A method for optimizing site selection of nuclear emergency reception points based on radiation risk perception

Abstract

The application discloses a kind of nuclear emergency pick-up point location optimization method based on radiation risk perception, belong to nuclear emergency management and intelligent optimization decision-making technical field.Construct multi-objective location model containing evacuation efficiency, radiation exposure risk, load balancing, combine Gaussian plume model to quantify dynamic radiation risk;Adopt the hybrid initialization strategy of elite and random combination to generate initial population, build cross-generation normalization dynamic fitness evaluation mechanism, through the improved hybrid genetic algorithm of adaptive selection, crossover and radiation gradient guided mutation iteration optimization, finally output optimal location scheme.The application solves the problems that traditional location method ignores dynamic radiation risk, algorithm converges slowly and optimization stability is poor, can greatly reduce the radiation exposure dose of personnel while ensuring evacuation efficiency, and has strong solution stability and engineering practicability.

Description

Technical Field

[0001] This invention belongs to the field of nuclear emergency management and intelligent optimization decision-making technology, specifically relating to a nuclear emergency pick-up point site selection optimization method based on radiation risk perception, and in particular to a pick-up point site selection method that integrates radiation risk modeling, multi-objective optimization and improved genetic algorithm. Background Technology

[0002] In the event of a severe accident at a nuclear power plant, a large amount of radioactive material may be released, forming a spatially and temporally variable radiation field due to atmospheric diffusion. In such a scenario, a large number of residents need to complete an emergency evacuation within a limited time. As a crucial transit point connecting residential areas with refuge facilities, the spatial layout of pick-up and drop-off points directly affects the efficiency of evacuation route organization, the level of radiation exposure for individuals, and the efficiency of emergency transportation resource allocation.

[0003] Currently, the problem of selecting pick-up and drop-off points is usually solved using p-median models, p-center models (Daskin M S. Location, dispatching, and routing models for emergency services with stochastic travel times [J]. Transportation Science, 1983, 17(1): 58-72), and multi-objective optimization methods based on genetic algorithms (Li Y, Cai W, Kana A A. Design of level of service on facilities for crowd evacuation using genetic algorithm optimization [J]. Safety Science, 2019, 120: 237-247), particle swarm optimization algorithms, and NSGA-II (Ruan F, Chen C, He C, et al. Optimization method of public decontamination location and allocation problem in off-site nuclear emergency based improved NSGA-II [J]. Journal of Hazardous Materials, 2025, 489: 137572). However, the existing methods mentioned above still have the following problems in nuclear accident emergency evacuation scenarios: (1) They mainly use distance or time as optimization objectives, which is difficult to reflect the radiation risk under nuclear accident conditions; (2) They lack explicit consideration of dynamic radiation fields; (3) The initial population of traditional genetic algorithms is highly random and the convergence speed is slow; (4) Fixed weights or simple normalization methods are often used in the multi-objective optimization process, which are easily affected by differences in target scale; (5) Existing mutation and search operators lack the utilization of radiation spatial distribution characteristics and cannot effectively guide the search process to converge to low-risk areas. Therefore, a nuclear emergency pick-up point location optimization method is needed that can simultaneously consider the dynamic radiation risk evolution characteristics and improve search efficiency and solution stability. Summary of the Invention

[0004] The purpose of this invention is to address the technical problems of existing site selection methods, such as ignoring dynamic radiation risks, unstable multi-objective evaluation, slow algorithm convergence, low solution quality, and inability to adapt to nuclear accident emergency scenarios. This invention provides a nuclear emergency pick-up point site selection optimization method based on radiation risk perception. By constructing a multi-objective site selection model under radiation risk constraints and designing an improved hybrid genetic algorithm that integrates cross-generational normalization and radiation gradient guidance mechanisms, the layout scheme of pick-up points under nuclear accident conditions can be optimized.

[0005] To achieve the above-mentioned objectives, the present invention provides the following technical solution:

[0006] S1: Establish a nuclear emergency pick-up and drop-off point location model to determine the candidate pick-up and drop-off point set, the residential area set, and related constraints;

[0007] S11. Determining the basic data for the model: Delineate the research area for the emergency impact of the nuclear accident, collect regional geographic information, road network information, population distribution information and candidate pick-up and drop-off point layout information, and clarify the scope and parameters of the set of settlements and the set of candidate pick-up and drop-off points within the research area;

[0008] S12. Variable definition: Define binary site selection decision variables and residential point and pick-up point matching and allocation variables to characterize the selection status of pick-up points and the corresponding relationship of resident evacuation allocation.

[0009] S13. Constraint Setting: Set constraints on the number of pick-up and drop-off points, the unique allocation of residential areas, and the effective point-to-point matching to ensure that the site selection and personnel allocation plan is compliant and feasible, and complete the overall construction of the nuclear emergency pick-up and drop-off point site selection model.

[0010] S2: Establish a multi-objective optimization model with evacuation efficiency, radiation exposure risk and load balancing at pick-up and drop-off points as objectives;

[0011] S21. Construction of evacuation efficiency target: Taking the minimization of the average evacuation distance of residents as the optimization target, the overall travel cost of residents to the pick-up and drop-off point is quantified to characterize the emergency evacuation and assembly efficiency.

[0012] S22. Radiation exposure risk target construction: Taking the minimization of the cumulative radiation risk of candidate pick-up and drop-off points and corresponding service populations as the optimization objective, the overall radiation exposure hazards during the residents' assembly process are quantified to ensure evacuation safety.

