Decision-making method for generating and optimizing dynamic data intelligent logistics scheme of urban industrial chain and supply chain

Abstract

The invention belongs to the technical field of intelligent logistics, and provides an urban industrial chain and supply chain dynamic data intelligent logistics scheme generation and optimization decision-making method comprising the following steps: S1, data acquisition and data preprocessing are carried out, and the data comprise supplier data, demand point data and traffic network data; s2, dynamic data modeling: constructing a dynamic data model and a multi-target mixed integer programming model of the urban industrial chain logistics system; s3, demand prediction is carried out based on Kalman filtering; and S4, constructing a hybrid optimization algorithm to generate and optimize a logistics scheme. According to the invention, intelligent generation and dynamic optimization of the logistics scheme are realized; therefore, dynamic data in an urban industrial chain can be processed, and the adaptability of a logistics scheme to real-time changes is improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of intelligent logistics, and particularly relates to a decision-making method for generating and optimizing an intelligent logistics scheme of dynamic data of an urban industrial chain supply chain. BACKGROUND

[0002] In a modern urban economic system, efficient operation of an urban industrial chain is one of the key factors to promote regional economic development, and a logistics link as an important support of the urban industrial chain is directly related to the operation efficiency and benefits of the entire industrial chain. The logistics link in the urban industrial chain is a complex system involving multiple subjects, multiple targets and multiple constraints. Traditional logistics schemes are often based on static data and experience rules, and are difficult to adapt to dynamic changes and unexpected situations of the urban industry. For example, in the process of formulating a logistics scheme, there are differences in interest demands and information asymmetry between subjects, which is difficult to coordinate. Moreover, the logistics scheme often needs to meet multiple targets to achieve better results. Among them, cost control, service quality, environmental protection and capacity utilization often need to be considered. The existing judgment and optimization methods based on static data and experience rules lack scientificity and flexibility, and are difficult to meet the optimization needs of multiple targets. In addition, experience rules are difficult to cover all possible unexpected situations, such as the impact of public health events and traffic control on logistics activities. In unexpected situations, traditional logistics schemes often fall into chaos due to the lack of effective response mechanisms, and cannot guarantee the stable operation of the urban industrial chain. Multiple constraints involve transportation capacity, storage space, time window and other limiting conditions, which further increase the complexity of logistics system optimization. Existing technologies often cannot be considered, and after the relevant conditions change, the logistics scheme based on static data is difficult to quickly adapt to changes, resulting in a disconnection between logistics activities and the actual needs of the industrial chain, affecting the overall efficiency of the industrial chain. SUMMARY

[0003] The purpose of the present application is to provide a decision-making method for generating and optimizing an intelligent logistics scheme of dynamic data of an urban industrial chain supply chain, to solve the technical problem that existing technologies are often based on static data and experience rules, and are difficult to adapt to dynamic changes and unexpected situations of the urban industry.

[0004] The decision-making method for generating and optimizing an intelligent logistics scheme of dynamic data of an urban industrial chain supply chain comprises the following steps:

[0005] S1, data acquisition and data preprocessing, the data including supplier data, demand point data and traffic network data;

[0006] S2, dynamic data modeling, including constructing a dynamic data model of the logistics system of the urban industrial chain and a multi-target mixed integer programming model;

[0007] S3, demand prediction based on Kalman filter;

[0008] S4, construction of a hybrid optimization algorithm for generation and optimization of the logistics scheme, the hybrid optimization algorithm finds the optimal solution that adapts to dynamic data changes by combining population evolution and local search, wherein the fitness formula for evaluating the logistics scheme is the objective function of the multi-objective mixed integer programming model.

[0009] Preferably, in step S2, the dynamic data model includes a supply end model, a demand end model, and a space-time dynamic network model of the logistics network; the expressions of the specific models are as follows:

[0010] 1) Supply end model: S(t) = {s1(t), s2(t), …, s1(t)}, wherein s i (t) represents the supply capacity vector of supplier i at time t, reflecting the fluctuation of supplier capacity;

[0011] 2) Demand end model: D(t) = {d1(t), d2(t), …, d1(t)}, wherein d j (t) represents the demand vector of demand point j at time t, reflecting real-time prediction of customer demand;

[0012] 3) Space-time dynamic network model: G(t) = (V, E(t)), E(t) = {(i, j) | d ij ·δ ij (t) ≤ maximum travel time}, wherein V represents a node set, E(t) represents an edge set at time t, (i, j) represents the nodes i and j at both ends of the edge e(t) in the edge set E(t), d ij represents the distance of the corresponding edge, and δ ij (t) represents the road congestion index of the corresponding edge.

