A hub and regional warehouse network oriented cost and carbon emission collaborative optimization method

Abstract

This invention belongs to the field of smart logistics technology and discloses a method for cost and carbon emission collaborative optimization for hubs and regional warehouse networks. It constructs a three-dimensional global collaborative optimization model of cost, carbon emission, and timeliness, and adjusts target weights in real time through a dynamic weight adaptive mechanism. It integrates cost and carbon emission elements across the entire supply chain to establish a three-flow linkage, coupled with a multi-dimensional constraint system. This breaks down optimization barriers in warehousing and transportation, and dynamically adapts to regional carbon emission quotas and cargo urgency. While ensuring timely order fulfillment, it achieves the collaborative goal of minimizing overall cost and keeping carbon emissions within quota limits, balancing economic benefits and environmental responsibility, and enhancing the overall competitiveness of the supply chain. Through edge nodes, it achieves heterogeneous data cleaning and standardization, dynamically calibrates carbon emission factors and quantifies them into unified units, and completes distributed storage, access control, and full-chain traceability through a consortium blockchain, ultimately constructing a digital twin data foundation.

Description

Technical Field

[0001] This invention belongs to the field of smart logistics technology, specifically a method for cost and carbon emission synergistic optimization of hub and regional warehouse networks. Background Technology

[0002] With the accelerated digital transformation of global supply chains, smart logistics hubs and regional warehouse networks have become the core carriers of modern logistics systems, and their operational efficiency directly determines the overall competitiveness of the supply chain. Under the industry's demand for sustainable development, the logistics industry faces the dual pressures of overall operational optimization and carbon emission control. Existing hub and regional warehouse network optimization solutions still have the following technical challenges: First, most solutions only focus on cost control or carbon emission management in a single link, without establishing a coordination mechanism between cost and carbon emission across the entire chain, resulting in mutual constraints between the two and failing to achieve global synergy and optimality. Secondly, the data perception and processing system has obvious shortcomings. The standardization of heterogeneous data is low, and there is a lack of trusted sharing and security control mechanisms, making it difficult to support accurate global collaborative decision-making. Finally, the optimization model lacks dynamic adaptability and has a lagging emergency response mechanism, making it unable to accurately match the actual needs of regional supply and demand fluctuations, carbon emission quota adjustments, and sudden operational situations (such as equipment failures and extreme weather). Summary of the Invention

[0003] The purpose of this invention is to provide a method for cost and carbon emission synergistic optimization of hub and regional warehouse networks, in order to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a cost and carbon emission collaborative optimization method for hub and regional warehouse networks, including a trusted data standardization stage, a global collaborative modeling stage, an intelligent algorithm solving stage, a virtual and real simulation verification stage, a closed-loop execution iteration stage, and a collaborative emergency dispatch stage; Preferably, the trusted data standardization stage is based on the integration of IoT, edge computing and consortium blockchain technology to build a trusted data perception network for the entire link of hub and regional warehouse networks. By deploying sensors and edge data acquisition terminals on warehousing equipment, transport vehicles (including new energy vehicles), loading and unloading nodes and carbon emission monitoring points in hub warehouses and regional warehouses, the system collects multi-dimensional data in real time, such as energy consumption of warehousing operations, fuel consumption or electricity consumption of transport vehicles, cargo loading and unloading volume, warehouse capacity occupancy rate, road conditions of transportation routes, regional carbon emission quotas, cargo temperature control data, supply and demand forecast data of each warehouse network node and historical optimization scheme execution feedback data. Sensor deployment rules: ① The storage equipment is equipped with current sensors and displacement sensors, with one sensor for each piece of equipment. The current sensor is used to monitor energy consumption, and the displacement sensor is used to monitor the operating status. ②The transport vehicles are equipped with GPS positioning sensors, fuel consumption / power sensors, and temperature and humidity sensors (specifically for cold chain vehicles), with one set of each sensor deployed on each vehicle; ③ Deploy weight sensors and infrared sensors at loading and unloading nodes, with two sensors deployed at each loading and unloading station, one at the inlet and one at the outlet; the weight sensors are used to monitor the loading and unloading volume; the infrared sensors are used to monitor the operation efficiency. ④ Gas sensors are deployed at carbon emission monitoring points: one per 500㎡ in hub warehouses, one per 300㎡ in regional warehouses, and one per 20km in core transportation routes; the gas sensors are used to monitor CO2 concentration.