[0013] S23. Load balancing target construction: The optimization objective is to minimize the load difference of the population served by each selected pick-up and drop-off point, so as to achieve balanced allocation of emergency service resources and avoid single-point overload and congestion.

[0014] S24. Multi-objective coupling modeling: The above three types of objectives are coupled to construct a collaborative optimization mathematical model to achieve synchronous optimization of three-dimensional indicators: evacuation efficiency, radiation safety, and resource balance.

[0015] S3: Establish a radiation risk assessment model and calculate the radiation risk indicators of candidate pick-up and drop-off points;

[0016] S31. Radiation field simulation: Based on the Gaussian plume model, combined with nuclear accident source term parameters and environmental meteorological parameters, the spatiotemporal atmospheric diffusion process of radioactive materials in a nuclear accident is simulated to obtain the spatiotemporal distribution characteristics of dynamic radiation dose rate in the study area.

[0017] S32. Radiation data extraction: Extract discrete instantaneous radiation dose rate data for each candidate pick-up and drop-off point during the entire evacuation period;

[0018] S33. Risk index calculation: The trapezoidal integral algorithm is used to perform integral approximation calculation on the discrete dose rate data to solve the cumulative radiation dose of each candidate reception point. This is used as a standardized radiation risk assessment index to complete the risk quantification assessment of all candidate points.

[0019] S4: A hybrid initialization strategy is adopted to generate the initial population. An elite population and a random population are constructed by combining radiation safety and service capability to improve the quality and diversity of the initial solution.

[0020] S41. Dual-dimensional evaluation of candidate pick-up and drop-off points: All candidate pick-up and drop-off points are quantitatively scored from two dimensions: radiation safety and service coverage capability, to obtain normalized radiation safety factor and normalized service capability index.

[0021] S42. Elite Pool Construction: Integrate radiation safety and service capability indicators to construct a comprehensive evaluation index, and select high-quality candidate pick-up and drop-off points based on the comprehensive score to form an elite candidate pool.

[0022] S43. Hybrid population generation: Elite individuals are drawn from the elite candidate pool and ordinary individuals are randomly generated at the same time. The elite population and the random population together form a complete initial population, taking into account both the quality of the initial solution and the diversity of the population.

[0023] S5: Construct a multi-objective dynamic fitness evaluation mechanism to calculate the fitness of individual solutions;

[0024] S51. Target data preprocessing: Logarithmic transformation is performed on radiation risk targets to reduce the evaluation bias caused by differences in the numerical scale of different targets.

[0025] S52, Cross-generational dynamic normalization: The normalization boundary of each objective is dynamically updated by using cross-generational quantiles combined with exponential moving average, and numerical truncation is used to avoid extreme value interference.

[0026] S53. Dynamic weight adaptive update: Based on the target distribution characteristics of each generation of the population, the weights of multiple targets are iteratively updated, a lower limit constraint on the weights is set and normalization is completed to avoid weight imbalance.

[0027] S54. Comprehensive Fitness Calculation: Based on dynamically updated target weights, the comprehensive fitness value of each individual in the population is calculated using a weighted average method, thus completing an accurate evaluation of the individual's merits.

[0028] S6: Perform genetic evolution operations based on fitness evaluation results, including selection, crossover, and radiation gradient-guided mutation.

[0029] S61. Elite Selection and Local Perturbation: A tournament selection mechanism is used to select high-quality parent populations, and compliant local perturbation optimization is performed on elite individuals on a regular basis to retain better elite solutions.

[0030] S62. Adaptive crossover operation: Based on the relative fitness of individuals, adaptively match the crossover probability, and complete the generation of offspring individuals through a two-point crossover method. The evolutionary intensity is dynamically adjusted according to the quality of the individuals.

[0031] S63. Radiation gradient-guided mutation: Based on the spatial gradient characteristics of the regional radiation field, calculate the radiation sensitivity factor of individuals, correct the basic mutation probability, guide the population to adaptively converge to low radiation risk areas, and improve the radiation safety of the solution.

[0032] S7: Repeat steps S5 and S6 until the termination condition is met, and output the optimal pick-up and drop-off point location scheme.

[0033] S71. Iterative Loop: Continuously loops the dynamic fitness evaluation and genetic evolution operations, iteratively updating the population individuals and the optimal address solution.

[0034] S72. Termination Judgment: The iteration process is judged in real time. When the preset maximum number of iterations is reached, or the fitness of the best individual has not been updated for several consecutive generations, the iteration optimization process is terminated.

[0035] S73. Scheme Output: Select the individual with the best overall fitness from the final converged population, parse the corresponding coding information, and output the optimal spatial layout and site selection scheme for the nuclear emergency pick-up and drop-off point.

[0036] The technical problem solved by this invention:

[0037] This invention addresses the problems existing in the selection of nuclear emergency pick-up and drop-off points, such as insufficient radiation risk modeling, unstable multi-objective evaluation mechanism, low initial population quality, and lack of radiation spatial information guidance in the genetic search process. It proposes a nuclear emergency pick-up and drop-off point site selection optimization method based on radiation risk perception to improve the safety, optimization efficiency, and solution stability of pick-up and drop-off point layout schemes under complex nuclear accident scenarios.