[0013] Preferably, the objective function of the multi-objective mixed integer programming model is as follows:

[0014] min Z = α·TC + β·CE + γ·(1-SL) + η·UC,

[0015] wherein Z is the result of the objective function, TC is the economic cost, CE is the environmental cost, SL is the service penalty, (1-SL) measures the service level, UC is the capacity utilization cost, and α, β, γ, and η are the respective weights of the economic cost, the environmental cost, the service level, and the capacity utilization cost, respectively.

[0016] Preferably, step S4 includes:

[0017] S4.1, initialization stage: generating an initial population based on dynamic data;

[0018] S4.2, Iterative optimization phase: gradually improve the quality of the solution through the alternating dominance of genetic algorithm and tabu search;

[0019] S4.3, Elite preservation strategy: retain the individual with the highest fitness in each generation;

[0020] S4.4, Update the global optimal solution: check if there is a better solution in the current population, if the fitness value of the current optimal solution is greater than that of the global optimal solution, update the global optimal solution.

[0021] Preferably, step 4.1 includes:

[0022] S4.1.1, Parameter setting: set the required parameters of genetic algorithm and tabu search;

[0023] S4.1.2, Initialization of population: if an initial solution is provided, use it to generate the population; otherwise, randomly generate the initial population;

[0024] S4.1.3, Evaluation of initial population: fitness calculation is performed on the initial population;

[0025] S4.1.4, Initialization of global optimal solution: select the individual with the highest fitness from the initial population as the global optimal solution;

[0026] S4.1.5, Initialization of tabu list: use the tabu list to store the prohibited solutions in tabu search, which is initially empty.

[0027] Preferably, in step 4.2, the genetic algorithm dominant stage performs global exploration through selection, crossover, and mutation operations to generate diverse solutions, and the tabu search dominant stage applies tabu search to each individual for local optimization to fine-tune the quality of the solution. The tabu list records recently searched solutions, and the memory period is controlled by the tabu length; the tabu search dominant stage adopts a double-layer search architecture, using a wide-range search operator at the strategic level and a fine-tuning operator at the tactical level.

[0028] Preferably, the tabu search dominant stage includes the following steps:

[0029] 1) Level judgment: determine the search strategy according to the input level parameter, the search strategy includes two search strategies of strategic level and tactical level;

[0030] 2) Neighborhood generation:

[0031] a. Level parameter for strategic level: generate candidate solutions related to facility layout and transportation mode;

[0032] b. Level parameter for tactical level: generate candidate solutions related to path adjustment and inventory allocation;

[0033] 3) Tabu screening and selection: select the optimal solution from the neighborhood, this step considers:

[0034] a. Tabu list: store the forbidden solutions in the tabu search based on the tabu list to avoid repeated search of recent operations;

[0035] b. aspiration criterion: when the forbidden solution meets the aspiration criterion, it is a high-quality solution, and the high-quality solution is allowed to break the tabu list limit;

[0036] 4) Update the solution: apply the selected movement operation to the current solution to generate and return the optimized solution.

[0037] Preferably, the decision-making method further comprises: S5, dynamically collecting data and dynamically adjusting the logistics scheme based on a dynamic adjustment mechanism; the final output of the previous step is the optimal logistics scheme, which is sent to the logistics execution system for execution, real-time monitoring of the execution and collection of related dynamic data, and update detection and processing based on the dynamic data; step S5 comprises:

[0038] 1) Detect whether the dynamic data has changed: if the data has changed, update the dynamic data and re-evaluate the population fitness;

[0039] 2) Self-adaptive adjustment of algorithm parameters: self-adaptively adjust the crossover probability, mutation probability and tabu length according to the current generation number;

[0040] 3) Output the optimal solution of the current generation: at the end of each iteration, output the optimal fitness value of the current generation, and the algorithm finally returns the updated global optimal solution.