[0005] The heterogeneous data collected is cleaned, deduplicated, and standardized in real time through edge nodes. The carbon emission factors are calibrated in real time using a dynamic factor decomposition model, and the carbon emission data is quantified into a unified unit with the cost data. The dynamic factor decomposition model includes a data preprocessing module, a factor extraction module, and a calibration output module. The inputs are historical carbon emission data, regional energy structure data, equipment energy consumption parameters, and ambient temperature / humidity data. The output is the calibrated carbon emission factor. The core parameters include the number of factors k=3~5, the iteration step size 0.001, and the convergence threshold 1e-5. The number of factors k=3~5 (determined based on the clustering of carbon emission influencing factors in the logistics scenario). The calibration steps are as follows: ① The input data is normalized by the preprocessing module; ② The factor extraction module uses the EM algorithm to decompose the carbon emission influencing factors; ③ The factor weights are corrected by combining real-time data from regional carbon emission monitoring points, and finally the accurate carbon emission factors are output.

[0006] The standardized core data is distributed and stored through consortium blockchain nodes. Smart contracts are used to control data access permissions and enable trusted sharing across nodes. Data traceability is achieved through timestamps and spatial tags. Finally, a digital twin data base for hubs and regional warehouse networks is built based on standardized and trusted data to achieve real-time synchronization of data between physical entities and virtual models.

[0007] Preferably, the global collaborative modeling stage is based on the trusted data output from the trusted data standardization stage, and constructs a three-dimensional global collaborative optimization model for cost, carbon emissions and timeliness for hubs and regional warehouse networks, with the three-dimensional optimization objectives being the lowest cost across the entire chain, total carbon emissions not exceeding the quota, and order fulfillment timeliness meeting the standards. Through a dynamic weight adaptive mechanism, the weight coefficients of three objectives are adjusted in real time based on the tightness of regional carbon emission quotas, the urgency of goods, and market supply and demand fluctuations. At the same time, cost elements such as warehouse construction costs, transportation costs, inventory holding costs, and sorting costs, as well as carbon emission elements such as transportation carbon emissions, warehouse energy consumption carbon emissions, and loading and unloading equipment carbon emissions, are integrated to establish a linkage equation between carbon flow, cargo flow, and capital flow, and to quantify the dynamic relationship between carbon emissions and cargo flow and cost consumption. The mathematical expression for the three-dimensional optimization objective is: ①minC=∑(Ci+Ct+Cs+Cr), where Ci=warehousing cost, Ct=transportation cost, Cs=sorting cost, and Cr=inventory cost; ②maxE≤E0, where E is the actual total carbon emissions and E0 is the regional carbon emission quota; ③maxT≥T0, where T is the actual performance time and T0 is the preset time threshold; The three-flow linkage equation is: E=λ1×Q×∑(di×ei)+λ2×∑(Pi×ti), where λ1 is the conversion coefficient between cargo flow and carbon flow, λ2 is the conversion coefficient between capital flow and carbon flow, Q is the total cargo volume, di is the transportation distance, ei is the carbon emission coefficient per unit distance, Pi is the equipment power, and ti is the equipment operating time; the dynamic weight coefficients satisfy ωC+ωE+ωT=1, where ωC, ωE, and ωT are the weights of cost, carbon emission, and timeliness, respectively, and their values ​​range from 0.2 to 0.6.

[0008] A three-dimensional global collaborative optimization model is formed by setting multi-dimensional constraints, including warehouse capacity constraints, vehicle carrying capacity constraints, supply and demand balance constraints, dynamic carbon emission quota constraints, priority scheduling constraints for new energy vehicles, route safety constraints under extreme weather conditions, and temperature control energy consumption constraints for cold chain goods.

