents the total number of leak points, which can be 6, 8, or 10. Indicates the leak point index. ; This indicates the total number of rescue points; 3, 4, or 5 are acceptable. Indicates the index of the rescue point. ; This indicates the total number of resource categories, and can be either 5 or 6. Represents the resource category index. .

[0024] Furthermore, let the first... The spatial coordinates of the leak points are: , It refers to the first The coordinates of the location of the leak point in the accident area coordinate system; let the first leak point be... The spatial coordinates of the rescue points are: , It refers to the first The location coordinates of each rescue point in the accident area coordinate system. The accident area coordinate system can be a two-dimensional plane coordinate system or a three-dimensional spatial coordinate system that includes height. When on-site dispatching is mainly carried out in the ground road network, the two-dimensional plane coordinate system is preferred.

[0025] 2> Get the first Resource demand matrix within a scheduling time window , This refers to a matrix composed of the quantities of various emergency resources required by each leak point, with dimensions of [missing information]. .in, Indicates the first The leak point is at the [number]th [location]. Within the scheduling time window, for the first The required quantity of emergency resources.

[0026] In specific implementation, the first Emergency resources include chemical protective suits, leak-stopping tools, neutralizing agents, first-aid equipment, isolation and warning equipment, and fire-fighting foam. The required quantity of these resources can be automatically estimated based on the type of leaked substance, the strength of the leak source, the number of injured persons, the radius of impact, and the emergency response plan; for example, when the first... When the first leak point involves an acidic corrosive medium leak, the required amount of neutralizing agent can be calculated based on the estimated mass of the leak and the amount of neutralizing agent used per unit mass; when the second leak point involves an acidic corrosive medium leak, the required amount of neutralizing agent can be calculated based on the estimated mass of the leak and the amount of neutralizing agent used per unit mass. When there is a risk of personnel exposure at a leak point, the required number of emergency medical equipment can be calculated based on the number of injured persons and the minimum protection factor.

[0027] 3> Get the first Inventory matrix within a scheduling time window , This refers to a matrix composed of the available emergency resource inventory at each rescue point, with dimension [missing information]. .in, Indicates the first The rescue point was at the The first one that can be called within the scheduling time window Quantity of emergency resources in stock.

[0028] Furthermore, if some resources have already been dispatched within the preceding scheduling time window, then during the calculation... The number of resources that have been scheduled but not yet returned is deducted in time, thereby avoiding the repeated use of the same batch of resources during the rolling scheduling process and ensuring that the discrete emergency resource scheduling scheme has practical significance.

[0029] 4> Obtain the first Matrix of estimated transport times within each scheduling window , This refers to a matrix consisting of the estimated arrival times of resources from each rescue point to each leak point, with dimension 1. .in, Indicates the first The rescue point to the first The estimated arrival time, in minutes, required to deliver emergency resources to each leak point.

[0030] In practical implementation, It can be calculated based on road distance, road traffic speed, loading preparation time, and traffic control detour time. The calculation method is expressed as follows: ; in, Indicates the first Within the scheduling time window, the first The first rescue point to the first The distance to the accessible road from each leak point; Indicates the first Within the scheduling time window, the first The first rescue point to the first Average traffic speed at each leak point; This represents the velocity stability constant, used to prevent the denominator from being zero, and can be taken as... ; Indicates the first Resource loading and preparation time at each rescue point; Indicates the first The first rescue point to the first Traffic control and detour times along the route to each leak point.

[0031] In one embodiment, such as Figure 2 As shown, a heatmap of the expected transportation time matrix is ​​analyzed; the horizontal axis represents the leak point (dimensionless), and the vertical axis represents the rescue point (dimensionless). The color intensity and labeled values ​​in the figure all represent transportation time (in minutes), which is obtained by adding the road distance, traffic speed, loading preparation time, and traffic control time. The experiment shows that the transportation time from different rescue points to the same leak point varies significantly, which directly affects the timeliness of resource scheduling and the calculation of available inventory gaps.

[0032] 5> Construct the first Continuous concentration field within a scheduling time window , This refers to the accident area in the [number]th [location]. Spatial distribution of toxic substance concentration within a scheduling time window, in units of , This represents the coordinates of any spatial location within the accident area.

[0033] In one implementation, if the leakage source strength, average wind speed, wind direction, and effective source height are known at the site, a Gaussian plume diffusion model can be used to calculate the continuous concentration field. In the wind direction coordinate system, the downwind coordinates are marked as... The crosswind coordinates are set as The vertical coordinates are marked as All units are Then the continuous concentration field It can be represented as: ; in, Indicates the first The leakage source strength within a scheduling time window is used to reflect the mass of toxic substances released into the environment per unit time, with units of [missing information]. ; Indicates the first The average wind speed within each scheduling time window, in units of ; This represents the lateral diffusion parameter, used to characterize the width of toxic substances spreading across the wind, and is measured in units of... ; This represents the vertical diffusion parameter, used to characterize the diffusion width of toxic substances in the vertical direction, and is measured in units of... ; Indicates the effective leak source height, in units of ; This represents the natural exponential function.

[0034] It should be noted that the Gaussian plume diffusion model is a continuous point source diffusion model based on statistical theory, used to simulate the concentration distribution of pollutants in the atmosphere along the downwind direction. This model assumes that the pollutant concentration follows a Gaussian distribution in both the crosswind and vertical directions.

[0035] S2. Extraction of Leakage Status Density Features and Biased Initialization of Gray Wolf Population In chemical spill accidents, the degree of danger at different leak points depends not only on the current concentration value, but also on the extent of the expansion of the concentration field boundary, the specificity of the resource demand structure, the degree of insufficiency of surrounding available inventory, and the priority of serious injury protection. Conventional random initialization of the gray wolf population does not utilize this situational information, which can easily generate a large number of low-quality scheduling schemes in the initial stage, resulting in subsequent searches requiring many iterations to approach the feasible area.

[0036] This step first calculates the resource demand density factor, then injects it into the initial gray wolf population's resource allocation ratio and hyperspherical manifold position. This ensures that the initial candidate scheduling scheme, while meeting basic inventory constraints, is more inclined to protect high-risk, high-demand, and high-gap leakage points. The specific steps are as follows: S21. Calculation of resource demand tightness factor based on concentration field changes, demand deviation, and available inventory gap. This step takes the continuous concentration field, resource demand matrix, inventory matrix, and expected transportation time matrix as inputs, and outputs the resource demand density factor corresponding to each leak point. The resource demand density factor is used to characterize the degree to which the leak point receives priority emergency resource support within the current scheduling time window. The specific steps are as follows: 1> In a continuous concentration field Read the first Spatial coordinates of each leak point The local concentration value at the leak point is obtained. , Indicates the first The leak point is at the [number]th [location]. The concentration level of toxic substances within a scheduling time window is used to characterize the intensity of exposure risk at the current location of the leak.

[0037] In practical implementation, if If it happens to be located on a regular grid point in the concentration field, then the concentration value of that grid point is read directly; if If the points are located between grid points, then bilinear or trilinear interpolation is used to obtain the result. For a two-dimensional regular mesh, let... The concentrations of the four surrounding grid points are respectively , , and Then, the four concentration values ​​can be weighted and summed according to the proportions of horizontal and vertical distances to obtain the result. By using interpolation to read data, we can avoid sudden changes in concentration or reading errors caused by the misalignment of the leak point location with the grid division.

[0038] In one embodiment, for example, if the leak point Located in a Within the concentration field grid, the concentration values ​​at its four corner points are respectively , , , (All units are...) If the leak point is located at the exact center in both the horizontal and vertical directions (with a ratio of 0.5 in both directions), then bilinear interpolation is used to calculate the concentration value at that point. .

[0039] 2> with the first Spatial coordinates of each leak point Centered on, in a continuous concentration field The magnitude of the concentration field gradient is calculated to obtain the intensity of the leakage situation change. , Used to characterize the The degree of drastic spatial change in the concentration of toxic substances near each leak point, and the intensity of changes in the leak situation. The larger the value, the steeper the concentration boundary near the leak point, and the more unstable the diffusion pattern. The unit is... .

[0040] In one implementation, if the continuous concentration field Stored using a two-dimensional regular grid, with the horizontal spacing of the grid set to . The vertical spacing of the grid is Then in the first Central difference calculation was used near each leak point. Directional concentration change and The change in directional concentration is calculated as follows: ; ; ; in, Indicates the first Spatial coordinates of each leak point The horizontal coordinate, in units of ; Indicates the first Spatial coordinates of each leak point The vertical coordinate, in units of ; Indicates the first Near the leak point The rate of change of concentration in the direction, in units of ; Indicates the first Near the leak point The rate of change of concentration in the direction, in units of .

[0041] It should be noted that this step does not only determine resource scheduling priority based on the concentration value at the leak point, but also calculates the intensity of changes in the leak situation. When the current concentration value at a leak point is not the highest, but the concentration field gradient amplitude is large, it indicates that the leak point may be located at the leak diffusion boundary or in a rapidly changing risk area, and should also receive higher attention in resource scheduling.