[0041] Preferably, in step S4, the fitness formula is dynamically adjusted based on the real-time collected data when evaluating the initial population, and the corresponding weights α, β, γ and η of each target are dynamically adjusted according to the real-time data, and the target weight self-adaptive formula used for dynamic adjustment is as follows:

[0042]

[0043] Wherein, α(t) represents the weight α dynamically adjusted with time t, and k is a constant used to adjust the sensitivity of the formula, and the target weight self-adaptive formula can also be applied to weights β, γ and η, and the results are β(t), γ(t) and η(t) in turn.

[0044] Preferably, in the step of self-adaptive adjustment of algorithm parameters, a dynamic parameter self-adaptive mechanism is applied, and the crossover probability P c (t), the mutation probability P m (t) and the tabu length L(t) are calculated dynamically according to the following formulas respectively:

[0045]

[0046] wherein, δ ij (t) represents the road congestion index of the corresponding edge, which is a positive number greater than 1, P c (t) is the intersection probability, which decreases with the increase of demand fluctuation, and is used to reduce the destructive operation, P c 0 represents the initial intersection probability, and λ represents the sensitivity coefficient; P m (t) is the mutation probability, which increases with the increase of traffic congestion, and is used to enhance the exploration ability, represents the initial mutation probability, β is the traffic influence coefficient, the traffic congestion index is a normalized value of 0-1, and is used to measure the real-time road congestion degree of the related area as a whole; L(t) is the taboo length, which is dynamically adjusted according to the road congestion index δ ij (t) of the corresponding edge, and is used to avoid falling into local optimum, and L0 is the basic taboo length, represents the upward rounding function.

[0047] The technical advantages of the present application are as follows:

[0048] 1. The present application realizes intelligent generation and dynamic optimization of the logistics scheme by dynamic data modeling, combined with a mixed integer programming optimization algorithm and a real-time feedback mechanism; therefore, dynamic data in the urban industrial chain can be processed, and the adaptability of the logistics scheme to real-time changes is improved.

[0049] 2. The present application realizes multi-objective optimization, and the fitness function applied in the optimization algorithm is calculated based on the objective function of the multi-objective mixed integer programming model, which can comprehensively reflect the advantages and disadvantages of individuals. The designed fitness function can balance multiple objectives such as cost control, service quality, environmental protection and capacity utilization, meet the multi-objective demand of the logistics scheme of the urban industrial chain, and realize the sustainable development of urban logistics.

[0050] 3. The genetic operation (selection, crossover, mutation) in the optimization algorithm of the present application directly acts on the decision variables, ensuring that the generated offspring meets the constraint conditions of the optimization model; and the applied taboo search performs local optimization in the solution space composed of decision variables; the alternately applied two optimization methods solve complex optimization problems in a dynamic environment, improving the solving efficiency and quality.

[0051] 4. The present application periodically collects real-time traffic data and demand data, and the demand prediction model based on Kalman filtering can accurately predict demand changes and reduce the impact of uncertainty.

[0052] 5、The application can monitor the execution after sending the optimized scheme to the logistics execution system, and further perform real-time processing and feedback on abnormal conditions. The optimization algorithm can realize dynamic response based on dynamic data, dynamically update the parameters of the optimization model, and further affect the fitness evaluation of the population, so that the scheme can adapt in time after the relevant conditions of the logistics system change, and the robustness of the system is improved. BRIEF DESCRIPTION OF DRAWINGS

[0053] Figure 1 The flowchart of the decision method for the urban industrial chain supply chain dynamic data intelligent logistics scheme generation and optimization of the application. DETAILED DESCRIPTION

[0054] The specific embodiments of the application will be further described in detail below with reference to the drawings, and by describing the embodiments, to help those skilled in the art to have a more complete, accurate and in-depth understanding of the inventive concept and technical scheme of the application.

[0055] As shown in Figure 1 , the application provides a decision method for urban industrial chain supply chain dynamic data intelligent logistics scheme generation and optimization, comprising the following steps.

[0056] S1, data acquisition and data preprocessing.

[0057] Collecting supplier data, demand point data, traffic network data and the like in the urban industrial chain, and performing cleaning and feature extraction.

[0058] S2, dynamic data modeling.

[0059] This step comprises the following sub-steps.

[0060] S2.1, constructing a dynamic data model of the urban industrial chain logistics system, the dynamic data model comprising a supply end model, a demand end model and a space-time dynamic network model of the logistics network. The expression of the specific model is as follows.