[0009] Methods for determining each constraint threshold: ① The warehouse capacity constraint is dynamically capped at 85% to 90% of the average daily inventory peak over the past three months, and is updated weekly based on weekly supply and demand forecast data; ② Vehicle carrying capacity constraints are set according to vehicle type: light vehicles ≤3 tons, medium vehicles 3~8 tons, heavy vehicles >8 tons, while reserving 10%~15% capacity redundancy to cope with sudden demand; ③ The supply and demand balance constraint requires that the order fulfillment rate of each regional warehouse be ≥90%, and the replenishment response time of the hub warehouse to the regional warehouse be ≤48 hours; ④ Energy consumption constraints for temperature control of cold chain goods: temperature control range fluctuation ≤ ±2℃, energy consumption not exceeding 30% of conventional transportation; ⑤ Extreme weather path safety constraints are triggered by an orange alert from the meteorological department, prohibiting the planning of transportation routes in dangerous sections such as flooded roads and areas with strong winds.

[0010] Preferably, the intelligent algorithm solution stage is designed for the nonlinear, multi-constraint, three-dimensional target global collaborative optimization model constructed in the global collaborative modeling stage. An improved NSGA-Ⅲ hybrid intelligent solution algorithm combining reinforcement learning and quantum annealing is designed (forming a hybrid intelligent solution system). The initial population that matches the real scene is generated by combining the actual operation data of the warehouse network and the virtual simulation data of the digital twin base, and the exploration and utilization mechanism of reinforcement learning is integrated simultaneously. The NSGA-III algorithm is used to achieve fast non-dominated sorting and crowding calculation for multiple objectives, and a preliminary Pareto optimal solution set is obtained. The quantum annealing algorithm is introduced to perform secondary optimization on the solution set to improve the global optimality of the solution set.

[0011] The scenario-based improvements are as follows: a warehouse network resource tension coefficient is introduced into the congestion calculation module. When the warehouse capacity occupancy rate of a certain area is ≥85% or the transportation vehicle capacity utilization rate is ≥90%, the coefficient is set to 0.8~1.0, amplifying the resource balance weight of the solution in this scenario. At the same time, the non-dominated ranking rules are optimized, and a ranking priority bonus is given to the solutions that meet the requirements of 'carbon emissions ≤ 95% of the quota and timeliness compliance rate ≥ 98%', ensuring that the core constraint objectives are met first.

[0012] The optimal Pareto solution set is comprehensively evaluated and ranked based on the entropy weight TOPSIS method. At the same time, the optimal reference solution suitable for the current operation scenario is selected by combining the reward mechanism of reinforcement learning. This includes cargo flow allocation scheme, transportation route planning scheme, and warehouse operation scheduling scheme.

[0013] Preferably, the virtual-real simulation verification stage constructs a virtual-real linkage digital twin simulation platform for hubs and regional warehouse networks, projects the optimal reference scheme output by the intelligent algorithm solution stage into the virtual twin space to achieve full-scene, multi-dimensional simulation verification, and restores all-scene elements such as warehouse network layout, equipment operation status, cargo flow process, transportation route, and carbon emission monitoring points based on the digital twin data base built in the trusted data standardization stage, realizing real-time data interaction between physical entities and virtual models; We will build an extreme scenario library that includes typical scenarios such as sudden surges in orders, transportation route congestion, tightening of carbon emission quotas, equipment failures, and extreme weather, and simulate the performance of the optimal solution in different scenarios, including cost consumption, total carbon emissions, operational efficiency, and timeliness compliance rate. Specific parameters for each typical scenario: ① Order surge: Order volume in a single period is 150% to 200% of that in a regular period, concentrated in 2 to 3 regional warehouses; ② Transportation route congestion: The traffic efficiency of the target route drops to 30%~50% of the normal level, and the congestion lasts for 2~4 hours; ③ Tightening of carbon emission quotas: Regional carbon emission quotas will be reduced by 20% to 30% compared to the baseline value, with a duration of 1 to 2 weeks; ④ Equipment failure: 5%~10% of the storage equipment or 2~3 transport vehicles in a single area warehouse cannot operate normally, and the failure lasts for 1~3 hours; ⑤ Extreme weather: Weather warnings for heavy rain, high temperatures, etc. will last for 4 to 8 hours, affecting 3 to 5 core transportation routes.