[0038] Compared with the prior art, the advantages of the present invention are as follows:

[0039] (1) This invention establishes a multi-objective optimization model for nuclear emergency evacuation points that takes into account radiation risks. It integrates evacuation time, cumulative radiation exposure risk and load balancing of evacuation points into the same optimization framework, thereby achieving coordinated optimization of three key indicators—efficiency, safety and balance—during nuclear accident emergency evacuation, and thus obtaining a evacuation point layout scheme that better meets actual needs.

[0040] (2) The present invention adopts a hybrid initialization strategy that integrates radiation risk and service capability. By constructing an initial structure that combines elite population and random population, the quality of the initial solution is improved while ensuring the diversity of the solution space, thereby enhancing the global search capability of the genetic algorithm and accelerating the convergence speed.

[0041] (3) This invention proposes a dynamic fitness evaluation mechanism based on cross-generational quantile normalization and exponential moving average, which effectively reduces the impact of differences in the dimensions of different objective functions and abnormal individuals on the evaluation results, improves the stability and convergence reliability of the multi-objective optimization process, and avoids the fluctuation problem that is easily generated by traditional normalization methods.

[0042] (4) The present invention designs a mutation guidance mechanism based on radiation gradient information. While maintaining the random search capability of the genetic algorithm, it introduces radiation field spatial gradient information, so that the population evolution process can adaptively adjust to the low radiation risk area, thereby improving the radiation safety of the solution.

[0043] (5) Experimental results show that, compared with the fixed-weight single-objective genetic algorithm, the intra-generation normalized weighted genetic algorithm and the NSGA-II algorithm, the present invention can further reduce the average radiation dose of residents while ensuring evacuation efficiency, and has good optimization stability and engineering application value. Attached Figure Description

[0044] Figure 1 This is a flowchart of the overall process for optimizing the location of nuclear emergency transport points based on radiation risk perception, as described in this invention.

[0045] Figure 2 This is a flowchart of the optimization process of the improved hybrid genetic algorithm (IHGA) described in this invention.

[0046] Figure 3 This is a schematic diagram of the multi-objective dynamic fitness evaluation mechanism described in this invention.

[0047] Figure 4 This is a structural diagram of the genetic operator described in this invention.

[0048] Figure 5 This is a schematic diagram of the location selection and evacuation road network for nuclear emergency pick-up and drop-off points in an embodiment of the present invention.

[0049] Figure 6This is a schematic diagram showing the performance comparison results of different algorithms under the evacuation time and radiation dose indicators in the embodiments of the present invention. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of this invention clearer, the following embodiments will be used in conjunction with the accompanying drawings to further illustrate the invention. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0051] Example 1:

[0052] This invention provides a method for optimizing the location of nuclear emergency evacuation points based on radiation risk perception, used for optimizing the layout of evacuation points in nuclear accident emergency evacuation scenarios. This method comprehensively considers factors such as population distribution, road network structure, and radiation risk distribution. It constructs a multi-objective optimization model and combines it with an improved hybrid genetic algorithm guided by radiation gradients to solve for the optimal evacuation point layout scheme. Figure 1 The diagram shown is an overall flowchart of the nuclear emergency pick-up point site selection optimization method based on radiation risk perception according to an embodiment of the present invention. The method includes steps S1-S7:

[0053] S1: Establish a nuclear emergency pick-up and drop-off point location model.

[0054] Based on the distribution of villages, population size, and locations of candidate pick-up and drop-off points within the affected area of ​​a nuclear accident, a nuclear emergency pick-up and drop-off point site selection model is established.

[0055] Define location decision variables ,satisfy ,in This represents the set of candidate pick-up and drop-off points. This indicates the number of pick-up and drop-off points to select; For the decision variables of pick-up and drop-off point location, =1 indicates that a pick-up / drop-off point has been selected. , =0 means not selected. This variable performs a global constraint on whether or not pick-up and drop-off points are selected, ensuring that the total number of selected pick-up and drop-off points remains fixed. .

[0056] Define allocation variables Each village is assigned to only one selected pick-up and drop-off point, thus satisfying the requirement. and satisfy This means that villages are only allowed to be assigned to selected pick-up and drop-off points, among which... Indicates a collection of villages; This represents "for all". The two constraints respectively ensure that the population allocation is without omissions and without multiple splits, while preventing the allocation of population to unselected locations, thus ensuring the physical feasibility of the allocation plan.

[0057] S2: Establish a multi-objective optimization model.

[0058] To simultaneously consider evacuation efficiency, radiation safety, and balanced service at pick-up and drop-off points, a multi-objective site selection model is established with average evacuation distance, cumulative radiation risk, and load balancing indicators at pick-up and drop-off points as optimization objectives.

[0059] The average evacuation distance target is defined as:

[0060]

[0061] in, Indicates the total number of villages; Refers to a single village; It is the set of all candidate pick-up and drop-off points. Refers to a single candidate pick-up / drop-off point; Indicates village and pick-up / drop-off point The distance between them Assign variables to villages and pick-up / drop-off points; this objective is used to characterize the average travel distance of residents to the pick-up / drop-off points, and the lower the value, the higher the overall evacuation and traffic efficiency.

[0062] The cumulative radiation risk target is defined as:

[0063]

[0064] in, Indicates pick-up and drop-off point Radiation dose rate at the location Indicates allocation to pick-up / drop-off point The total number of people; this target is used to measure the overall radiation exposure risk faced by residents during the assembly phase, and the smaller the value, the lower the total radiation damage to the entire population.