[0061] 1) Supply end model: S(t) = {s1(t), s2(t), …, s1(t)}, wherein s i (t) represents the supply capacity vector of supplier i at time t, reflecting the fluctuation of the production capacity of the supplier.

[0062] 2) Demand end model: D(t) = {d1(t), d2(t), …, d1(t)}, wherein d j (t) represents the demand vector of demand point j at time t, reflecting the real-time prediction of customer demand.

[0063] 3) Space-time dynamic network model: G(t) = (V, E(t)), E(t) = {(i, j) | d ij• δ ij (t)≤max travel time}, where V represents a set of nodes, E(t) represents a set of edges (i.e., a set of paths) at time t, (i,j) are nodes i,j at both ends of an edge e(t) in the set of edges E(t), d ij represents the distance of the corresponding edge, δ ij (t) represents the road congestion index of the corresponding edge, the edge e(t) also has a corresponding dynamic weight w e (t), E(t) is dynamically updated over time, reflecting the impact of real-time traffic conditions on logistics paths, and through the dynamic weight w e (t), the city traffic flow data involves dynamic parameters such as road congestion index δ ij (t), supplier capacity fluctuation s i (t) (i.e., the supply capacity vector of supplier i changing over time), real-time prediction of customer demand d j (t) (i.e., the demand vector of demand point j changing over time), carbon trading price λ(t), etc.

[0064] 4) Time window constraints of the spatiotemporal dynamic network model: where TW ij represents the feasible time window of node i to node j, is the start time of the feasible time window, is the end time of the feasible time window.

[0065] The above dynamic data model introduces multi-dimensional dynamic data fusion through infrastructure innovation, and implements the impact of real-time logistics data decision-making reflecting road conditions (congestion), inventory (supply capacity and customer demand), etc.

[0066] S2.2, build a multi-objective mixed integer programming model.

[0067] Currently, in terms of multi-objective scheme optimization, it involves cost control, service quality, environmental protection, and capacity utilization in multiple aspects, so it is necessary to set up corresponding models for optimization. The objective function of model optimization is as follows:

[0068] min Z = α·TC + β·CE + γ·(1-SL) + ηUC,

[0069] The objective function needs to minimize economic cost, minimize environmental cost, maximize service level and minimize capacity utilization cost, wherein Z is the result of the objective function, TC is the economic cost, CE is the environmental cost, SL is the service penalty, when the service penalty is maximum, the service level is optimal, therefore the service level is measured by (1-SL), UC is the capacity utilization cost, the smaller the capacity utilization cost is, the higher the capacity utilization is, and a, b, g and h are the corresponding weights of the economic cost, the environmental cost, the service level and the capacity utilization cost respectively.

[0070] The economic cost TC is the sum of all costs of the logistics scheme, including transportation cost, warehousing cost and handling cost in logistics. This part is an optimization objective commonly used in the prior art. The carbon emission cost in logistics

[0071] The environmental cost CE is the carbon emission cost generated by logistics, and in the present scheme, dynamic carbon cost is adopted, which can be dynamically calculated according to the collected data, and the corresponding formula is as follows:

[0072]

[0073] Wherein, λ(t) is the carbon trading price, used to encourage low-carbon path selection, E(t) represents the edge set, m represents the transportation mode (road, railway, waterway, air), is a binary decision variable, which represents whether the corresponding transportation mode is selected for the path (i, j), and the value of 1 represents selection, and the value of 0 represents non-selection; the carbon emission factor m represents the unit carbon emission coefficient of the corresponding transportation mode m.

[0074] The service quality of the traditional evaluation logistics scheme is either based on user feedback score or based on binary judgment of “whether the time window is met”, the former has the problems of hysteresis and weak objectivity, and the latter as a binary judgment, it is difficult to finely reflect the influence of service quality caused by time delay. The present scheme introduces the service penalty SL as a continuous service level index, which more finely reflects the influence of delay. The corresponding formula is as follows:

[0075]

[0076] Wherein, t j represents the time of arriving at the demand point j to complete the logistics service, and X represents the set of demand points.