[0014] If the simulation results fail to meet the target, the results are fed back to the intelligent algorithm solution stage in real time through the digital twin platform, and the algorithm parameters are automatically adjusted and the solution is solved again; if all the targets are met, the solution is output to the closed-loop execution iteration stage.

[0015] Preferably, the closed-loop execution iteration stage inherits the optimal solution verified in the virtual and real simulation verification stage, and constructs a full-process closed-loop execution system with blockchain smart contracts, real-time monitoring, and dynamic evaluation. The optimal solution verified in the simulation is decomposed into specific control instructions, which are sent to automated equipment, transportation vehicle scheduling system, and warehouse management system through industrial Ethernet. At the same time, the instructions are automatically triggered and executed based on the preset rules of the blockchain smart contract. The trusted sensing network, built through the trusted data standardization phase, collects real-time data on equipment operating status, cargo flow progress, cost consumption, and carbon emissions. This data is then compared in real-time with the preset indicators of the optimization plan. When the deviation exceeds the preset threshold, the smart contract automatically triggers a local adjustment instruction. Establish a multi-dimensional dynamic evaluation system that includes economic, environmental, operational, and safety dimensions. Set the evaluation cycle (e.g., daily / weekly) according to the characteristics of the operational scenario. Use the analytic hierarchy process (AHP) combined with reinforcement learning feedback results to periodically evaluate the optimization effect. Feedback the evaluation results to the global collaborative modeling stage and update the objective function weight coefficients and constraint parameters in real time to achieve adaptive iteration of the optimization model.

[0016] Preferably, the collaborative emergency dispatch phase constructs a distributed collaborative emergency dispatch system based on a blockchain consortium chain and multiple intelligent agents. The blockchain consortium chain enables data sharing and trusted collaboration among multiple entities such as hub warehouses, regional warehouses, transportation companies, and carbon trading platforms, ensuring real-time synchronization and reliable interaction of cargo flow information, inventory information, carbon emission information, and emergency resource information. Deploy a multi-agent model to simulate the operational status and emergency needs of each warehouse network node. When an emergency occurs, the multi-agent collaboratively assesses the emergency needs and automatically calls the hybrid intelligent algorithm in the intelligent algorithm solution stage to generate an emergency optimization plan. Each regional warehouse agent reports node status data to the main agent of the hub warehouse every 10 seconds, and reports abnormal situations such as stockouts and malfunctions in real time. After receiving the information, the main agent forms an evaluation group with three adjacent regional warehouse agents and determines the emergency demand level through a voting mechanism (≥2 votes in favor). Based on the demand level, the main agent initiates a scheduling request to the regional warehouse agents with redundant resources. The requested agent should provide a response within 1 minute. If no response is provided within the time limit, the next candidate node will be automatically matched.

[0017] After the emergency response is completed, the system automatically reverts to the regular optimized scheduling mode of the closed-loop execution iteration phase. At the same time, the emergency response data is uploaded to the consortium blockchain for storage and fed back to the dynamic evaluation system of the closed-loop execution iteration phase to update the extreme scenario library and model parameters.

[0018] The beneficial effects of this invention are as follows: 1. This invention constructs a three-dimensional global collaborative optimization model of cost, carbon emissions, and timeliness. It adjusts the target weight in real time through a dynamic weight adaptive mechanism, integrates the cost and carbon emission factors of the entire supply chain to establish a three-flow linkage, and is equipped with a multi-dimensional constraint system. It breaks down the optimization barriers in warehousing, transportation and other links, and can dynamically adapt to regional carbon emission quotas and the urgency of goods. Under the premise of ensuring that the order fulfillment timeliness meets the standards, it achieves the synergistic goal of minimizing the cost of the entire supply chain and keeping carbon emissions within the quota, thus balancing economic benefits and environmental responsibility and enhancing the overall competitiveness of the supply chain.

[0019] 2. This invention integrates IoT, edge computing, and consortium blockchain technologies to build a trusted data sensing network across the entire chain. It achieves heterogeneous data cleaning and standardization through edge nodes, dynamically calibrates carbon emission factors and quantifies them into unified units, and completes distributed evidence storage, access control, and full-chain traceability through consortium blockchain, ultimately constructing a digital twin data foundation. This foundation enables real-time synchronization of physical and virtual data, providing data support for global modeling and algorithm solving, and improving the accuracy and reliability of collaborative decision-making.