[0042] In one embodiment, such as Figure 3 As shown, an analysis of the continuous concentration field and the distribution of leak points and rescue points in the accident area is presented. The horizontal axis represents the horizontal coordinate (unit: meters), and the vertical axis represents the vertical coordinate (unit: meters). The color intensity in the figure represents the concentration of toxic substances (unit: milligrams per cubic meter). Blue dots represent leak points, and green triangles represent rescue points. The experiment shows that the high concentration area is mainly concentrated near leak point 4, which may be the main source of the leak. The rescue points are distributed around the leak point to facilitate timely rescue.

[0043] 3> According to the first The resource requirement vectors of all leak points within the scheduling time window are used to calculate the i-th... Resource demand deviation at each leak point , Used to characterize the The degree of difference between the resource demand structure of each leak point and the average resource demand structure of all leak points. Resource demand deviation. The larger the value, the more likely it is to be the first. Each leak point has a more prominent need for certain specific resources, and if an average resource allocation method is adopted, it is more likely to cause resource misallocation.

[0044] In practical implementation, we first start with the resource demand matrix. Take the first one from the middle Okay, we get the first Resource requirement vector for each leak point , It refers to the first A vector composed of the number of leakage points and the demand for various emergency resources is a vector with dimension [missing information]. The column vectors; then, the mean vector of all leak point resource demand vectors is calculated. ,Right now .in, Indicates the first The average resource demand structure of all leak points within a scheduling time window is a dimension-based structure. Column vectors.

[0045] Furthermore, calculate the covariance matrix of the resource demand vectors for all leakage points. , Used to characterize the correlation between the demand quantities of different resource categories, it is a dimension-based approach. The matrix is ​​calculated as follows: ; in, This represents the transpose of the corresponding vector.

[0046] Furthermore, Mahalanobis distance with regularization terms is used to calculate the deviation of resource demand. The calculation method is expressed as follows: ; in, This represents the regularization parameter of the covariance matrix, used to avoid the covariance matrix from becoming non-invertible. It can be set to a value that... ; express An identity matrix of order 1, whose dimensions are equal to the total number of resource categories. Consistent; This represents the matrix inversion operation.

[0047] 4> Calculate the number of rescue points based on their inventory and estimated transport time. The achievable inventory gap at each leak point , Used to characterize, under current traffic and inventory conditions, the first The difficulty of obtaining resource support from surrounding aid stations at each leak point. The extent of achievable inventory shortage. The larger it is, the more likely it is to be the first The more difficult it is for the resource needs corresponding to a leak point to be met by a rescue point that can be reached in a short period of time.

[0048] In the specific implementation, first calculate the first... The leak point is related to the first Time discounts on emergency resources can be applied to inventory. The calculation method is expressed as follows: ;

[0049] in, Indicates the first The leak point is at the [number]th [location]. Within the scheduling time window, for the first Time discounts on emergency resources can be applied to inventory; This parameter represents the response time scale, used to control the discount strength of transportation time on available inventory; it can be 15 minutes, 30 minutes, or 60 minutes. This represents the natural exponential function.

[0050] Furthermore, based on the quantity of demand And time-limited discounts available in stock Calculate the relative gap for each resource category and average it across resource categories to obtain the achievable inventory gap. The calculation method is expressed as follows: ;

[0051] in, This represents a non-negative truncation function, used to count only the portion of resources that are insufficient. This represents the demand stability constant, used to prevent the denominator from being zero, and can be taken as... .

[0052] 5> The concentration values ​​at the leakage points were respectively... Intensity of changes in leakage status Resource demand deviation and available inventory gap Normalization is performed to obtain concentration risk weights. Diffusion change weight Demand deviation weight and inventory gap weight In one implementation, the min-max normalization method is preferred for normalization processing.

[0053] 6> Weight concentration risk Diffusion change weight Demand deviation weight and inventory gap weight Fusion, to obtain the first Resource demand density factor for each leak point Resource demand intensity factor Used for comprehensive characterization of the first The leak point is at the [number]th [location]. Resource guarantee priority within each scheduling time window.

[0054] In the actual implementation, the unnormalized tightness score is first calculated. The calculation method is expressed as follows: ; in, Indicates the first The leak point is at the [number]th [location]. The corresponding unnormalized density score within each scheduling time window; This represents the concentration risk weighting coefficient, used to control the strength of the current concentration's impact on priority. This represents the diffusion change weighting coefficient, used to control the intensity of the influence of spatial changes in the concentration field on the priority. This represents the demand deviation weighting coefficient, used to control the intensity of the impact of the specificity of resource demand structure on priority. This represents the inventory gap weighting coefficient, used to control the intensity of the impact of insufficient available inventory on priority; the above four weighting coefficients satisfy... And all are not less than 0. In a convenient implementation method, one can take... , , , If the accident is in its early, rapid spread stage, it can improve... If resources are severely scarce in the accident area, it can be improved. .

[0055] Furthermore, Softmax normalization is performed on the unnormalized density scores of all leakage points to obtain the resource demand density factor. The calculation method is expressed as follows: ; in, This represents the density differentiation coefficient, used to control the degree of amplification of the difference between high-priority and low-priority leak points, and can be taken from 1 to 3. This represents the normalized stability constant of the density, which can be taken as... ; This represents the natural exponential function.

[0056] It should be noted that the resource demand tightness factor uses continuous numerical weights, which can avoid simply classifying leakage points into high-risk and low-risk categories, thus preventing sudden changes in scheduling and enabling the resource scheduling scheme to have a smoother adjustment capability when risks change.

[0057] S22. Discrete scheduling population with biased initialization based on resource demand density factor. This step takes the resource demand density factor, resource demand matrix, and inventory matrix as inputs to generate an initial discrete emergency resource scheduling scheme that satisfies the basic constraints of rescue point inventory, and maps it to a hyperspherical manifold location. The specific steps are as follows: 1> Regarding the first The gray wolf generates an initial discrete emergency resource scheduling scheme. , It refers to the first The candidate resource scheduling scheme corresponding to the gray wolf in the initialization phase describes the quantity of various emergency resources transferred from each rescue point to each leak point, and is a dimension-based scheme. Column vectors; This represents the individual index of the gray wolf. ; This indicates the size of the gray wolf population, and can be expressed as 30, 40, or 60.

[0058] In practical implementation, a single decision variable in the initial discrete emergency resource scheduling scheme is denoted as... , Indicates the first In the initial scheduling scheme corresponding to the first gray wolf, the first... The rescue point to the first The first leak point was transferred. The quantity of emergency resources. (All) According to the "Rescue Point Index" Resource Category Index Leakage point index Expanding the sequence of "" yields the initial discrete emergency resource scheduling scheme. For example, when , , hour, The expansion dimension is .

[0059] 2> For each rescue point and each resource category Based on the resource demand density factor and demand quantity Constructing the probability of leak point assignment , Used to control the The initial allocation tendency of emergency resources among different leak points is calculated as follows: ; in, Indicates the first Within the scheduling time window, the first Emergency resources are allocated to the first The probability weight of each leakage point; Represents the probability stability constant, which can be taken as... ; The density influence index is used to control the strength of the influence of the resource demand density factor on the initial allocation probability. It can be 0.5 to 2.0, with 1 being preferred. This represents the demand quantity impact index, used to control the strength of the influence of actual demand quantity on the initial allocation probability. It can be taken from 0.5 to 2.0, with 1 being the preferred value.

[0060] It should be noted that leak points with high resource demand density factors and high demand quantities will receive a greater initial allocation probability; leak points with low resource demand density factors or no actual demand will not be completely excluded, but rather... A smaller probability is retained to ensure that the initial population has the necessary diversity.

[0061] In one embodiment, such as Figure 4 As shown, a heatmap of the initial allocation probability weights of various resources to each leak point is analyzed; the horizontal axis represents the leak point (dimensionless), and the vertical axis represents the resource category (dimensionless). The color intensity and labeled values ​​in the figure represent the allocation probability weights (dimensionless). The experiment shows that leak points with high density and high demand receive a greater initial allocation probability, but other leak points still retain a smaller probability, reflecting the characteristic of biased initialization that ensures both focus and diversity.

[0062] 3> According to the first The first rescue point Emergency resource inventory quantity Assigning probability to leakage points Generate the initial resource quantity corresponding to each leak point. .

[0063] In one implementation, first according to Generate allocation ratios with random perturbations. The calculation method is expressed as follows: ; in, Indicates the first Within the scheduling time window, the first The first gray wolf in The first rescue point, the first Category of emergency resources, the first The initial allocation ratio after adding random perturbations to each leakage point dimension; Indicates the first The first gray wolf in The first rescue point, the first Category of emergency resources, the first The random perturbation factor on each leakage point dimension can be obtained from... Uniform sampling within the range.