[0077] The capacity utilization cost UC is used to measure the capacity utilization, and the smaller the value is, the higher the capacity utilization is, so as to avoid excess or insufficient capacity of the factory and improve the efficiency of the industrial chain. The corresponding formula is as follows:

[0078]

[0079] Wherein, F represents a factory set, containing a specific factory f, represents the quantity of logistics goods (raw materials or parts) supplied by supplier i to factory f at time t, W f represents the rated capacity of factory f.

[0080] In this way The deviation degree of the capacity utilization rate of the corresponding factory f is embodied:

[0081] When the utilization rate is 100%, the deviation degree is 0, and the capacity utilization rate cost is the lowest.

[0082] When the factory capacity is idle, there is equipment idling.

[0083] When the factory is overloaded, there is overtime, and the equipment wear is intensified.

[0084] S3, demand prediction based on Kalman filtering.

[0085] This step is based on Kalman filtering to predict demand, and the Kalman filtering prediction model is as follows.

[0086]

[0087] P t|t =(I-K t H t )P t|t-1 ,

[0088] Wherein, is the predicted state, P t|t-1 is the prediction error covariance, K t is the Kalman gain, z t is the observation value, F t is the state transition matrix, H t is the state transition matrix, Q t process noise covariance, R t observation noise covariance.

[0089] S4, construct a hybrid optimization algorithm based on improved genetic algorithm and tabu search to generate and optimize the logistics scheme.

[0090] The method provides a hybrid optimization algorithm based on improved genetic algorithm and tabu search, which combines the characteristics of genetic algorithm (Genetic Algorithm, GA) and tabu search (Tabu Search, TS), and is used for solving dynamic optimization problems. The algorithm finds the optimal solution suitable for dynamic data changes by combining population evolution and local search. Specifically, the following contents are included.

[0091] S4.1, Initialization phase: generate initial population based on dynamic data (e.g. real-time traffic, demand fluctuation). Specifically, the following steps are included.

[0092] S4.1.1, Parameter setting: set required parameters such as population size, maximum iteration number, mutation rate, and tabu length.

[0093] S4.1.2, Initialize population: if an initial solution is provided, use it to generate the population; otherwise, randomly generate the initial population.

[0094] S4.1.3, Evaluate initial population: calculate the fitness of the initial population.

[0095] The fitness formula used in this step is the objective function of the multi-objective mixed integer programming model, and can be dynamically adjusted based on real-time data collected. The corresponding weights of each objective, a, β, γ, and η, are dynamically adjusted according to real-time data. The adaptive formula for adjusting the target weight is as follows:

[0096]

[0097] where a(t) represents the dynamically adjusted weight a over time t, and k is a constant used to adjust the sensitivity of the formula. This adaptive formula for adjusting the target weight can also be applied to weights β, γ, and η, with the results being β(t), γ(t), and η(t) in turn.

[0098] S4.1.4, Initialize global optimal solution: select the individual with the highest fitness from the initial population as the global optimal solution.

[0099] S4.1.5, Initialize tabu list: use the tabu list to store prohibited solutions in the tabu search, which is initially empty.

[0100] S4.2, Iterative optimization phase: gradually improve the quality of the solution through the alternating dominance of genetic algorithm (GA) and tabu search (TS). In the genetic algorithm dominant stage, global exploration is performed through selection, crossover, and mutation operations to generate diverse solutions. The genetic algorithm dominant stage accounts for 2 / 3 of the iteration times, and is performed twice every three iterations. In the tabu search dominant stage, each individual is subjected to tabu search for local optimization to fine-tune the quality of the solution. The tabu search dominant stage accounts for 1 / 3 of the iteration times, and is performed once every three iterations.

[0101] The genetic algorithm dominant stage includes the following operations:

[0102] S4.2.1, Selection operation: select parent individuals from the current population based on fitness values for generating the next generation.

[0103] S4.2.2, crossover operation: gene recombination is performed on the selected parents to generate offspring.

[0104] S4.2.3, mutation operation: random perturbation is performed on the offspring genes with a certain probability to increase the diversity of solutions.

[0105] The taboo search dominant stage includes the following operations:

[0106] S4.2.4, local search: taboo search is performed on each offspring individual, and each new solution generated by genetic operation is locally optimized. This step uses a taboo list to record recently searched solutions to avoid repeated search into local optimum; at the same time, the memory period is controlled by the taboo length to balance exploration and development.