[0020] 3. This invention designs a hybrid intelligent solution algorithm to improve the efficiency and global optimality of the solution. It verifies the solution through virtual-real linkage simulation to cover extreme scenarios and ensure the feasibility of the solution. At the same time, it constructs a closed-loop execution iteration system to achieve adaptive model updates based on multi-dimensional evaluation. Combined with a distributed collaborative emergency system, it can quickly respond to sudden situations such as order surges and equipment failures, automatically generate emergency optimization solutions, and return to normal scheduling and update model parameters after handling. This improves the adaptability of the warehouse network to supply and demand fluctuations, carbon emission quota adjustments, and sudden situations, ensuring operational stability and robustness. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating the global optimization process of this invention; Figure 2This is a flowchart illustrating the trusted data standardization process of the present invention. Figure 3 This is a flowchart of the intelligent algorithm solution of the present invention; Figure 4 This is a flowchart of the closed-loop execution iteration process of the present invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] like Figures 1 to 4 As shown, this embodiment of the invention provides a method for cost and carbon emission collaborative optimization for hub and regional warehouse networks, including a trusted data standardization stage, a global collaborative modeling stage, an intelligent algorithm solution stage, a virtual and real simulation verification stage, a closed-loop execution iteration stage, and a collaborative emergency dispatch stage. The specific implementation of each stage is as follows: The trusted data standardization phase is based on the integration of IoT, edge computing and consortium blockchain technology to build a trusted data perception network for the entire link of hub and regional warehouse networks. By deploying sensors and edge data acquisition terminals on storage equipment, transport vehicles (including new energy vehicles), loading and unloading nodes and carbon emission monitoring points in hub warehouses and regional warehouses, it collects multi-dimensional data in real time, such as energy consumption of storage operations, fuel consumption or electricity consumption of transport vehicles, cargo loading and unloading volume, storage capacity occupancy rate, road conditions of transportation routes, regional carbon emission quotas, cargo temperature control data (for cold chain scenarios), as well as supply and demand forecast data and historical optimization scheme execution feedback data of each warehouse network node. The supply and demand forecast data is obtained by integrating historical order data from the past 6 months, regional industry prosperity data (such as peak sales season data in the retail industry), holiday arrangements, promotional activity forecasts, and other multi-source information, and using the gradient boosting tree (GBT) model for forecasting. In daily scenarios, the forecast data for the past 7 days is updated once a day, while in special scenarios such as promotions and holidays, the forecast data for the past 3 days is updated every 6 hours to ensure that the forecast results are consistent with the fluctuations in actual demand.

[0024] The heterogeneous data collected is cleaned, deduplicated, and standardized in real time through edge nodes. The carbon emission factors are calibrated in real time using a dynamic factor decomposition model, and the carbon emission data is quantified into a unified unit comparable to the cost data, such as yuan / ton CO2. The specific processing rules are as follows: ① The 3σ criterion is used to identify outliers. For example, if the energy consumption data exceeds the mean of the same scene by ±3 times the standard deviation, it is marked as an outlier and automatically replaced with the mean of the last 10 minutes. ②Use timestamp + collection node ID + data type as unique identifier, retain the first collection data, and remove duplicate records; ③ The unified data format is “YYYY-MM-DDHH:MM:SS-collection node-data type-value-unit”, where the energy consumption unit is uniformly kWh, the carbon emission data unit is uniformly kgCO2, and the distance unit is uniformly km, to ensure that cross-node data can be directly compared.

[0025] Standardized carbon emission data, cost data, and supply and demand data are distributed and stored through consortium blockchain nodes. Smart contracts are used to control data access permissions and ensure trusted sharing across nodes, avoiding data tampering and information asymmetry. Data traceability is achieved through timestamps and spatial tags. Finally, a digital twin data base for hubs and regional warehouse networks is built based on standardized and trusted data, enabling real-time synchronization of data between physical entities and virtual models.