[0065] The load balancing target for pick-up and drop-off points is defined as follows:

[0066]

[0067] in, This indicates the number of pick-up and drop-off points to select. This represents the average population served by a single pick-up / drop-off point. Indicates allocation to pick-up / drop-off point Total number of people This indicates the selection of a set of pick-up and drop-off points. Indicates village The target is used to measure the load balancing between pick-up and drop-off points. The closer the indicator is to 0, the smaller the difference in the population that each pick-up and drop-off point can accommodate, and there will be no congestion or overload at the points.

[0068] The above three objectives together constitute a multi-objective optimization model for the location of nuclear emergency evacuation points.

[0069] S3: Establish a radiation risk assessment model.

[0070] In this embodiment, a Gaussian plume model is used to simulate the diffusion process of radioactive materials after a nuclear accident, in order to obtain the radiation dose rate distribution within a spatial region. The Gaussian plume model can reproduce the spatiotemporal diffusion patterns of radioactive pollutants with wind direction and meteorological parameters, and can obtain the radiation dose rate of any point in the entire region that changes dynamically over time. This differs from static single-point radiation values ​​and closely matches the real characteristics of the continuous evolution of radiation in a nuclear accident.

[0071] Let the location of the candidate pick-up point be... In discrete time series The instantaneous dose rate at that location was calculated. To characterize the cumulative radiation risk over the entire exposure period, the dose rate over discrete time is approximated by integration, and the cumulative radiation dose is obtained using the trapezoidal integral formula. It serves as a radiation risk assessment indicator for pick-up and drop-off points, and is used in the subsequent site selection optimization process. This represents the total number of discrete time points. Candidate pick-up and drop-off points At any moment The instantaneous radiation dose rate; Candidate pick-up and drop-off points The cumulative radiation dose over the entire time period calculated using trapezoidal integral; The time interval is defined as the interval between two adjacent discrete time points. The trapezoidal integral is used to adapt the time-series radiation curve numerically, accurately summarizing the total radiation exposure over the entire evacuation period.

[0072] S4: Generate the initial population using a hybrid initialization strategy.

[0073] To balance the quality of initial solutions with population diversity, a hybrid initialization strategy combining elite initialization and random initialization is adopted to generate the initial population. Traditional genetic algorithms initialize completely randomly, resulting in a very low proportion of high-quality initial solutions and slow convergence. This scheme combines elite and random individuals, relying on elite points to provide high-quality starting solutions to accelerate convergence, while relying on random individuals to retain global search diversity and avoid premature convergence.

[0074] First, the radiation safety of the candidate pick-up and drop-off points is evaluated. Let the candidate pick-up and drop-off points be... The dose rate at the site Then its normalized radiation safety factor is ; For a single candidate pick-up / drop-off point, This is a collection of all candidate pick-up and drop-off points; The maximum radiation dose rate among all candidate pick-up and drop-off points. The minimum radiation dose rate among all candidate pick-up and drop-off points; Candidate pick-up and drop-off points The normalized radiation safety factor ranges from 0 to 1. The lower the radiation level at a location, the higher the value of this factor, and the better the safety.

[0075] Then, the service capabilities of the candidate pick-up and drop-off points are evaluated. Let the candidate pick-up and drop-off points be... The population within the service area is ,in, Indicates village population, It meets the requirements of the candidate pick-up and drop-off points. The set of all villages within a distance of the maximum service radius. In the formula This includes all villages in the entire region. It is a village with candidate pick-up points Spatial distance, To constrain the maximum service radius, the service population is normalized. This represents a normalized service capability indicator. Any candidate pick-up / drop-off point The population covered; , These represent the maximum and minimum population coverage of all candidate pick-up and drop-off points, respectively; normalization maps the population values ​​to 0~1, which can be used for weighted calculation with radiation safety indicators to eliminate dimensional differences.

[0076] Further construct comprehensive evaluation indicators ,in These are the weighting coefficients. Candidate pick-up and drop-off points Overall evaluation score; This refers to the normalized radiation safety factor obtained in the previous section. Candidate pick-up and drop-off points The normalized service capability indicators. Weight Priority can be adjusted as needed. Larger sites place greater emphasis on radiation safety. Smaller locations prioritize population carrying capacity to accommodate different nuclear emergency control needs. Candidate pick-up and drop-off points are ranked in descending order based on comprehensive scores, and the highest-scoring point is selected. A pool of candidate pick-up and drop-off points is formed. Points are randomly selected from this elite pool. Each node generates an elite individual.

[0077] Let the population size be The initial population consists of an elite population and a random population. ,in This represents the first complete initial population; For the elite population, For random populations, Include An elite individual, This is the elite proportion parameter, used to control the proportion of elite solutions in the initial population. This is the floor operator, used to round a real number down to the nearest integer. Includes the remainder A randomly generated individual, The number of individuals is random. After the initial population is built, the genetic optimization phase begins. The improved hybrid genetic algorithm process is as follows: Figure 2 As shown, the algorithm's operating logic is as follows: First, the initial population generated by the hybrid initialization strategy is input into the multi-objective dynamic fitness evaluation module to calculate the comprehensive fitness of all individuals in the population. Then, the algorithm enters the judgment node to check whether the iteration process meets the preset termination condition. If it does, the optimal pick-up point selection scheme is directly output and the algorithm ends. If the termination condition is not met, the selection operation, adaptive crossover, and radial gradient-guided mutation are executed in sequence. After all operators are completed, a new generation of population is generated. The new generation of population returns to the multi-objective dynamic fitness evaluation module along the return arrow and iterates until the termination condition is triggered.