[0107] In this step, an innovative double-layer search architecture is adopted, which uses a wide-range search operator at the strategic side and a fine-tuning operator at the tactical layer, while traditional relaxation algorithms cannot balance global optimization and local fine-tuning. Specifically, the following steps are included.

[0108] 1) Level judgment: according to the input level parameter, the search strategy is determined. The search strategy includes two search strategies of strategic layer and tactical layer.

[0109] 2) Neighborhood generation:

[0110] a. Level parameter is strategic layer: generate candidate solutions related to facility layout and transportation mode, which are applied to distribution center layout and transportation mode selection.

[0111] b. Level parameter is tactical layer: generate candidate solutions related to path adjustment and inventory allocation, which are applied to path planning and inventory control.

[0112] 3) Taboo screening and selection: use the select_best_move function to select the optimal solution from the neighborhood, which considers the following:

[0113] a. Taboo list: based on the taboo list, the forbidden solutions in the taboo search are stored to avoid repeated search in recent operations.

[0114] b. Desire criterion: when the forbidden solution meets the desire criterion, it is a high-quality solution, which allows the high-quality solution to break through the taboo list limit.

[0115] 4) Update the solution: apply the selected move operation to the current solution through the apply_move function to generate and return the optimized solution.

[0116] S4.3, elite reservation strategy is executed: the individual with the highest fitness is reserved in each generation to ensure that the optimal solution is not lost in the iteration process. Specifically, the following contents are included.

[0117] S4.3.1, evaluate offspring: calculate the fitness value of the offspring individuals.

[0118] S4.3.2, merge parent and offspring: merge the parent and offspring.

[0119] S4.3.3, elite reservation: select the next generation population according to the fitness value, and reserve the individuals with the highest fitness value in each generation.

[0120] S4.4, update the global optimal solution: check whether there is a better solution in the current population, if the fitness value of the current optimal solution is greater than that of the global optimal solution, update the global optimal solution. The above steps gradually approach the global optimum by continuously accumulating high-quality solutions.

[0121] S5, dynamically collect data and adjust the logistics scheme based on the dynamic adjustment mechanism.

[0122] The initial scheme generated in the previous step is finally output as the optimal logistics scheme after scheme evaluation and optimization by the fitness function. The optimal logistics scheme is sent to the logistics execution system for execution. The system applying this method monitors the execution in real time and collects relevant dynamic data. This method has dynamic data response capability, and updates, detects and processes based on dynamic data. The specific steps are as follows.

[0123] 1) Detect whether the dynamic data has changed: if the data has changed, update the dynamic data and re-evaluate the fitness of the population.

[0124] 2) Self-adaptive adjustment of algorithm parameters: adaptively adjust the crossover probability, mutation probability and taboo length according to the current generation number.

[0125] This step applies a dynamic parameter self-adaptive mechanism, and the crossover probability P c (t), the mutation probability P m (t) and the taboo length L(t) are calculated dynamically as follows:

[0126]

[0127] Where δ ij (t) represents the road congestion index of the corresponding edge, which is a positive number greater than 1, such as 1.5, which means the road congestion degree is 50%, P c (t) is the crossover probability, which decreases with the increase of demand fluctuation, and is used to reduce destructive operations, P c 0 is the initial crossover probability, and λ is the sensitivity coefficient; P m (t) is the mutation probability, which increases with the increase of traffic congestion, and is used to enhance the exploration ability, represents the initial mutation probability, β is the traffic influence coefficient, the traffic congestion index is obtained from the urban traffic big data, is used for measuring the real-time road congestion degree of the whole relevant area, is a 0~1 normalized value, 1 represents serious congestion; L(t) is the taboo length, according to the road congestion index δ ij (t) is dynamically adjusted, is used for avoiding falling into local optimum, L0 is the basic taboo length, represents the upward rounding function.

[0128] 3) output the optimal solution of the current generation: at the end of each iteration, the optimal fitness value of the current generation is output, and the algorithm finally returns the updated global optimal solution.

[0129] The application is exemplarily described above in combination with the drawings, and it is obvious that the specific implementation of the application is not limited by the above manner, and various non-essential improvements using the inventive concept and technical solution of the application, or directly applying the inventive concept and technical solution of the application to other occasions without improvement, are all within the protection scope of the application.