[0064] Further, calculate the number of consecutive initial allocations, i.e. ;in, Indicates the first The first gray wolf in The first rescue point, the first Category of emergency resources, the first The number of initial contiguous allocations that are not integerized across each leak point dimension.

[0065] Furthermore, on Perform floor rounding to obtain the initial integer allocation quantity, and record the decimal remainder. If the same rescue point... and the same resource category If there is still remaining inventory after rounding, one resource will be added one by one in descending order of the decimal remainder, until the added quantity reaches the remaining inventory or the corresponding leak point demand limit.

[0066] In one embodiment, as an example, suppose that for rescue point 1 and resource 1, the number of consecutive allocations to the three leak points is generated according to the allocation probability. , , First, round down to the nearest integer to get the initial allocation quantities of 5, 12, and 0, with a total allocation of 17. At this point, the remaining inventory is... The decimal remainders are 0.8, 0.3, and 0.9 respectively. The leak points are sorted from largest to smallest as follows: Leak Point 3 (0.9), Leak Point 1 (0.8), and Leak Point 2 (0.3). One resource is added to each of Leak Point 3 and Leak Point 1 in turn. After inspection, it is found that the demand limit of Leak Point 3 is 1. Therefore, after adding one resource, the limit is reached and no more resources can be added. The remaining resource slot is allocated to Leak Point 2. The final allocation is 6 resources for Leak Point 1, 13 resources for Leak Point 2, and 1 resource for Leak Point 3.

[0067] 4> Initial discrete emergency resource scheduling scheme Perform demand ceiling correction to avoid allocating significantly more resources than required to a leak point during the initialization phase.

[0068] In the actual implementation, for each leak point and each resource category Calculate the number of rescue points to be transferred to the leak point. Total amount of emergency resources If the total quantity exceeds the demand quantity With allowable redundancy coefficient The product of these factors will be used to deduct any excess amount from rescue points with longer transport times or higher unit transport costs, until the following condition is met: ;in, This represents the redundancy factor, used to allow for a small amount of safe redundancy allocation, and can be taken as 1.05 to 1.20. Based on this, it can be avoided that the initial population concentrates a large amount of resources on a single high-density leak point, causing the basic needs of other leak points to be squeezed out.

[0069] In one embodiment, as an example, assuming that leak point 1 requires 10 sealing tools, with an allowable redundancy factor... That is, a maximum of 11 items can be allocated; in the current plan, 15 items were allocated to the 3 rescue points, with 4 items exceeding the limit; after inspection, it was found that the transportation time of rescue point 3 was the longest, so the allocation of rescue point 3 was deducted first until the total amount did not exceed 11 items.

[0070] 5> Based on the resource demand density factor Constructing the expansion compactness weight vector Expand the density weight vector This refers to the weight vector formed by extending the resource demand density factor at the leak point level to each scheduling decision dimension, and its dimension is the same as that of the initial discrete emergency resource scheduling scheme. Consistent.

[0071] In specific implementation, for the first The first rescue point, the first Emergency resources and the first The scheduling dimension corresponding to each leak point can be set to the expansion tightness weight. ;in, Represents the expansion density weight vector The corresponding number in the middle The first rescue point, the first Type of emergency resources and the first The weight value of each leakage point; This represents the initial bias enhancement coefficient, used to control the influence of the resource demand density factor on the initial population distribution, and can be taken as 1.0 to 2.0.

[0072] It should be noted that the following is adopted: The calculation method can improve the continuous mapping strength of the scheduling dimension corresponding to the high resource demand density leakage point, while avoiding the dimension corresponding to the low resource demand density leakage point being compressed to 0, thus taking into account both key protection and overall coverage.

[0073] In one embodiment, for example, if , , , Then for all and The dimension weight corresponding to leak point 1 is The dimension weight corresponding to leak point 2 is .

[0074] 6> Initial discrete emergency resource scheduling scheme Convert to the first The initial hyperspherical manifold position of the gray wolf , It refers to the first The initial position of the lead gray wolf in the continuous search space is used to participate in subsequent gray wolf position updates.

[0075] In the specific implementation, the initial discrete emergency resource scheduling scheme is first... Perform logarithmic compression to reduce the dominance of large inventory resources on continuous location scales, and then combine it with the expanded density weight vector. Element-wise multiplication yields a weighted continuous vector. The calculation method is expressed as follows: ; in, Indicates the first The weighted continuous vector corresponding to the gray wolf. Dimensions and Consistent; Represent the natural logarithm function; This indicates element-wise multiplication.

[0076] Furthermore, for weighted continuous vectors Perform L2 norm normalization to obtain the initial hyperspherical manifold position. , Dimensions and Consistent.

[0077] It should be noted that the hyperspherical manifold here refers to a hypersphere with a unit radius centered at the origin, where all gray wolf positions are distributed. Constraining the gray wolf positions to a hyperspherical manifold, i.e., using hyperspherical manifold normalization, ensures that the continuous positions of different gray wolf individuals have a uniform scale. This helps improve search stability and convergence speed, while preventing rescue points with large inventory sizes or resource categories with large orders of magnitude from having an excessively strong influence on continuous position updates. In addition, logarithmic compression and resource demand density weight injection together ensure that the initial search direction reflects both the resource quantity structure and the priority of the incident situation.

[0078] 7> Repeat steps 1> through 6> until the generated size is [size missing]. The initial gray wolf population. The initial gray wolf population included An initial discrete emergency resource scheduling scheme and its corresponding The initial hyperspherical manifold position.

[0079] S23. Discrete scheduling back projection that satisfies inventory constraints and resource coupling constraints. Since the hyperspherical manifold position in the Grey Wolf optimization process is a continuous search variable and cannot be directly used as an emergency resource scheduling instruction, this step converts the hyperspherical manifold position into a discrete emergency resource scheduling scheme in non-negative integer form. During the conversion process, the upper limit of inventory, the upper limit of demand, the minimum guarantee of critical materials, and resource coupling constraints are corrected. The specific steps are as follows: 1> Construct an inverse scaling vector based on the inventory quantity at rescue points and the required quantity at leak points. , Used to restore the location of a hyperspherical manifold to the scale of actual resource quantity, with its dimension consistent with the location of the hyperspherical manifold.

[0080] In specific implementation, for the first The first rescue point, the first Type of emergency resources and the first The dimension corresponding to each leak point will be inversely scaled to the vector elements. Set to the smaller value between the inventory quantity and the demand quantity, i.e. ;in, Represents the inverse scaling vector The corresponding number in the middle The first rescue point, the first Type of emergency resources and the first The scale value of each leak point. Based on this, it is possible to avoid directly using a large inventory quantity when backprojecting consecutive locations, which could lead to a leak point receiving resources that significantly exceed its needs.

[0081] 2> The first Hyperspherical position of the gray wolf With inverse scaling vector Element-wise multiplication yields a continuous vector of resource quantities. , This refers to the number of non-integerized resource schedules recovered from consecutive search positions.

[0082] In the actual implementation, non-negative truncation is performed first, and then the continuous resource quantity vector is obtained. ,Right now ;in, This indicates element-wise non-negative truncation, used to correct negative values ​​to 0; Indicates the first The current hyperspherical manifold position of the gray wolf; This indicates element-wise multiplication.

[0083] It should be noted that while the gray wolf position update may generate a negative component, the emergency resource allocation quantity cannot be negative. Therefore, a non-negative truncation must be performed during backprojection. This allows for greater exploration freedom in the continuous search process while ensuring that the discrete scheduling results satisfy the basic physical meaning.

[0084] 3> For continuous resource quantity vectors Integerization is performed to obtain the initial discrete resource quantity. The integerization process uses a combination of floor function and maximum remainder compensation to reduce inventory overruns caused by simple rounding.

[0085] In practical implementation, this applies to each rescue point. and each resource category First, check all leak points. Corresponding number of continuous resources Perform a floor function to get And calculate the decimal remainder. Then, calculate the compensable quantity of that resource category at the rescue point, expressed as follows: ; in, Indicates the first The first gray wolf in The first rescue point, the first The integer quantity that can still be added under the category of emergency resources; This indicates an indicator function that takes the value 1 if the condition within the parentheses is true, and 0 otherwise. This represents the function that takes the minimum value.

[0086] Furthermore, according to the decimal remainder Add one resource to the corresponding leak point in descending order of size, until the number of additions reaches a certain threshold. There may be no more leak points to add. Based on this, we can try to retain the proportion of consecutive search positions after integerization.

[0087] 4> Project the inventory upper limit onto the initial discrete resource quantity to ensure that the total amount of each type of emergency resource mobilized from each rescue point does not exceed the corresponding inventory quantity.