[0078] S5: Based on the population individuals generated in step S4 and combined with the multi-objective optimization model established in step S2, calculate the fitness value of each individual. For example... Figure 3 As shown, this invention constructs a multi-objective dynamic fitness evaluation mechanism. First, it inputs three types of raw target data: average evacuation distance target T, cumulative radiation risk target D, and load balancing target B. Second, it performs logarithmic transformation preprocessing on radiation target D. Next, it calculates the upper and lower bounds of the quantiles for each target value and updates the normalized boundary using exponential moving average. Then, it calculates the normalized target value based on the updated boundary. Next, it performs dynamic updating of the target weights. Finally, it calculates the weighted solution to obtain the individual's comprehensive fitness, completing the quantitative evaluation of the dynamic fitness of the population. Specifically:

[0079] S5.1: Radiation risk target transformation, for cumulative radiation risk targets Perform logarithmic transformation in This represents the scaling parameter; the logarithmic transformation compresses the extremely large numerical range of radiation risk, preventing the radiation target from being too large and completely crowding out the evaluation weights of the other two targets. Used for fine-tuning the compression amplitude.

[0080] S5.2: Cross-generational normalization of the objective function, for the objective function Calculate the first The lower quantile of the target value for the next generation population and upper quantile The normalized boundary is updated using exponential moving average. , ,in, The target is the average evacuation distance. This is the target value for radiation risk after logarithmic transformation. The goal is to achieve load balancing at pick-up and drop-off points. The index of the objective function takes the value of one of the three; For the first The target obtained by updating the moving average of the economic index Normalize the lower bound. For the first The target obtained by updating the moving average of the economic index Normalization of the upper boundary, It represents the smoothing coefficient; the cross-generational exponential moving average replaces the single-generation boundary, weakens the disturbance of single-generation outliers on normalization, allows the boundary to change smoothly with iteration, and improves the stability of the evaluation.

[0081] Calculate the normalized objective value based on the updated boundary. ,in Indicates the current individual, Represents an individual In the objective function The value of , when hour ,when hour ,when hour .

[0082] To prevent the normalization results from exceeding the preset range, the normalization results are truncated. ; For the first Individual Corresponding target The normalized target value; , These are operations to take the maximum and minimum values, respectively, used to force the normalization result to be limited to the range of 0 to 1, thus eliminating interference from extreme abnormal values.

[0083] S5.3: Target weights are dynamically updated, calculating the first... Objective function in generation population normalized median Describe the objective function The median evaluation index, among which, For the first target The median of normalized values. The calculation is performed to find the median; relying on the median to resist outliers, it is less likely to be biased by individual extreme values ​​compared to the mean.

[0084] Calculate temporary weights based on the median. ,in, This represents the sensitivity coefficient for weight adjustment. For the target index variable, .

[0085] Use exponential moving average to update weights ,in This indicates the adaptive adjustment rate of the weights. For the first Substitute the objective function weights, For the first Substitute objective function Historical weight, The rate is adjusted adaptively for weights. Historical weights are used to smooth out temporary weights, avoiding drastic weight oscillations between iterations. Control the weighting ratio between new and old.

[0086] Set a lower bound constraint on the weights. After the weights are updated, normalization is performed to ensure that they meet the requirements. ,in This is the minimum weight threshold. The lower bound constraint ensures that the weights of the three objectives will not reach zero, ensuring that all three types of optimization objectives participate in the evaluation throughout the process, and preventing any objective from becoming completely ineffective.

[0087] S5.4: Overall fitness calculation, calculating the individual's overall fitness based on the updated target weights:

[0088] ;

[0089] in, For the current individual within the population, Refers to the first Individual Overall adaptability; , , The order is number 1 The weights corresponding to the average evacuation distance target, the cumulative radiation risk target after logarithmic transformation, and the load balancing target at pick-up and drop-off points are determined. , , The order is number 1 Individual The normalized objective values ​​of the three objectives are used. A weighted summation is performed to achieve multi-objective fusion; the lower the fitness value, the better the overall performance of the location selection scheme.

[0090] S6: Perform genetic evolution operations based on the fitness values ​​obtained in step S5. For example... Figure 4 As shown, this invention employs a genetic evolution mechanism that combines selection, crossover, and radiation gradient-guided mutation. The genetic operators are executed sequentially from top to bottom: taking the parent population as input, tournament selection is first carried out, and then elite local perturbation is performed on elite individuals; then, adaptive crossover probability and basic mutation probability are calculated based on the relative fitness of individuals; then, radiation gradient-guided mutation is completed by combining the radiation field spatial gradient, and mutation probability is corrected by relying on radiation sensitivity factors; finally, offspring population is generated, completing a single complete genetic evolution process.

[0091] S6.1: Selection operation, using tournament selection to construct the parent population. Every interval... During this period, local perturbation operations are performed on elite individuals with high fitness rankings.

[0092] For elite individuals, a gene locus corresponding to a pre-selected pick-up / drop-off point is randomly chosen. A gene locus corresponding to an unselected pick-up point Perform a swap operation, with a perturbation count of . ,in Indicates the percentage of disturbance. This indicates the number of pick-up and drop-off points to select.