[0026] The real-time synchronization mechanism is as follows: the synchronization frequency is set according to the data type. Dynamic data such as energy consumption and carbon emissions are synchronized once every 1 minute, while static data such as storage capacity and inventory are synchronized once every 30 minutes. Data transmission adopts the MQTT protocol to ensure low-latency transmission. When data transmission is interrupted for more than 5 minutes, local cached data is automatically enabled to temporarily fill the gap. After the connection is restored, the cached data is synchronized first and the consistency is verified to avoid the virtual model from becoming disconnected from the physical entity.

[0027] The data storage adopts a hierarchical storage architecture. The real-time layer stores dynamic data such as energy consumption, carbon emissions, and road conditions for nearly one hour, and uses an in-memory database to ensure read speed. The near real-time layer stores operational data for nearly seven days, using distributed database storage. The historical layer stores historical data and optimization solutions for more than seven days, using low-cost object storage. The data is organized by a three-dimensional index of warehouse network nodes, data types, and timestamps, and supports fast retrieval and access by node, time, and indicator type.

[0028] The global collaborative modeling stage is based on the credible data output from the credible data standardization stage. It constructs a three-dimensional global collaborative optimization model for cost, carbon emissions, and timeliness for hubs and regional warehouse networks, with the three-dimensional optimization objectives being the lowest cost across the entire chain, total carbon emissions not exceeding the quota, and order fulfillment timeliness meeting the standards. The fast-moving consumer goods (FMCG) industry focuses on balancing cost and timeliness, with the weight of inventory holding costs reduced by 5% to 10% and the timeliness compliance threshold increased to 98%. The cold chain industry strengthens temperature control energy consumption constraints and timeliness weight, tightening the temperature control fluctuation threshold to ±1℃ and increasing the timeliness weight by 10% to 15% compared to conventional industries. The manufacturing bulk cargo industry focuses on carbon emissions and transportation cost optimization, with the priority scheduling weight of new energy vehicles increased by 20% and the carbon emission quota utilization threshold set at ≥90%. The cross-border logistics industry adds customs clearance timeliness constraints (clearance time ≤48 hours) and prioritizes routes with dense customs hub nodes in route planning.

[0029] The dynamic weight adaptive mechanism adjusts the weight coefficients of the three objectives in real time based on the tightness of regional carbon emission quotas, the urgency of goods, and market supply and demand fluctuations. For example, when carbon emission quotas are tightened, the weight of carbon emission control is increased, and the weight of timeliness is increased when transporting fresh goods. The specific adjustment logic is as follows: ① The tightness of carbon emission quotas is defined by the quota surplus rate. A surplus rate of ≥60% is considered loose, with ωE of 0.2~0.3; a surplus rate of 30%~60% is considered moderate, with ωE of 0.3~0.4; and a surplus rate of <30% is considered tight, with ωE of 0.4~0.6. ② The urgency of goods is determined by the remaining time of the performance period. If the remaining time is ≤24 hours, it is extremely urgent, and ωT is 0.4~0.6; if the remaining time is 24~72 hours, it is urgent, and ωT is 0.3~0.4; if the remaining time is >72 hours, it is normal, and ωT is 0.2~0.3. ③ Market supply and demand fluctuations are defined by the order fulfillment rate. A fulfillment rate of ≥95% indicates supply and demand balance, with ωC ranging from 0.3 to 0.4; a fulfillment rate of 85% to 95% indicates supply and demand tension, with ωC ranging from 0.4 to 0.5; and a fulfillment rate of <85% indicates supply and demand imbalance, with ωC ranging from 0.5 to 0.6.

[0030] Simultaneously, it integrates cost elements such as warehouse construction costs, transportation costs, inventory holding costs, and sorting costs, as well as carbon emission elements such as transportation carbon emissions, warehouse energy consumption carbon emissions, and loading and unloading equipment carbon emissions, to establish a three-flow linkage equation for carbon flow, cargo flow, and capital flow, and quantify the dynamic relationship between carbon emissions and costs. A three-dimensional global collaborative optimization model is formed by setting up multi-dimensional constraints, including warehouse capacity constraints, vehicle carrying capacity constraints, supply and demand balance constraints, dynamic carbon emission quota constraints (combined with real-time price adjustments in the regional carbon trading market), priority scheduling constraints for new energy vehicles, route safety constraints under extreme weather conditions, and temperature control energy consumption constraints for cold chain goods.