[0088] In specific implementation, if there exists Then, for the first The first rescue point and the first Emergency resources are allocated proportionally to each leak point, and the calculation method is expressed as follows: ; in, Indicating the first step in the process of projecting the inventory ceiling The first gray wolf plan The rescue point to the first The first leak point was transferred. The quantity of temporary resources for emergency response; Indicates the number of times the inventory ceiling projection is executed. The first gray wolf plan The rescue point to the first The first leak point was transferred. The quantity of discrete resources in the category of emergency resources; This represents the inventory projection stability constant, used to prevent the denominator from being zero, and can be taken as... ; This represents a floor function; if there is still remaining allocable inventory after reduction, it is allocated according to the principle of prioritizing replenishment of leakage points with greater losses before and after reduction, until the inventory limit is not exceeded. Based on this, the resource allocation after the inventory limit is projected can still be as close as possible to the original candidate scheme.

[0089] 5> Implement a minimum guarantee projection for critical materials on the initial discrete resource quantity to ensure a minimum satisfaction rate of life-saving materials for severely injured leak points.

[0090] In specific implementation, let's assume Indicates the first The set of critical leak points within a scheduling time window; This refers to a collection of life-saving supplies, which may include first aid equipment, chemical protective suits, and necessary medical protective equipment. Indicates the first The minimum satisfaction rate for life-saving supplies at a single, severely injured leak site can be taken as 0.9 to 1.0. For each... The calculation method for the allocated quantity of life-saving supplies is as follows: ; in, Indicates the first The first gray wolf corresponds to the scheduling scheme of the first gray wolf. The total quantity of life-saving supplies obtained from the most severely damaged leak points. If If supplies are still in stock, life-saving supplies will be replenished from relief points that still have remaining inventory; replenishment will be based on the estimated arrival time. Rescue points are selected from shortest to longest time, with priority given to those with larger remaining resource inventories. If the remaining inventory at all rescue points is sufficient to meet the minimum guarantee requirements, a dispatch plan that meets the minimum guarantee projection for critical materials is formed after supplementary allocation. If the remaining inventory is insufficient, available life-saving materials are prioritized for allocation to severely injured leak points with more casualties or higher hazard levels, and the gap is recorded in the subsequent hard constraint violation record.

[0091] In one embodiment, for example, if the set of severely damaged leak points... The leak point is 1, and the total demand for life-saving supplies is 100 pieces, with a minimum fulfillment rate of [missing information]. That is, at least 90 items are needed, but the current scheduling scheme only allocates 70 items, leaving a gap of 20 items. The algorithm will first replenish from the rescue point with the shortest expected arrival time and still have inventory. If rescue point 1 is expected to arrive in 15 minutes and has 10 items in inventory, then all items will be transferred. If rescue point 2 is expected to arrive in 30 minutes and has sufficient inventory, then the remaining 10 items will be replenished. If the total available inventory is only 15 items, then only 15 items can be replenished, and a gap of 5 items will be recorded in the hard constraint violation degree.

[0092] 6> Perform resource coupling constraint projection on the initial discrete resource quantity to ensure that the basic operation ratio relationship is met among the resources that need to be used together.

[0093] In practical implementation, if the first Type of emergency resources and the first Emergency resources of this type have a coupling relationship (for example, leak-stopping tools require chemical protective suits, and neutralizing agents require protective equipment), then let... Indicates the first Emergency resources of this type are relative to the first Minimum allocation ratio of emergency resources. For each leak point. Check if the following relationship is satisfied: ; in, Indicates the first The first gray wolf in the corresponding scheduling scheme The rescue point to the first The first leak point was transferred. The quantity of emergency resources; Indicates the first The first gray wolf in the corresponding scheduling scheme The rescue point to the first The first leak point was transferred. The number of emergency resources of each type. If this constraint is not met, priority will be given to those with shorter estimated arrival times and still some time remaining. Replenishment and allocation of remaining emergency resources at rescue points Emergency resources; if insufficient, the amount will be reduced according to the coupling ratio. The quantity of emergency resources should be carefully considered to avoid situations where there are tools for plugging leaks but a lack of protective resources, or neutralizing agents but a lack of supporting protective equipment, making such deployment combinations impossible to execute.

[0094] In one embodiment, for example, if the minimum matching ratio of leak-sealing tools to chemical protective suits is 1:2, that is, each set of leak-sealing tools requires at least 2 sets of chemical protective suits, and a candidate solution allocates 10 sets of leak-sealing tools to the first leak point but only allocates 12 sets of chemical protective suits, then the resource coupling constraint projection will prioritize the allocation of 8 sets of chemical protective suits; if the inventory of chemical protective suits is insufficient, the number of leak-sealing tools will be reduced accordingly to make the final solution feasible for on-site operation.

[0095] 7> Expand the resource quantities after completing the projection of the upper limit of inventory, the minimum guarantee of key materials, and the resource coupling constraint into a discrete emergency resource scheduling scheme. , It refers to the first The non-negative integer emergency resource scheduling scheme corresponding to the gray wolf is used to enter the subsequent fitness evaluation.

[0096] In one implementation, the discrete scheduling back-projection process can be represented as: ; in, This represents the feasible region projection operator, used to correct the initial integer resource quantity to the feasible region for resource scheduling. ; The feasible region for resource scheduling refers to the set of scheduling schemes jointly limited by the upper limit of inventory, the upper limit of demand, the minimum guarantee of critical materials, resource coupling, and non-negative integer constraints. Integerization operator, used to map the number of contiguous resources to the number of integer schedules; This represents the function that takes the maximum value. This indicates element-wise multiplication.

[0097] It should be noted that discrete scheduling back projection is not a simple rounding, but rather a step-by-step correction according to the actual scheduling constraints after rounding. This allows the continuous search capability of the Grey Wolf algorithm to be combined with the discrete constraints of emergency resource scheduling, reducing the interference of unexecutable candidate solutions on the search process.

[0098] S3. Iterative Search of Gray Wolves Based on Density Fluctuation Awareness and Hard-Constraint Priority Scheduling Strategy After S2 processing, the initial gray wolf population already has a scheduling tendency towards high resource demand and dense leak points. However, chemical leak accidents have strong time-varying characteristics. Wind speed, diffusion boundary, road traffic conditions, and inventory consumption may all change the merits of candidate solutions. Conventional gray wolf optimization algorithms use fixed convergence factors and single fitness ranking, which are prone to premature convergence when the accident situation fluctuates. They are also prone to sacrificing life rescue protection at severely injured leak points due to better cost objectives.

[0099] This step uses fitness distribution entropy to perceive the degree of population dispersion, uses fluctuations in resource demand density to perceive changes in the accident situation, and uses a hard-constraint priority evaluation mechanism to ensure that life-saving protection takes precedence over cost and time optimization. The specific steps are as follows:

[0100] S31. Population search state perception based on fitness distribution entropy This step involves statistically analyzing the fitness distribution of discrete emergency resource scheduling schemes within the current gray wolf population and calculating the normalized fitness distribution entropy. This entropy is used to determine whether the gray wolf population is in a state of global exploration, local exploitation, or overcrowding. The specific steps are as follows: 1> In the In the nth iteration, for the th The discrete emergency resource scheduling scheme corresponding to the gray wolf Calculate temporary fitness value ,in, This indicates that the Grey Wolf optimizes the iterative index. ; This represents the maximum number of iterations, which can be between 100 and 300. Indicates the first The first gray wolf in Discrete emergency resource scheduling scheme in the next iteration; Indicates the first The first gray wolf in Temporary fitness value in the next iteration.

[0101] In practical implementation, the temporary fitness value This is used only for statistical analysis of population distribution and does not directly replace the evaluation of the final scheduling scheme. The provisional fitness value can be obtained by weighting the hard constraint violation rate, normalized total scheduling cost, and normalized maximum response time, as expressed in the following formula: ; in, Represents a discrete emergency resource scheduling scheme The degree of violation of hard constraints; Represents a discrete emergency resource scheduling scheme The normalized total scheduling cost; Represents a discrete emergency resource scheduling scheme Normalized maximum response time; This represents the penalty coefficient for violating hard constraints, and can be taken from 100 to 1000. This indicates the cost statistical weight, which can range from 0.1 to 1.0; This represents the time-based statistical weight, which can range from 0.1 to 1.0.

[0102] 2> Based on the size of the gray wolf population Determine the number of fitness partitions , This indicates the number of intervals into which the current gray wolf population's temporary fitness values ​​are divided, used to calculate the empirical distribution. (Number of fitness intervals) Press It is confirmed that, among them, This represents the floor function; This represents a logarithmic function with base 2. For example, when... hour, Therefore, it is acceptable. .

[0103] In the actual implementation, first read the current population Find the minimum and maximum values, and then divide the interval between the minimum and maximum values ​​into equal-width segments. There are fitness intervals; if all temporary fitness values ​​in the current population are exactly the same, it means that the population is highly concentrated in the target space, and the normalized fitness distribution entropy can be directly set to 0.

[0104] 3> Count the number of gray wolves within each fitness interval to obtain the interval count. , Indicates the first In the next iteration, it falls into the first The number of individual gray wolves within each fitness interval; Indicates the fitness range index. .

[0105] Furthermore, based on the number of intervals Calculate the fitness distribution entropy The calculation method is expressed as follows: ; in, Indicates the first The fitness distribution entropy at the next iteration is used to characterize the degree of dispersion of the current gray wolf population in the target space; Represents the logarithmic function with base 2; This represents the stable parameter for entropy calculation, used to avoid taking the logarithm with respect to 0, and can be taken as... .

