CN-122022247-A - Logistics carrier self-adaptive configuration method based on LightGBM and logistics data hybrid driving

CN122022247ACN 122022247 ACN122022247 ACN 122022247ACN-122022247-A

Abstract

The invention relates to the field of intelligent manufacturing and warehouse logistics, and discloses a logistics carrier self-adaptive configuration method based on LightGBM and logistics data mixed driving, which comprises the following steps of preprocessing multi-source heterogeneous data to form a demand sequence and a characteristic data set, and constructing a LightGBM-based demand prediction model to obtain prediction data; based on prediction data, a multi-objective optimization decision model is built, NSGA-II is adopted for optimization solution, the minimum physical space required by particle swarm optimization algorithm is calculated, the optimal storage layout is obtained, iteration test is carried out through a discrete event simulation model, AGV fleet configuration is obtained, flow bottleneck is identified through constraint theory, optimal goods picking port configuration is obtained, discrete event simulation and iteration optimization are carried out, and optimal self-adaptive configuration meeting requirements is obtained.

Inventors

SHEN JIAJING
ZHONG ZHIYANG
WANG JIN
ZHANG HAIYUN
LI XIAOFEI
LU GUODONG

Assignees

余姚市机器人研究中心
浙江大学

Dates

Publication Date: 20260512
Application Date: 20251222

Claims (9)

1. A logistics carrier self-adaptive configuration method based on LightGBM and logistics data hybrid driving is characterized by comprising the following steps: step 1, preprocessing multi-source heterogeneous data to form a demand sequence and a characteristic data set, and constructing a LightGBM-based demand prediction model to obtain prediction data; Step 2, constructing a multi-objective optimization decision model based on the prediction data, and carrying out optimization solution by adopting NSGA-II; Step 3, calculating a required minimum physical space through a particle swarm optimization algorithm to obtain an optimal storage layout; Step 4, performing iterative test through a discrete event simulation model to obtain AGV fleet configuration; step 5, identifying flow bottlenecks through constraint theory to obtain optimal goods picking port configuration; And 6, carrying out simulation and iterative optimization based on the optimal warehouse layout, the AGV fleet configuration and the optimal picking port configuration to obtain the optimal self-adaptive configuration meeting the requirements.
2. The method is characterized in that step 1 specifically comprises the steps of collecting multi-source heterogeneous data, including historical order data, a product bill of materials, a production plan, physical characteristics of materials and equipment files, cleaning the multi-source heterogeneous data, processing missing values, eliminating outliers, reconstructing a dimension unified time sequence, forming a demand sequence and a characteristic dataset taking SKU-time period as granularity, dividing the dataset into a training set and a verification set when a demand prediction model is constructed, selecting sales volume data subjected to logarithmic transformation processing as target variables, performing model training by using LightGBM regression algorithm, automatically executing feature selection operation in a model training link, evaluating importance of each feature, screening out first key features with high importance ranking, evaluating performance for the model training, adjusting and optimizing key parameters of the demand prediction model based on evaluation results, predicting key information in a future period through the optimized model, and outputting the prediction data.
3. The method is characterized in that step 2 specifically comprises the steps of receiving the prediction data, constructing a mathematical planning model, configuring multi-objective modeling at a system level, taking warehouse scale parameters, shelf layout parameters, AGV models and quantity, picking ports and workstation quantity as decision variables, taking minimum equipment investment and operation cost, maximum system throughput and minimum order average response time as objective functions, taking warehouse building area, inventory capacity, equipment utilization rate and service level as constraint conditions, constructing a multi-objective optimization decision model configured by system level equipment and warehouse units, and adopting NSGA-II to perform optimization solution.
4. The method for adaptively configuring the logistics carrier based on LightGBM and logistics data hybrid driving of claim 1, wherein step 3 is characterized in that a particle swarm optimization algorithm is adopted to encode and solve the number of rows and columns of a goods shelf, the channel position and the high-frequency area setting, the weighted sum of the average AGV driving distance and the goods shelf cost is taken as a fitness function, one particle represents a layout scheme, and the position and the speed of the particle are continuously adjusted by simulating the searching behavior of the particle in a solution space, so that the optimal storage layout is obtained.
5. The logistics carrier self-adaptive configuration method based on LightGBM and logistics data hybrid driving is characterized in that step 4 specifically comprises the steps of establishing a model selection rule base according to material weight, volume and carrying modes, determining candidate AGV types, constructing a discrete event simulation model to simulate order execution processes under different fleet scales, designing an adaptability function based on order completion time, AGV utilization rate and blocking rate indexes, searching for the number combination of each model by adopting the particle swarm optimization algorithm, encoding the number of each model into particle positions, calling the discrete event simulation model for each candidate solution, counting the number of completed tasks, average task completion time and AGV utilization rate in unit time, and comprehensively considering operation performance and cost to obtain the configuration of the AGV fleet.
6. The method for adaptively configuring the logistics carrier based on LightGBM and logistics data hybrid driving of claim 1, wherein step 5 is specifically implemented by identifying a bottleneck process through constraint theory, establishing a hybrid integer linear programming model taking a pick port opening state and order allocation as decision variables, solving by utilizing Gurobi, and optimizing the number and the position of the pick ports to obtain the optimal pick port configuration.
7. The method is characterized in that step 6 specifically comprises mapping the optimal warehouse layout, AGV fleet configuration, optimal goods-picking port configuration and order allocation scheme into Flexsim simulation environment to perform discrete event simulation, obtaining performance indexes including throughput, average response time, resource utilization rate and work-in-process in a target operation period, adjusting continuous parameters in the multi-target optimization decision model based on Bayesian optimization if the performance indexes do not meet a preset performance threshold, recommending new parameter combinations by Bayesian optimization in each iteration, simulating and evaluating performance and updating, and recalling the particle swarm optimization algorithm and discrete event simulation model to perform iterative calculation until the performance indexes meet requirements or reach an iteration upper limit.
8. The method for adaptively configuring a logistics carrier based on LightGBM and logistics data hybrid driving as set forth in claim 4, wherein the update formula of the particle speed and the position is: , , Wherein, the Is the first The particles are at the first The distribution of the bin bits at the time of the iteration, Is the first The velocity of the individual particles is such that, Is the first The optimal solution over the history of the individual particles, For a globally optimal solution over the population history, As the weight of the inertia is given, 、 In order for the learning factor to be a function of, 、 Is a random number ranging from 0 to 1.
9. The method for adaptively configuring a logistics carrier based on LightGBM and logistics data hybrid driving as set forth in claim 6, wherein the function of the mixed integer linear programming model is: , Wherein, the To pick up goods A decision variable of whether to turn on or not, To open the goods picking port Is used for the production of the high-density polyethylene, To pick up goods A cost factor for handling the unit load, To pick up goods The amount of tasks processed during the optimization cycle.