[0093] Since exchange-based perturbation operations may violate coverage constraints, the algorithm will initiate a feasibility repair process, attempting at most [number] times. The process involves several repairs. If the feasibility of a solution cannot be restored within the specified number of attempts, the perturbed solution is discarded, the original elite individuals are retained, and only compliant and better perturbed individuals are kept to ensure the stability of the elite solution. A perturbed individual is only accepted if it satisfies all constraints and its fitness value has improved. Elite local perturbations only optimize high-quality individuals, and the addition of repair and selective acceptance rules both explores the optimization space of elite solutions and prevents the destruction of high-quality feasible solutions.

[0094] S6.2: Calculate the relative fitness of an individual. calculate ,in Represents an individual Overall adaptability and These represent the maximum and minimum overall fitness of the current population, respectively. To assign a unique number to an individual within the population; Representative of individuals The relative fitness of individuals; normalized to the 0~1 interval, as a unified adjustment scale for adaptive probability, to adapt to the two sets of probability calculations for crossover and mutation.

[0095] S6.3: Adaptive crossover probability calculation, based on the individual's relative fitness, calculates the first... Individual crossover probability :

[0096]

[0097] in and This represents the pre-defined upper and lower bounds of the crossover probability. This represents the maximum number of iterations. Based on the calculated crossover probabilities, a two-point crossover operation is performed to generate offspring individuals. High-quality individuals. Smaller size and higher crossover probability enable superior individuals to participate in more gene recombination, improving the ability to develop superior search regions; at the same time, as the iteration process progresses, the crossover probability gradually decreases, enhancing the global search capability in the early stage and reducing population disturbance in the later stage, promoting stable convergence of the algorithm.

[0098] S6.4: Calculation of basic variation probability, based on individual relative fitness. Individual mutation probability :

[0099]

[0100] in and This represents the pre-set upper and lower boundary values ​​of the basic mutation probability; high-quality individuals have a lower mutation probability, ensuring the stable retention of good solutions; mutation is weaker in the early stages of iteration and stronger in the later stages, improving the ability to escape local optima in the later stages.

[0101] S6.5: Radiation gradient-guided variation for selected pick-up / drop-off points. In the set of unselected candidate pick-up and drop-off points Define local neighborhood:

[0102] ,

[0103] in, Selected pick-up and drop-off points This is a collection of all selected pick-up and drop-off points. Selected pick-up and drop-off points The corresponding local candidate neighborhood, Candidate pick-up and drop-off points within the neighborhood. Representative has selected pick-up and drop-off point and pick-up / drop-off point Spatial distance; For the first The neighborhood correction radius corresponding to the iteration. Indicates the initial neighborhood radius. Used to control the radius decay rate; within this neighborhood, the dose gradient is calculated:

[0104]

[0105] in, Selected pick-up and drop-off points Candidate pick-up and drop-off points in the surrounding area The radiation dose gradient between them; Representative pick-up and drop-off point radiation dose rate, Representative pick-up and drop-off point ; Pick-up and drop-off point , The spatial straight-line distance between the two; only the gradient towards the lower radiation point is identified, and the gradient in the direction of increasing radiation is set to zero, guiding only optimization towards a safer point.

[0106] Calculate the maximum dose gradient Further calculation of radiation sensitivity factor ,in ; This is the set of all selected pick-up and drop-off points. Refers to any selected pick-up / drop-off point within the set; the sensitivity factor is normalized to 0~1 to quantify the potential for radiation reduction that can be optimized around the current point.

[0107] S6.6: Correcting the basic variability probability using radiation sensitivity factors:

[0108] ,

[0109] in, This refers to the individual baseline variance probability calculated previously. This represents the final mutation probability after correction for the radiation sensitivity factor. The radiation gradient influence coefficient is used to regulate the intensity of the intervention of radiation spatial features on the mutation probability. Mutation operations are performed based on the corrected mutation probability to complete genetic evolution. The higher the radiation optimization potential, the greater the mutation probability, and the more likely the algorithm is to replace the pick-up point. High-quality low-radiation points have a lower mutation probability and are not easily modified, enabling the population to actively converge towards low-radiation areas.

[0110] S7: Output the optimal pick-up / drop-off point location scheme. Repeat the selection, crossover, and mutation operations in step S6, and update the population fitness values. Iteration link reference. Figure 2The population is iteratively updated repeatedly. The genetic evolution process stops when a preset termination condition is met. The individual with the best fitness is selected from the final population, and the final pick-up point location scheme is determined based on the binary chromosome code corresponding to the best individual. The set of selected pick-up points is then output.

[0111] Example 2:

[0112] This embodiment is used to verify the effectiveness and stability of the nuclear emergency evacuation point site selection optimization method based on radiation risk perception proposed in this invention in actual nuclear accident evacuation scenarios.

[0113] like Figure 5 As shown, a nuclear accident emergency road network was constructed, encompassing all road intersections, candidate points, evacuation routes, shelters, and the accident site. The study area covers approximately 100 residential communities within the emergency planning zone, with a total population of about 25,000, and includes 86 candidate pick-up and drop-off points. During the evacuation, 200 evacuation buses were deployed, each with a rated passenger capacity of 50 people, and the vehicles were set to travel at a speed of 50 km / h under normal traffic conditions.