[0106] 4> The fitness distribution entropy Dividing by the maximum possible entropy yields the normalized fitness distribution entropy. ,Right now ;in, Indicates the first The normalized fitness distribution entropy at the next iteration has a value range close to... . The larger the value, the more dispersed the discrete emergency resource scheduling schemes for the gray wolf population are across multiple fitness ranges, indicating that exploratory capabilities are still needed. The smaller the value, the more concentrated the gray wolf population is within a small fitness range, indicating that local development can be strengthened or diversity correction can be triggered.

[0107] In one embodiment, for example, if the temporary fitness values ​​corresponding to 40 gray wolves are relatively evenly distributed within 8 fitness intervals, then the normalized fitness distribution entropy is... A value close to 1 indicates that the current search has not yet converged; if all 35 gray wolves are concentrated in two adjacent intervals, then... The significant decrease indicates that the current search has entered a state of local concentration.

[0108] S32. Nonlinear convergence factor driven by both fitness entropy fluctuation and tightness fluctuation. This step adaptively adjusts the search control factor of the Gray Wolf optimization algorithm based on the historical fluctuations of the normalized fitness distribution entropy and resource demand density factor. This ensures that the search process maintains strong exploration capabilities when the accident situation is unstable and accelerates convergence when the situation is stable. The specific steps are as follows: 1> Record recent The normalized fitness distribution entropy of each iteration forms a local entropy sequence. , Used to characterize recent changes in the current search status of the gray wolf population. This represents the length of the entropy fluctuation statistical window, which can be 5 to 10.

[0109] In practical implementation, when When using the first to the second The normalized fitness distribution entropy of the next iteration constitutes ;when When, use the first Next to The normalized fitness distribution entropy of the next iteration constitutes .

[0110] 2> Based on local entropy sequence Calculate entropy fluctuation value , The method used to characterize the fluctuations in the search status of the gray wolf population in recent iterations is expressed as follows: ; in, Represents a local entropy sequence The number of elements in; Represents a local entropy sequence The first in A normalized fitness distribution entropy element; Represents a local entropy sequence The mean of the local entropy sequence. If the relative merits of different discrete emergency resource scheduling schemes change frequently with the accident situation, then the local entropy sequence... Large fluctuations, entropy fluctuation value Increase; if the search process gradually stabilizes, the entropy fluctuation value... Decrease.

[0111] 3> Based on the first The resource demand density factor between the current scheduling time window and the previous scheduling time window is used to calculate the resource demand density factor of the current scheduling time window. The fluctuation value of resource demand density of each scheduling time window relative to the previous scheduling time window , The method used to characterize the changes in accident status and resource priority between adjacent scheduling time windows is expressed as follows: ; in, Indicates the first The fluctuation value of resource demand density within each scheduling time window; Indicates the first The resource demand density factor of each leakage point in the previous scheduling time window; This represents the absolute value operation. When... At that time, it can be made .

[0112] It should be noted that fitness distribution entropy mainly reflects the internal state of the optimization population, while resource demand density fluctuation mainly reflects changes in the external accident situation. Combining the two can prevent the algorithm from determining the convergence rhythm solely based on whether the population is concentrated, thus making it more suitable for rolling scheduling scenarios in chemical leak accidents.

[0113] 4> Based on the current normalized fitness distribution entropy Entropy fluctuation value Fluctuation value of resource demand tightness Calculate the search retention index The calculation method is expressed as follows: ; in, Indicates the first The scheduling time window, the first The search hold exponent corresponding to each iteration is used to control the nonlinear decay rate of the search control factor; This represents the minimum entropy compensation parameter, used to prevent the search control factor from prematurely reaching zero when the normalized fitness distribution entropy is too low; it can be set to 0.05. This represents the entropy fluctuation scaling factor, which can range from 0.5 to 1.0; This represents the scaling factor for density fluctuations, which can range from 0.5 to 1.5.

[0114] 5> Based on search retention index Calculate search control factors The calculation method is expressed as follows: ; in, Indicates the first The scheduling time window, the first The search control factor corresponding to the next iteration is used to adjust the magnitude of the gray wolf's position update. When the time approaches 2, the gray wolf population maintains a strong global exploration capability; When the population approaches zero, the gray wolf population shifts to localized exploitation.

[0115] In one embodiment, for example, if wind speed changes significantly in the early stages of a leak, causing frequent shifts in the ranking of resource demand density factors at different leak points, then... Larger, search retention index Increase, search control factor As the descent rate slows, the gray wolf population retains its exploration capabilities for a longer period; if the leakage is controlled, the resource demand density factor changes little, and the normalized fitness distribution entropy decreases, then the search control factor... Accelerated descent allows the algorithm to converge to a stable scheduling scheme more quickly.

[0116] S33. Hyperspherical manifold position update incorporating resource demand density modulation This step updates the hyperspherical manifold position based on the gray wolf pack hierarchical coordination mechanism, and continuously modulates it using a resource demand density factor after the position update, so that high-priority leakage points continuously obtain resource allocation preferences during the iterative search process. The specific steps are as follows:

[0117] 1> Screening based on the current evaluation results of the gray wolf population Wolf, wolves and Wolves. Among them, The wolf represents the leader wolf, which has the best overall performance in the current population. The wolf represents the alpha wolf, which has the second-best overall performance in the current population. The wolf represents the leader wolf, which has the third best overall performance in the current population; Wolf, wolves and The wolves collectively guide the other gray wolves to update the position of the hyperspherical manifold.

[0118] In its implementation, the leader wolf selection adopts a hard constraint priority principle: if there are discrete emergency resource scheduling schemes with a hard constraint violation degree of 0 in the current population, then only among the candidate schemes with a hard constraint violation degree of 0, non-dominated ranking is performed based on total scheduling cost, maximum response time, and non-critical resource gap, combined with congestion degree for selection. Wolf, wolves and Wolves; if there are no candidate solutions with a hard constraint violation degree of 0 in the current population, the candidate solution with a lower hard constraint violation degree is selected as the leader wolf to guide subsequent searches to approach the feasible region of resource scheduling. Crowding degree is used to characterize the sparsity of candidate solutions in the multi-objective space. The greater the crowding degree, the more conducive the candidate solution is to maintaining the diversity of the solution set.

[0119] In one embodiment, for example, in the first In this iteration, 10 solutions in the population satisfy the hard constraints. Through non-dominated sorting, schemes 1, 2, and 3 are in the first non-dominated level and do not dominate each other. Among them, scheme 1 has the highest crowding (i.e., its neighboring schemes are the farthest away), followed by scheme 2, and scheme 3 has the lowest. Therefore, scheme 1 is selected. Wolf, Option 2 is Wolf, Option 3 is Wolf.

[0120] 2> Regarding the first The first gray wolf in Hyperspherical manifold position at the next iteration Perform gray wolf position update to obtain candidate consecutive positions. , Indicates the first The first gray wolf in Candidate consecutive positions obtained by the leader wolf in the next iteration.

[0121] In specific implementation, respectively Wolf, wolves and The position of the wolf's hyperspherical manifold is updated with reference. Let... express wolves in Position of the hyperspherical manifold in the next iteration; express wolves in Position of the hyperspherical manifold in the next iteration; express wolves in The position of the hyperspherical manifold in the next iteration. For any leader wolf Generate random vectors and Each element of the random vector is from Uniform sampling; then calculate the coefficient vector. and The calculation method is expressed as follows: ; ; in, Indicates the first The search stride coefficient vector corresponding to the leader wolf; Indicates the first The perturbation coefficient vector corresponding to the leader wolf class.

[0122] Furthermore, calculate the first... Alpha gray wolf relative to leader wolf Distance vector and candidate guiding position The calculation method is expressed as follows: ; ; in, Indicates the first In the nth iteration Alpha gray wolf relative to leader wolf The continuous search distance; Indicates the first During the nth iteration The alpha gray wolf is led by the alpha wolf Candidate positions generated by guidance; Indicates the first During the nth iteration The hyperspherical manifold position of the wolf-like leader; This indicates taking the absolute value of each element; This indicates element-wise multiplication.

[0123] Finally, the average of the three candidate guiding positions is used to obtain the candidate continuous positions. The calculation method is expressed as .

[0124] 3> Based on the resource demand density factor Constructing a density modulation vector , This is used to further strengthen the scheduling dimension corresponding to the high resource demand density leakage point after the location is updated. Its dimension is consistent with the location of the hyperspherical manifold.

[0125] In specific implementation, for the first The first rescue point, the first Type of emergency resources and the first The dimension corresponding to each leakage point can be set to the density modulation weight. ;in, Represents the density modulation vector The corresponding number in the middle The first rescue point, the first Type of emergency resources and the first Modulation weights of each leakage point; This represents the density modulation enhancement factor, which can be taken from 0.3 to 1.0.