Description

Logistics carrier self-adaptive configuration method based on LightGBM and logistics data hybrid driving Technical Field The invention relates to the field of intelligent manufacturing and warehouse logistics, in particular to a logistics carrier self-adaptive configuration method based on LightGBM and logistics data hybrid driving. Background In modern production logistics systems, enterprises generally face the conditions of multiple varieties, small batches, shortened exchange period, enhanced demand fluctuation and the like. In warehouse and internal logistics systems, factors such as shelf size and layout, mobile robot AGV fleet configuration, number and location of pick ports, etc. have important effects on system throughput and response time. The traditional design method is usually based on experience and static calculation for one-time planning, once the production rhythm or order structure is changed, the system performance is easy to be obviously reduced, and quick reconstruction and self-adaptive configuration of equipment and storage units are difficult to realize. Conventional warehouse configurations typically rely on fixed rules and historical experience, but in the face of changing product requirements, order volume, warehouse equipment, etc., conventional approaches often have difficulty achieving quick response and efficient utilization of resources. In addition, the existing equipment configuration optimization work is usually only aimed at a certain local link, such as only optimizing the number of AGVs or only optimizing the layout of shelves, and the storage scale, the equipment configuration and the order structure are not unified in the same optimization frame, so that deviation exists between local optimization and global optimization, and the system is difficult to adjust in time when the demand changes. Disclosure of Invention Aiming at the technical problems that the prior art is difficult to realize quick response, high-efficiency utilization of resources, local and global deviation, difficulty in timely adjustment when the demand changes and the like, the invention provides a logistics carrier self-adaptive configuration method based on LightGBM and logistics data hybrid driving, which has the following technical scheme: step 1, preprocessing multi-source heterogeneous data to form a demand sequence and a characteristic data set, and constructing a LightGBM-based demand prediction model to obtain prediction data; Step 2, constructing a multi-objective optimization decision model based on the prediction data, and carrying out optimization solution by adopting NSGA-II; Step 3, calculating a required minimum physical space through a particle swarm optimization algorithm to obtain an optimal storage layout; Step 4, performing iterative test through a discrete event simulation model to obtain AGV fleet configuration; step 5, identifying flow bottlenecks through constraint theory to obtain optimal goods picking port configuration; And 6, carrying out simulation and iterative optimization based on the optimal warehouse layout, the AGV fleet configuration and the optimal picking port configuration to obtain the optimal self-adaptive configuration meeting the requirements. The method comprises the steps of (1) collecting multi-source heterogeneous data, including historical order data, a product bill of materials, a production plan, physical characteristics of materials and equipment files, cleaning the multi-source heterogeneous data, processing missing values, removing outliers, unifying dimensions and reconstructing time sequences to form a demand sequence and a characteristic data set taking SKU-time periods as granularity, dividing the data set into a training set and a verification set when constructing a demand prediction model, selecting sales volume data after logarithmic transformation processing as target variables, carrying out model training by using LightGBM regression algorithm, automatically executing feature selection operation in a model training link, evaluating the importance of each feature, screening out the first key features with high importance ranking, carrying out adjustment and optimization on key parameters of the demand prediction model based on evaluation results after model training is completed, predicting key information in a future period through the optimized model, and outputting prediction data. Further, the step 2 is specifically that the prediction data is received, a mathematical programming model is built, a multi-objective modeling is configured at a system level, storage scale parameters, shelf layout parameters, AGV model and number, a picking port and workstation number are used as decision variables, equipment investment and operation cost are minimum, system throughput is maximum and order average response time is minimum as objective functions, warehouse building area, inventory capacity, equipment utilization rate and service level are