[0114] In this embodiment, the input data for each algorithm is kept consistent, including the candidate pick-up and drop-off point set, village population size, road network structure, and radiation field parameters, to ensure the comparability and fairness of the experimental results. The following four methods are used to optimize the pick-up and drop-off point location:

[0115] (1) Fixed-weight single-objective genetic algorithm (B1);

[0116] (2) Intra-generation normalized weighted genetic algorithm (B2);

[0117] (3) Multi-objective genetic algorithm based on NSGA-II (B3);

[0118] (4) The improved hybrid genetic algorithm (IHGA, B4) proposed in this invention.

[0119] To ensure fairness in the comparison, the pick-up and drop-off point selection results generated by each algorithm were input into the same evacuation route planning and vehicle scheduling module for unified evaluation, so that performance differences only stemmed from the pick-up and drop-off point selection methods themselves. Each algorithm was run independently 5 times, and the evacuation completion time and the average cumulative radiation dose per person were recorded as two evaluation indicators. The results are shown in Table 1.

[0120] Table 1

[0121]

[0122] As shown in Table 1, the method proposed in this invention outperforms the comparative algorithms in both evacuation performance and safety. Specifically, in terms of evacuation time, compared to the three traditional algorithms B1, B2, and B3, the evacuation completion time of the B4 algorithm of this invention is reduced by approximately 6.2%, 6.7%, and 5.9%, respectively; and in terms of average radiation dose, it is reduced by approximately 57.2%, 74.4%, and 54.6% compared to B1, B2, and B3, respectively, significantly optimizing radiation safety protection and achieving a dual improvement in efficiency and safety.

[0123] At the same time, such as Figure 6 As shown, the box plot further illustrates the statistical distribution characteristics and robustness differences of each comparative algorithm. Results from multiple independent runs of different algorithms demonstrate that the method of this invention exhibits less dispersion and a more stable median in both evacuation time and radiation dose, indicating superior statistical stability and resistance to random disturbances. Experimental results show that the method of this invention can effectively reduce radiation risk while ensuring evacuation efficiency and improve the stability and reliability of pick-up point location schemes, making it suitable for the pick-up point optimization problem in nuclear accident emergency evacuation scenarios.

[0124] In summary, this invention effectively addresses various shortcomings of traditional nuclear emergency pick-up point site selection methods through an integrated improved design that combines dynamic radiation risk modeling, multi-objective dynamic evaluation, and radiation gradient-guided evolution. It is adaptable to complex dynamic nuclear accident radiation scenarios, and the output site selection scheme is safe, efficient, and balanced, with stable and reliable solution results, making it highly valuable for engineering applications.

[0125] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for optimizing the location of nuclear emergency transport points based on radiation risk perception, characterized in that, Includes the following steps: S1. Construct a basic model for the selection of nuclear emergency pick-up and drop-off points, determine the set of candidate pick-up and drop-off points and the set of residential areas in the area to be optimized, define the location decision variables and allocation variables, and set the location constraints. S2. Construct a multi-objective optimization model, with evacuation efficiency, radiation exposure risk, and load balancing at pick-up and drop-off points as the core optimization objectives, and establish a multi-objective collaborative optimization mathematical model. S3. Construct a radiation risk assessment model to simulate the atmospheric diffusion process of radioactive materials in a nuclear accident and quantify the radiation risk indicators of each candidate delivery point. S4. A hybrid initialization strategy is adopted to generate an initial population. An elite population is constructed by combining the radiation security and service capabilities of candidate pick-up and drop-off points. This is combined with a random population to form a complete initial population, taking into account both the quality of the initial solution and the diversity of the population. S5. Establish a multi-objective dynamic fitness evaluation mechanism to dynamically calculate and evaluate the fitness of each solution in the initial population. S6. Perform genetic evolution operations based on fitness evaluation results, including selection operations, adaptive crossover operations, and radiation gradient-guided mutation operations. S7. Iterate through steps S5 and S6 until the preset iteration termination condition is met, and output the optimal nuclear emergency pick-up and drop-off point location scheme.

2. The method for optimizing the location of nuclear emergency transport points based on radiation risk perception according to claim 1, characterized in that, In step S1, the decision variables, allocation variables, and constraints are specifically defined as follows: Define decision variables ,satisfy ,in This represents the set of candidate pick-up and drop-off points. This indicates the number of pick-up and drop-off points to select; Define allocation variables Each village is assigned to only one selected pick-up and drop-off point, thus satisfying the requirement. and satisfy This means that villages are only allowed to be assigned to selected pick-up and drop-off points, among which... It represents a collection of villages.

3. The method for optimizing the location of nuclear emergency transport points based on radiation risk perception according to claim 1, characterized in that, In step S2, the multi-objective optimization model includes three optimization objectives, the specific expressions of which are as follows: Objective 1: Average evacuation distance is defined as follows: in Indicates village and pick-up / drop-off point The distance between them; this objective is used to characterize the average travel distance of residents to the pick-up and drop-off points; Objective 2, the cumulative radiation risk target is defined as: in Indicates pick-up and drop-off point Radiation dose rate at the location Indicates allocation to pick-up / drop-off point The total number of people; this target is used to measure the overall radiation exposure risk faced by residents during the assembly phase; Objective 3: Load balancing at pick-up and drop-off points is defined as follows: in, This indicates the number of pick-up and drop-off points to select. This represents the average population served by a single pick-up / drop-off point. Indicates allocation to pick-up / drop-off point Total number of people This indicates the selection of a set of pick-up and drop-off points. Indicates village The population; this target is used to measure the degree of load balancing among the pick-up and drop-off points.