[0126] 4> Select consecutive candidate positions With density modulation vector Multiply element-wise and reproject onto the unit hypersphere to obtain the updated hyperspherical manifold position. The calculation method is expressed as follows: ; ; in, Indicates the first The first gray wolf in Position of the hyperspherical manifold in the next iteration; Represents the L2 norm; This indicates element-wise multiplication.

[0127] It should be noted that this step uses continuous density modulation, which can avoid drastic changes in the resource allocation structure, so that the scheduling scheme can protect high-risk leakage points while still retaining the necessary basic protection for medium and low-risk leakage points.

[0128] 5> Update the position of the hyperspherical manifold Input the discrete scheduling back projection process in S203 to obtain the first... The first gray wolf in Discrete emergency resource scheduling scheme in the next iteration .

[0129] In one embodiment, for example, if the resource demand density factor of the fourth leak point increases significantly within the current scheduling time window, then the density modulation vector... This will increase the continuous position weight of each resource dimension corresponding to the fourth leakage point; after unit hypersphere renormalization and discrete scheduling back projection, the critical material configuration level of the fourth leakage point will be enhanced under the premise of meeting inventory and coupling constraints; if the risk of the fourth leakage point decreases in the subsequent scheduling time window, the corresponding modulation weight will naturally weaken, and resources can be gradually shifted to other high-density leakage points.

[0130] S34. Fitness Evaluation of the Two-Level Scheduling Scheme with Hard Constraint Priority

[0131] This step first evaluates the hard constraint violation rate of discrete emergency resource scheduling schemes, and then evaluates the total scheduling cost, maximum response time, and non-critical resource gaps of the scheduling schemes that meet the hard constraints, so as to achieve coordinated optimization of safety priority, timeliness, and economy. The specific steps are as follows: 1> For the discrete emergency resource scheduling scheme to be evaluated Calculate the degree of hard constraint violation , It is used to measure the degree to which discrete emergency resource allocation schemes violate inventory limits, minimum guarantees of life-saving supplies, and resource coupling relationships.

[0132] In practical implementation, for discrete emergency resource scheduling schemes Its hard constraint violation degree Inventory violation Violation of life rescue and protection standards and resource coupling violation The weighted average is obtained, and the calculation method is expressed as follows: ; in, This indicates that the inventory violates the weighting rules; This indicates that life-saving and protection measures violate weighting; This indicates that resource coupling violates weights. To emphasize the priority principle of life-saving, it can be made... Greater than and For example, it is advisable , , .

[0133] In this implementation, a discrete emergency resource scheduling scheme is used. Corresponding inventory violation rate It can be represented as: ; in, Represents a discrete emergency resource scheduling scheme The Middle The rescue point to the first The first leak point was transferred. The quantity of emergency resources; This represents the function that takes the maximum value.

[0134] In this implementation, a discrete emergency resource scheduling scheme is used. Corresponding violation of life-saving guarantee It can be represented as: ; in, Indicates the first The importance weight of each serious injury leak point is used to reflect the difference in the number of injured or the level of hazard. It can be set according to the proportion of injured or the level of hazard. For example, if there are two serious injury leak points, with 30 people at leak point 1 and 10 people at leak point 2, then the importance weight can be set accordingly. ; This represents the function that takes the maximum value.

[0135] In this implementation, a discrete emergency resource scheduling scheme is used. Corresponding resource coupling violation This can be obtained by summing the gaps in all resource coupling relationships. Let... This represents a set of resource coupling relationships, where each element is... , indicating the first Emergency resources need the first Emergency resources at the minimum proportion Matching It can be represented as: .

[0136] It should be noted that when At that time, discrete emergency resource scheduling scheme Satisfy all hard constraints; when At that time, discrete emergency resource scheduling scheme If a hard constraint is violated, the solution will be given lower priority than the solution that satisfies the hard constraint when selecting the leader wolf and outputting the final strategy.

[0137] It should also be noted that, Indicates the first The first gray wolf in the corresponding scheduling scheme The first rescue point towards the The first leak point was transferred. The number of emergency resources, It is the first The significance of the wolf-headed plan; Represents a discrete emergency resource scheduling scheme The Middle The first rescue point towards the The first leak point was transferred. The number of emergency resources, It is a component of any scheme. When there is no need to distinguish individual gray wolves, it is only used to represent the first... The rescue point to the first The first leak point was transferred. When considering the quantity of emergency resources, adopt To represent; when it is necessary to clearly distinguish individual gray wolves, use To represent. If ,but .

[0138] 2> For a discrete emergency resource scheduling scheme with a hard constraint violation degree of 0, calculate the total scheduling cost. , Used to evaluate resource allocation costs, transportation costs, and additional costs associated with emergency dispatch. In practical implementation, the total dispatch cost... The calculation method is expressed as follows: ; in, Indicates the first The rescue point called the first The unit cost of deploying emergency resources; Indicates the first Within the scheduling time window, the first The rescue point to the first The transportation cost of a single resource from a leak point can be determined by road distance, vehicle cost, and detour cost. Therefore, long-distance, large-volume, and high-cost resource allocation increases the total scheduling cost, thus guiding scheduling schemes to reduce unnecessary resource waste and long-distance transport while meeting safety constraints.

[0139] In one embodiment, such as Figure 5As shown, the convergence curve of the total scheduling cost is analyzed; the horizontal axis represents the number of iterations (dimensionless), and the vertical axis represents the total scheduling cost (dimensionless). Experiments show that the total scheduling cost is relatively high in the early stage of the search, gradually decreases and tends to stabilize with iterations, proving that the algorithm can optimize resource scheduling and continuously improve the economy of the scheduling scheme after satisfying hard constraints.

[0140] 3. For discrete emergency resource scheduling schemes with a hard constraint violation degree of 0, calculate the maximum response time. , This is used to evaluate the time when the latest batch of transported resources arrived at the leak point. In practice, only combinations of actual transported rescue points and leak points are counted, and the calculation method is expressed as follows: ; in, Indicates the first The rescue point to the first At least one type of emergency resource was actually deployed to each leak site.

[0141] It should be noted that this step uses the maximum response time as one of the soft objectives, which can prevent certain leak points from being in the latest arrival state for a long time, thereby improving the ability of the emergency resource scheduling plan to ensure the overall timeliness.

[0142] In one embodiment, such as Figure 6 As shown, the maximum response time convergence curve is analyzed; the horizontal axis represents the number of iterations (dimensionless), and the vertical axis represents the maximum response time (in minutes). Experiments show that the maximum response time decreases continuously with iteration, proving that the algorithm can gradually improve global timeliness while ensuring safety.

[0143] 4> For discrete emergency resource scheduling schemes with a hard constraint violation rate of 0, calculate the non-critical resource shortage rate. , This is used to evaluate the degree to which the demand for resources at non-serious injury leak points or non-life-saving locations is met. In specific implementation, let... If the set represents all resource categories, then the non-critical resource gap rate is... The calculation method is expressed as follows: .

[0144] It should be noted that this calculation can further improve the overall satisfaction of ordinary leak points and ordinary resource categories, on the basis that the hard constraints of life rescue have been met, thus avoiding the optimization process from only satisfying the minimum hard constraints and ignoring the overall quality of disposal.

[0145] In one embodiment, such as Figure 7As shown, the convergence curve of the non-critical resource shortage rate is analyzed; the horizontal axis represents the number of iterations (dimensionless), and the vertical axis represents the non-critical resource shortage rate (dimensionless). The experiment shows that the non-critical resource shortage rate gradually decreases with iteration, indicating that the algorithm can continuously improve the satisfaction of ordinary leakage points and non-life rescue resources under the premise that life rescue guarantee is already met.

[0146] 5. Compare different discrete emergency resource scheduling schemes according to the hard constraint priority principle. If the hard constraint violation degrees of two discrete emergency resource scheduling schemes are different, the discrete emergency resource scheduling scheme with the lower hard constraint violation degree is preferred; if the hard constraint violation degrees of both discrete emergency resource scheduling schemes are 0, then the scheme with the lower hard constraint violation degree is selected based on the total scheduling cost. Maximum response time Non-critical resource gap rate Perform Pareto comparisons.

[0147] In practical implementation, if both Scheme 1 and Scheme 2 satisfy the hard constraints, and Scheme 1 is not inferior to Scheme 2 in all three objectives—total scheduling cost, maximum response time, and non-critical resource shortage rate—and is superior to Scheme 2 in at least one objective, then Scheme 1 dominates Scheme 2. If Scheme 1 has a lower cost but a longer response time, and Scheme 2 has a shorter response time but a higher cost, then both can be retained as non-dominant candidate schemes. If Scheme C has the lowest cost but has insufficient emergency equipment support at the point of serious injury leakage, then Scheme C cannot be prioritized over any scheme with a hard constraint violation degree of 0 because its hard constraint violation degree is greater than 0.