Claims

1. A decision method for urban industrial chain supply chain dynamic data intelligent logistics scheme generation and optimization, characterized in that: Comprise the following steps: S1, data acquisition and data preprocessing, data including supplier data, demand point data, traffic network data; S2, dynamic data modeling, including the construction of city industry chain logistics system dynamic data model and multi-objective mixed integer programming model; S3, based on Kalman filtering for demand prediction; S4, the construction of hybrid optimization algorithm for the generation and optimization of logistics scheme, the hybrid optimization algorithm is combined with population evolution and local search to find the optimal solution suitable for dynamic data changes, wherein the fitness formula for evaluating the logistics scheme is the objective function of the multi-objective mixed integer programming model.

2. The decision method for generating and optimizing a dynamic data intelligent logistics scheme of urban industrial chain supply chain according to claim 1, characterized in that: In step S2, the dynamic data model includes the supply end model, the demand end model and the space-time dynamic network model of the logistics network; the expressions of the specific models are as follows: 1) Supply side model: S(t) = {s1(t), s2(t), …, si(t)}, where si(t) represents the supply capacity vector of supplier i at time t, embodying the supplier capacity fluctuation; i (t) represents the supply capacity vector of supplier i at time t, embodying the supplier capacity fluctuation; 2) Demand side model: D(t) = {d1(t), d2(t), …, d1(t)}, where d j (t) represents the demand vector of demand point j at time t, reflecting real-time prediction of customer demand; 3) Spatiotemporal dynamic network model: G(t)=(V,E(t)),E(t)={(i,j)|d ij ·δ ij Let V be the set of nodes, E(t) be the set of edges at time t, and (i,j) be the nodes i,j at both ends of edge e(t) in edge set E(t). ij δ represents the distance to the corresponding edge. ij (t) represents the road congestion index of the corresponding side.

3. The decision method for generating and optimizing a dynamic data intelligent logistics scheme of urban industrial chain supply chain according to claim 1, characterized in that: The objective function of the multi-objective mixed integer programming model is as follows: minZ=α·TC+β·CE+γ·(1-SL)+ηUC, Wherein, Z is the result of the objective function, TC is the economic cost, CE is the environmental cost, SL is the service penalty, (1-SL) measures the service level, and UC is the capacity utilization cost, and α, β, γ and η are the corresponding weights of the economic cost, the environmental cost, the service level and the capacity utilization cost, respectively.

4. The decision method for generating and optimizing a dynamic data intelligent logistics scheme of urban industrial chain supply chain according to claim 1, characterized in that: Step S4 includes: S4.1, initialization stage: generating initial population based on dynamic data; S4.2, iterative optimization stage: through the alternative dominance of genetic algorithm and tabu search, gradually improving the quality of solution; S4.3, elite reservation strategy is executed: the individual with the highest fitness is reserved in each generation; S4.4, update the global optimal solution: check whether there is a better solution in the current population, if the fitness value of the current optimal solution is greater than that of the global optimal solution, update the global optimal solution.

5. The decision method for generating and optimizing a dynamic data intelligent logistics scheme of urban industrial chain supply chain according to claim 4, characterized in that: Step 4.1 includes: S4.1.1, parameter setting: setting the required parameters of genetic algorithm and tabu search; S4.1.2, initialize the population: if the initial solution is provided, use the initial solution to generate the population; otherwise, generate the initial population randomly; S4.1.3, evaluate the initial population: calculate the fitness of the initial population; S4.1.4, initialize the global optimal solution: select the individual with the highest fitness from the initial population as the global optimal solution; S4.1.5, initialize the tabu list: use the tabu list to store the prohibited solutions in tabu search, which is initially empty.

6. The decision method for generating and optimizing a dynamic data intelligent logistics scheme of urban industrial chain supply chain according to claim 4, characterized in that: In step 4.2, the genetic algorithm dominant stage carries out global exploration through selection, crossover and mutation operations to generate diverse solutions, the tabu search dominant stage applies tabu search to each individual for local optimization to fine-tune the quality of the solution, the tabu list is used to record the recently searched solutions, and the memory period is controlled by the tabu length; the tabu search dominant stage adopts a double-layer search architecture, using a wide-range search operator at the strategic side and a fine-tuning operator at the tactical layer.