4. The method for optimizing the location of nuclear emergency transport points based on radiation risk perception according to claim 1, characterized in that, In step S3, the radiation risk assessment model is constructed based on the Gaussian plume model, specifically including: The spatiotemporal diffusion process of radioactive materials in a nuclear accident was simulated using a Gaussian plume model to obtain the spatiotemporal distribution of spatial radiation dose rate in the area to be optimized. The instantaneous radiation dose rate of candidate pick-up and drop-off points is extracted from the discrete time series. The discrete time dose rate data is approximated by integral calculation using the trapezoidal integral formula to obtain the cumulative radiation dose of the candidate pick-up and drop-off points. The cumulative radiation dose is used as the radiation risk assessment index of the candidate pick-up and drop-off points.

5. The method for optimizing the location of nuclear emergency transport points based on radiation risk perception according to claim 1, characterized in that, In step S4, the hybrid initialization strategy specifically includes the following steps: S41. Evaluate the radiation safety of candidate pick-up and drop-off points; [List candidate pick-up and drop-off points] The dose rate at the site Then its normalized radiation safety factor is ; S42. Evaluate the service capabilities of candidate pick-up and drop-off points; set candidate pick-up and drop-off points. The population within the service area is ,in In the formula To constrain the maximum service radius, the service population is normalized. This represents a normalized service capability indicator; S43. Constructing comprehensive evaluation indicators ,in The weighting coefficients are used to rank the candidate pick-up and drop-off points in descending order based on their comprehensive scores, and the highest-scoring points are selected. A pool of candidate pick-up and drop-off points is formed; points are randomly selected from the elite pool. Each node generates an elite individual; S44. Randomly select nodes from the elite pool to generate a preset number of elite individuals. The remaining individuals in the population are obtained by random generation. The elite population and the random population together constitute a complete initial population.

6. The method for optimizing the location of nuclear emergency transport points based on radiation risk perception according to claim 1, characterized in that, In step S5, the multi-objective dynamic fitness evaluation mechanism specifically includes: S51. Radiation Risk Target Transformation: A logarithmic transformation is performed on the cumulative radiation risk targets to mitigate the impact of numerical scaling bias in radiation risk. The transformation formula is as follows: ,in Indicates the scaling adjustment parameters; S52, Cross-generational dynamic normalization processing: For the objective function Calculate the first The lower quantile of the target value for the next generation population and upper quantile The normalized boundary is updated using exponential moving average. , ,in Indicates the smoothing coefficient; Calculate the normalized objective value based on the updated boundary. ,in Indicates the current individual, Represents an individual In the objective function The value of , when hour ,when hour ,when hour ; To prevent the normalization results from exceeding the preset range, the normalization results are truncated. ; S53, Target Weight Dynamic Adaptive Update: Calculate the... Objective function in generation population normalized median Describe the objective function The median evaluation index; Calculate temporary weights based on the median. ,in This represents the sensitivity coefficient for weight adjustment; Use exponential moving average to update weights ,in This indicates the adaptive adjustment rate of the weights. For the first Replace the weights of the objective function; Set a lower bound constraint on the weights. After the weights are updated, normalization is performed to ensure that they meet the requirements. ,in The minimum weight threshold; S54. Overall Fitness Calculation: Based on the dynamically updated target weights, the overall fitness value of each individual in the population is calculated using a weighted average. 。 7. The method for optimizing the location of nuclear emergency transport points based on radiation risk perception according to claim 1, characterized in that, In step S6, the selection operation specifically involves: using a tournament selection mechanism to screen and construct the parent population; and performing a local perturbation operation on elite individuals with high fitness rankings every preset number of iterations. The local perturbation operation includes randomly swapping the selected and unselected pick-up / drop-off points corresponding to the gene loci of elite individuals, performing a feasibility repair process, and retaining only the perturbation individuals that comply with constraints and have improved fitness.

8. The method for optimizing the location of nuclear emergency transport points based on radiation risk perception according to claim 1, characterized in that, In step S6, the adaptive crossover operation specifically involves: calculating the relative fitness of individuals, combining the preset crossover probability boundary value and the maximum number of iterations, adaptively calculating the real-time crossover probability of each individual, and performing a two-point crossover operation based on the real-time crossover probability to generate offspring individuals.

9. The method for optimizing the location of nuclear emergency transport points based on radiation risk perception according to claim 1, characterized in that, In step S6, the radiation gradient-guided mutation operation specifically includes: S61. Define the local candidate neighborhood of the selected pick-up and drop-off points based on the preset initial neighborhood radius and radius decay rate. S62. Calculate the radiation dose gradient and the maximum dose gradient within the local candidate neighborhood, and solve for the individual radiation sensitivity factor. S63. Based on the radiation sensitivity factor and the preset influence coefficient, correct the individual basic variation probability obtained from the individual relative fitness; S64. Perform mutation operations based on the corrected mutation probability to complete the population mutation update and guide the population to converge towards a low radiation risk area.

10. The method for optimizing the location of nuclear emergency transport points based on radiation risk perception according to claim 8 or 9, characterized in that, The relative fitness of the individual, for the individual calculate ,in Represents an individual Overall adaptability and These represent the maximum and minimum fitness of the current population, respectively; the crossover probability and the basic mutation probability are adaptively calculated based on the individual relative fitness.

Images