[0148] S305, Leader Wolf Selection and Optimal Emergency Resource Allocation Strategy Generation This step selects the leader wolf based on the hard constraint priority evaluation results, iteratively updates the gray wolf population, and outputs the optimal emergency resource scheduling strategy after the termination condition is met. The specific steps are as follows: 1> After each iteration, based on the hard constraint violation degree Filter the set of scheduling schemes that satisfy hard constraints , represented as ;in, Indicates the first The set of scheduling schemes that satisfy the hard constraints in each iteration.

[0149] In practical implementation, if If not empty, then... Internally based on total scheduling cost Maximum response time Non-critical resource gap rate Perform non-dominated sorting and combine it with crowding selection. Wolf, wolves and Wolf; if If empty, then the hard constraint violation degree applies. Sort the solutions from smallest to largest and select the three candidate solutions with the lowest hard constraint violation as... Wolf, wolves and Wolves, in turn, can prioritize moving in directions that satisfy life-saving and resource coupling constraints when no feasible solution has been found. Crowding degree is used to characterize the sparsity of candidate solutions in the multi-objective space; the higher the crowding degree, the more conducive the candidate solution is to maintaining the diversity of the solution set.

[0150] In one embodiment, as an example, after one iteration, all satisfying... The proposed solutions constitute a feasible solution set. In this set, Solution 1 (cost 100, time 30 minutes, gap 0.1) and Solution 2 (cost 120, time 25 minutes, gap 0.15) are not mutually exclusive and are both retained. Solution 1 has fewer surrounding solutions, thus its crowding is high, while Solution 2 has more surrounding solutions, thus its crowding is low. If we were to choose one of them as... The wolf should choose solution 1, which has a higher crowding density, to maintain the diversity of solutions.

[0151] 2> Utilizing the results obtained through filtering Wolf, wolves and The wolf performs search control factor adjustment, hyperspherical manifold position update, density modulation, and discrete scheduling back projection in S302 and S303 to obtain the discrete emergency resource scheduling scheme for the next iteration, and continues to perform hard constraint priority evaluation.

[0152] In practical implementation, if the maximum number of iterations is reached... If the iteration terminates, then the iteration ends; if the iteration continues... If the set of non-dominated feasible candidates does not improve significantly in the next iteration, the iteration can be terminated early. The value represents the length of the early stop window, which can be 15 to 30. The lack of significant improvement in the non-dominated feasible candidate set means that the optimal maximum response time, optimal total scheduling cost, and optimal non-critical resource gap rate have all failed to reach the preset improvement threshold, for example, the improvement is less than 1%.

[0153] 3> Select the optimal emergency resource scheduling strategy from the final set of non-dominated feasible candidates. , This refers to the discrete emergency resource scheduling scheme that is ultimately output to the emergency command system for execution or review.

[0154] In one implementation, if the emergency command objective leans towards rapidly controlling the accident, then the maximum response time is prioritized. The minimum feasible solution is taken as the optimal emergency resource allocation strategy. If the emergency command objective leans towards resource conservation, then the total dispatch cost should be prioritized. The minimum feasible solution is taken as the optimal emergency resource allocation strategy. If there is no human preference input, the three soft objectives are normalized and then weighted and summed. The feasible solution with the smallest weighted sum is selected. The calculation method is expressed as follows: ; in, This represents the final set of non-dominated feasible candidates; This represents the normalized total scheduling cost; Indicates the normalized maximum response time; This represents the normalized non-critical resource deficit rate; Indicates the weight of cost targets; Indicates the target weight for response time; Indicates the target weight of the gap; the above three weights satisfy... It is acceptable , , .

[0155] 4> Implement the optimal emergency resource allocation strategy Restore to a three-dimensional scheduling table. Each record in the three-dimensional scheduling table is... The rescue point to the first The first leak point was transferred. The number of emergency resources is calculated, and the degree of violation of the hard constraints corresponding to the scheme is output simultaneously. Total scheduling cost Maximum response time Non-critical resource gap rate .

[0156] In one embodiment, such as Figure 8 As shown, the resource allocation table of rescue point 1 in the optimal scheduling scheme is analyzed, and the quantity of various emergency resources allocated by rescue point 1 to each leak point in the final optimized emergency resource scheduling scheme is displayed. The horizontal axis represents the emergency resource category (dimensionless), and the vertical axis represents the leak point number (dimensionless). The color intensity and labeled values ​​in the figure all represent the resource allocation quantity (unit: pieces). The experiment shows that there are significant differences in the allocation quantity of rescue point 1 to different leak points and different resource categories, and some leak points can obtain more critical material allocations.

[0157] S4. Rolling Dispatch Application Process for Sudden Chemical Leakage Accidents After optimizing emergency resource scheduling within a single scheduling time window, this step embeds the method into the rolling scheduling process, enabling the emergency resource scheduling strategy to be continuously updated according to the accident situation, road conditions, and inventory status. The specific steps are as follows:

[0158] 1> In the At the start of each scheduling time window, the continuous concentration field is updated. Resource Demand Matrix Inventory matrix and the expected transportation time matrix The scheduling time window can be set to 5 minutes, 10 minutes, or 15 minutes; for leaks with rapid spread or high toxicity, the scheduling time window can be shortened.

[0159] 2> Recalculate the resource demand density factor according to S201 ;like Then, the fluctuation value of resource demand tightness is calculated simultaneously. This is used for subsequent search control factor adjustment. Based on this, changes in wind speed, leakage source strength, road control, and inventory consumption can be reflected in the scheduling optimization process in a timely manner.

[0160] 3> In the The initial gray wolf population is generated within a scheduling time window. If Then, perform biased initialization according to S202; if Then, the excellent scheduling schemes that have not yet been executed or are still adjustable in the previous scheduling time window are used as part of the initial gray wolves, and several new biased initial gray wolves are generated in addition. This method can inherit the effective search results of the previous time window, while retaining the ability to adapt to new situations.

[0161] In practical implementation, 20% to 50% of the gray wolves in the population can be obtained from the non-dominated feasible candidate schemes of the previous time window, while the remaining gray wolves are regenerated according to the current resource demand density factor. If the accident situation changes significantly, i.e. If the threshold is exceeded (e.g., 0.2 to 0.3), the inheritance ratio of the previous time window scheme can be reduced, and the proportion of newly generated gray wolves can be increased.

[0162] 4> Execute steps S301 to S305 to obtain the optimal emergency resource scheduling strategy for the current scheduling time window. , It refers to the first The optimal discrete emergency resource scheduling scheme is output within each scheduling time window.

[0163] 5> Based on the optimal emergency resource allocation strategy Generate execution instructions, which include the rescue point number, leak point number, resource type, resource quantity, estimated departure time, and estimated arrival time. Irrevocable scheduling instructions that have already been issued are recorded as fixed resource occupancy in the next scheduling time window and deducted from the inventory matrix; scheduling instructions that have not yet been executed or can be adjusted can be re-optimized in the next scheduling time window.

[0164] It should be noted that the rolling scheduling application process does not start from scratch and optimize each time. Instead, it inherits existing feasible solutions based on updated accident situation information and adjusts the search intensity using fluctuations in resource demand density. Based on this, it can balance the real-time nature of emergency response with the stability of the solution, avoiding on-site execution chaos caused by frequent and drastic changes in scheduling instructions.

[0165] In one embodiment, for example, if a chemical plant experiences a chlorine leak, the scheduling time window is set to 15 minutes. In the first time window ( Within the region, leak points P1, P2, and P3 were detected. An initial population was generated and optimized, and a solution was output. The instructions were to urgently transfer neutralizing agents and chemical protective suits from rescue point M1 to P1, and to transfer first-aid equipment from M2 to P2, and to execute the orders. In the second time window ( Within the area, information was updated: wind direction changed, P3 concentration increased sharply, and M1's chemical protective suit inventory decreased due to the previous round of shipments; calculations were performed. If the value exceeds the threshold of 0.25, it indicates a drastic change in the situation; the algorithm will inherit 30% of the unexecuted adjustment schemes from the previous round into the new population, and the remaining 70% will be adjusted according to the new density factor. Reinitialization; Optimized output scheme The command redirected more resources to P3 and adjusted the resource allocation path of M2.

[0166] It should be noted that, through the above-mentioned acquisition of accident situation data, calculation of resource demand density factor, biased initialization of gray wolf population, discrete scheduling back projection, fitness entropy fluctuation adjustment, continuous density modulation position update, hard constraint priority two-level evaluation, and rolling scheduling output, this invention can generate an emergency resource scheduling scheme that meets the principles of life rescue priority, inventory constraints, and resource coupling constraints in the event of a sudden chemical leak accident, and maintain high scheduling feasibility, response efficiency, and resource matching stability when the accident situation changes dynamically.