7. The decision method for generating and optimizing a dynamic data intelligent logistics scheme of urban industrial chain supply chain according to claim 6, characterized in that: The tabu search dominant stage includes the following steps: 1) level judgment: according to the input level parameter, determine the search strategy, the search strategy includes two search strategies of strategic layer and tactical layer; 2) neighborhood generation: a. level parameter is strategic layer: generate candidate solutions related to facility layout and transportation mode; b. The hierarchical parameter is the tactical level: generate candidate solutions related to path adjustment and inventory allocation; 3) Tabu screening and selection: select the optimal solution from the neighborhood, this step considers: a. Tabu list: store the forbidden solutions in the tabu search based on the tabu list to avoid repeated search of recent operations; b. Desirability criteria: when the forbidden solution meets the desirability criteria, it is a high-quality solution, which is allowed to break the tabu list limit; 4) Update the solution: apply the selected movement operation to the current solution, generate and return the optimized solution.

8. The decision method for generating and optimizing a dynamic data intelligent logistics scheme of urban industrial chain supply chain according to claim 1, characterized in that: Further comprising: S5, dynamically collect data and dynamically adjust the logistics scheme based on the dynamic adjustment mechanism; The final output of the previous step is the optimal logistics scheme, which is sent to the logistics execution system for execution, real-time monitoring of the execution and collection of relevant dynamic data, and update detection and processing based on dynamic data; Step S5 includes: 1) Detect whether the dynamic data has changed: if the data has changed, update the dynamic data and re-evaluate the fitness of the population; 2) Self-adaptive adjustment of algorithm parameters: adjust the crossover probability, mutation probability and tabu length according to the current generation number; 3) Output the optimal solution of the current generation: at the end of each iteration, output the optimal fitness value of the current generation, and the algorithm finally returns the updated global optimal solution.

9. The decision-making method for generating and optimizing intelligent logistics solutions based on dynamic data of urban industrial chain and supply chain as described in claim 8, characterized in that: In step S4, when evaluating the initial population, the fitness formula is dynamically adjusted based on the real-time collected data, and the corresponding weights α, β, γ and η of each target are dynamically adjusted according to the importance of the real-time data. The target weight adaptive formula used for dynamic adjustment is as follows: Wherein, α(t) represents the weight α dynamically adjusted with time t, k is a constant used to adjust the sensitivity of the formula, this target weight adaptive formula can also be applied to weights β, γ and η, the results obtained are β(t), γ(t) and η(t) in turn.

10. The decision method for generating and optimizing a dynamic data intelligent logistics scheme of urban industrial chain supply chain according to claim 8, characterized in that: The dynamic parameter adaptive mechanism is applied in the step of adaptively adjusting the algorithm parameters, the crossover probability P c (t), the mutation probability P m (t) and the tabu length L(t) are respectively as follows: P c (t) = P c 0 • e -λ·需求波动指数 , where δ ij (t) is the road congestion index of the corresponding edge, which is a positive number greater than 1, P c (t) is the intersection probability, which decreases with the increase of demand fluctuation, and is used to reduce the destructive operation, P c 0 λ is the sensitivity coefficient; P m (t) is the mutation probability, which increases with the increase of traffic congestion, and is used to enhance the exploration ability, λ is the sensitivity coefficient; P m (t) is the mutation probability, which increases with the increase of traffic congestion, and is used to enhance the exploration ability, L(t) is the taboo length, which is dynamically adjusted according to the road congestion index δ ij (t) of the corresponding edge, and is used to avoid falling into local optimum, L0 is the basic taboo length, L(t) is the taboo length, which is dynamically adjusted according to the road congestion index δ ij (t) of the corresponding edge, and is used to avoid falling into local optimum, L0 is the basic taboo length, L(t) is the taboo length, which is dynamically adjusted according to the road congestion index δ ij (t) of the corresponding edge, and is used to avoid falling into local optimum, L0 is the basic taboo length, L(t) is the taboo length, which is dynamically adjusted according to the road congestion index δ ij (t) of the corresponding edge, and is used to avoid falling into local optimum, L0 is the basic taboo length, L(t) is the taboo length, which is dynamically adjusted according to the road congestion index δ ij (t) of the corresponding edge, and is used to avoid falling into local optimum, L0 is the basic taboo length, L(t) is the taboo length, which is dynamically adjusted according to the road congestion index δ ij (t) of the corresponding edge, and is used to avoid falling into local optimum,

Images