[0167] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for generating emergency resource scheduling strategies to respond to sudden leakage accidents, characterized in that, Includes the following steps: S1. Obtain situational data including at least the leak point, rescue point, resource type, resource requirement, inventory, expected transportation time moment, and continuous concentration field, and construct a data structure to characterize the scheduling scheme; S2. Calculate the resource demand density factor for each leakage point based on the situation data, and inject the density factor into the resource allocation ratio and hyperspherical manifold position of the initial gray wolf population to generate an initial candidate scheduling scheme that meets the basic inventory constraints, making the scheme biased towards high-risk, high-demand and high-gap leakage points. S3. During the iteration of the Grey Wolf algorithm, the individual position update strategy is adjusted according to the dynamic fluctuation of the density factor, and the hard inventory constraint is satisfied first to generate the optimal scheduling strategy for the current iteration. S4. Apply the scheduling strategy obtained through iterative optimization to the rolling time-domain window, and repeat S1 to S3 as the accident situation data is updated to output the real-time updated emergency resource scheduling strategy.

2. The method for generating an emergency resource scheduling strategy to respond to a sudden leakage accident according to claim 1, characterized in that, S2 specifically refers to: S21. Taking the continuous concentration field, resource demand matrix, inventory matrix and expected transportation time matrix as input, output the resource demand density factor corresponding to each leakage point; S22. Using the resource demand density factor, resource demand matrix, and inventory matrix as inputs, generate an initial discrete emergency resource scheduling scheme that satisfies the basic constraints of the rescue point inventory, and map it to the location of a hyperspherical manifold. S23. Convert the hyperspherical manifold position into a discrete emergency resource scheduling scheme in non-negative integer form, and correct the upper limit of inventory, upper limit of demand, minimum guarantee of key materials and resource coupling constraints during the conversion process.

3. The method for generating an emergency resource scheduling strategy to respond to a sudden leakage accident according to claim 1, characterized in that, S3 specifically refers to: S31. Statistically analyze the fitness distribution of discrete emergency resource scheduling schemes in the current gray wolf population, calculate the normalized fitness distribution entropy, and use it to determine whether the gray wolf population is in a state of global exploration, local development, or overcrowding. S32. Based on the historical fluctuations of the normalized fitness distribution entropy and resource demand density factor, the search control factor of the gray wolf optimization algorithm is adaptively adjusted so that the search process maintains a strong exploration capability when the accident situation is unstable and accelerates convergence when the situation is stable. S33. Update the position of the hyperspherical manifold based on the gray wolf pack hierarchical coordination mechanism, and continuously modulate it using the resource demand density factor after the position update, so that the high-priority leakage point continuously obtains resource allocation tendency during the iterative search process. S34. First, evaluate the hard constraint violation degree of discrete emergency resource scheduling schemes, and then evaluate the total scheduling cost, maximum response time and non-critical resource gap of the scheduling schemes that meet the hard constraints. S35. Select the leader wolf based on the hard constraint priority evaluation results, iteratively update the gray wolf population, and output the optimal emergency resource scheduling strategy after the termination condition is met.

4. The method for generating an emergency resource scheduling strategy to deal with a sudden leakage accident according to claim 2, characterized in that, S21 specifically refers to: Reading the first in a continuous concentration field The local concentration values ​​at the spatial coordinates of each leak point are used to obtain the concentration values ​​at the leak points. With the first Using the spatial coordinates of each leak point as the center, the gradient amplitude of the concentration field is calculated in a continuous concentration field to obtain the intensity of the leak situation change. According to the The resource requirement vectors of all leak points within the scheduling time window are used to calculate the i-th... Deviation in resource requirements at each leak point; Based on the supplies at the relief point and the estimated transport time, calculate the first... The achievable inventory gap at each leakage point; Normalize the concentration value at the leakage point, the intensity of the change in leakage status, the deviation of resource demand, and the available inventory gap, respectively, to obtain the concentration risk weight, diffusion change weight, demand deviation weight, and inventory gap weight. By integrating the concentration risk weight, diffusion change weight, demand deviation weight, and inventory gap weight, we obtain the first... Resource demand density factor for each leakage point.

5. The method for generating an emergency resource scheduling strategy to respond to a sudden leakage accident according to claim 2, characterized in that, S22 specifically refers to: Regarding the first The gray wolf generates an initial discrete emergency resource scheduling scheme; For each rescue point and each resource category, construct the probability of leak point allocation based on the resource demand density factor and the demand quantity; Based on the quantity of emergency resources of type I at the i-th rescue point and the probability of allocation of leak points, generate the initial quantity of resources corresponding to each leak point; Perform demand ceiling correction on the initial discrete emergency resource scheduling scheme to avoid allocating resources that significantly exceed the demand to a leak point during the initialization phase; Construct an extended density weight vector based on the resource demand density factor; The initial discrete emergency resource scheduling scheme is transformed into the first... The initial hyperspherical manifold position of the gray wolf; Repeat the above steps until the generated size is [size missing]. The initial gray wolf population.

6. The method for generating an emergency resource scheduling strategy to respond to a sudden leakage accident according to claim 2, characterized in that, S23 specifically refers to: Construct an inverse scaling vector based on the inventory quantity at the rescue point and the required quantity at the leak point; The first The position of the gray wolf's hyperspherical manifold is multiplied element-wise with the inverse scaling vector to obtain a continuous resource quantity vector. Integerization is performed on the continuous resource quantity vector to obtain the initial discrete resource quantity; The initial discrete resource quantity is projected onto the inventory limit to ensure that the total amount of each type of emergency resource mobilized from each rescue point does not exceed the corresponding inventory quantity. Apply a minimum guarantee projection for critical materials to the initial discrete resource quantity to ensure a minimum satisfaction rate of life-saving materials for severely injured leak points; Perform resource coupling constraint projection on the initial discrete resource quantity to ensure that the basic operation ratio relationship is met among the resources that need to be used in combination. The resource quantity after completing the projection of the upper limit of inventory, the minimum guarantee of key materials, and the resource coupling constraint is expanded into a discrete emergency resource scheduling scheme.

7. The method for generating an emergency resource scheduling strategy to respond to a sudden leakage accident according to claim 3, characterized in that, S31 specifically refers to: In the In the nth iteration, for the th Calculate the temporary fitness value for the discrete emergency resource scheduling scheme corresponding to the gray wolf; Based on the size of the gray wolf population Determine the number of fitness partitions ; Count the number of gray wolves in each fitness interval to obtain the number of intervals; fitness distribution entropy Divide by the maximum possible entropy to obtain the normalized fitness distribution entropy.

8. The method for generating an emergency resource scheduling strategy to respond to a sudden leakage accident according to claim 3, characterized in that, S32 specifically refers to: Record recent The normalized fitness distribution entropy of each iteration forms a local entropy sequence; Calculate entropy fluctuation values ​​based on local entropy sequences; Based on the The resource demand density factor between the current scheduling time window and the previous scheduling time window is used to calculate the resource demand density factor of the current scheduling time window. The fluctuation value of resource demand density of each scheduling time window relative to the previous scheduling time window; The search retention index is calculated based on the current normalized fitness distribution entropy, entropy fluctuation value, and resource demand tightness fluctuation value. The search control factor is calculated based on the search retention index.

9. The method for generating an emergency resource scheduling strategy to deal with a sudden leakage accident according to claim 3, characterized in that, S33 specifically refers to: Screening based on the current evaluation results of the gray wolf population Wolf, wolves and Wolf; The wolf represents the leader wolf, which has the best overall performance in the current population. The wolf represents the alpha wolf, which has the second-best overall performance in the current population. The wolf represents the leader wolf, which has the third best overall performance in the current population; For the The first gray wolf in In the next iteration, the hyperspherical manifold position is updated using the grey wolf position to obtain candidate continuous positions; Construct a density modulation vector based on the resource demand density factor; The candidate continuous positions are multiplied element-wise by the compactness modulation vector and then reprojected onto the unit hypersphere to obtain the updated hyperspherical manifold positions. The updated hyperspherical manifold configuration is input into the discrete scheduling back projection process to obtain the first... The first gray wolf in Discrete emergency resource scheduling scheme in the next iteration.

10. The method for generating an emergency resource scheduling strategy to deal with a sudden leakage accident according to claim 3, characterized in that, S34 specifically refers to: For the discrete emergency resource scheduling scheme to be evaluated Calculate the degree of hard constraint violation; For a discrete emergency resource scheduling scheme with a hard constraint violation degree of 0, calculate the total scheduling cost; For a discrete emergency resource scheduling scheme with a hard constraint violation degree of 0, calculate the maximum response time; For a discrete emergency resource scheduling scheme with a hard constraint violation rate of 0, calculate the non-critical resource shortage rate; S35 specifically refers to: After each iteration, a set of scheduling schemes that satisfy the hard constraints is selected based on the degree of hard constraint violation. obtained by filtering Wolf, wolves and The wolf performs search control factor adjustment, hyperspherical manifold position update, density modulation, and discrete scheduling back projection to obtain the discrete emergency resource scheduling scheme for the next iteration, and continues to perform hard constraint priority evaluation; Select the optimal emergency resource scheduling strategy from the final set of non-dominated feasible candidates; The optimal emergency resource scheduling strategy is restored to a three-dimensional scheduling table.

Images