[0031] The specific adjustment criteria are as follows: when the regional carbon trading market price increases by ≥20% compared to the benchmark price, the carbon emission quota utilization threshold will be lowered by 5%~8% to encourage the reduction of carbon emissions to obtain trading revenue; when the price falls by ≥20%, the threshold will be raised by 3%~5% to appropriately relax carbon emission constraints to balance costs; when the price fluctuation is within ±20%, the threshold will remain unchanged to ensure the stability of the constraints.

[0032] In the intelligent algorithm solution stage, for the nonlinear, multi-constraint, three-dimensional target global collaborative optimization model constructed in the global collaborative modeling stage, an improved NSGA-Ⅲ hybrid intelligent solution algorithm combining reinforcement learning and quantum annealing is designed (forming a hybrid intelligent solution system) to improve solution efficiency and solution robustness. The initial population that matches the real scene is generated by combining the actual operation data of the warehouse network and the virtual simulation data of the digital twin base, and the exploration and utilization mechanism of reinforcement learning is integrated in time to avoid the initial population from getting trapped in local optima. The improvements of the improved NSGA-Ⅲ are as follows: the congestion calculation module is optimized, and a warehouse network scenario adaptation coefficient (with a value of 0.1~0.3) is introduced to adjust the sorting priority; the state space S of the reinforcement learning module is {warehouse network capacity occupancy rate, remaining carbon emission quota, remaining order time}, the action space A is {path adjustment, inventory allocation, equipment scheduling}, and the reward function R = 0.4 × cost saving rate + 0.3 × carbon emission reduction rate + 0.3 × timeliness compliance rate; The core parameters of the quantum annealing algorithm are: initial temperature T0=100, cooling rate α=0.95, and termination temperature Tf=1. The fusion mechanism of the three is a serial linkage. First, an initial solution set is generated by NSGA-Ⅲ, then the quantum annealing algorithm optimizes the global optimality of the solution set, and finally the reinforcement learning reward mechanism is used to select the appropriate solution. The fusion cycle is synchronized with the warehouse network data acquisition cycle, once every 30 minutes.

[0033] The NSGA-III algorithm is used to achieve fast non-dominated sorting and crowding calculation for multiple objectives, and a preliminary Pareto optimal solution set is obtained. To address the local optima problem of the solution set, the quantum annealing algorithm is introduced to perform secondary optimization on the solution set to improve the global optimality of the solution set.

[0034] Iteration termination conditions: ① The maximum number of iterations is reached, which is set to 200 for normal scenarios and 100 for emergency scenarios; ② No new solutions are generated in the Pareto optimal solution set after 20 consecutive iterations; ③ The fluctuation range of the objective function value is ≤1%, and the cost, carbon emission, and time efficiency indicators must all be met; If any condition is met, the iteration stops and the current optimal solution set is output.

[0035] The optimized Pareto optimal solution set is comprehensively evaluated and ranked based on the entropy weight TOPSIS method. The scenario-based application logic is as follows: ① The objective weight of each evaluation indicator is calculated using the entropy weight method, with cost indicators (total cost, unit cost of goods) accounting for 35%~40%, carbon emission indicators (total carbon emissions, carbon intensity) accounting for 30%~35%, and timeliness indicators (compliance timeliness, compliance rate) accounting for 25%~30%; ② The proximity is used as the core judgment criterion. The closer the proximity is to 1, the better the solution. Solutions with a proximity of ≥0.85 are selected first. If there are multiple solutions that meet the conditions, the carbon emission reduction rate is further compared, and the solution with better carbon emission is selected.

[0036] At the same time, by combining the reward mechanism of reinforcement learning, the optimal reference solution suitable for the current operation scenario is selected with the reward indicators of cost saving rate, carbon emission reduction rate and timeliness compliance rate. This includes cargo flow allocation solution, transportation route planning solution and warehousing operation scheduling solution. In the transportation route planning solution, the new energy vehicle route is recommended first based on the priority scheduling constraint of new energy vehicles.

[0037] The implementation rules are as follows: Priority will be given to planning transportation routes within the range of new energy vehicles, with pure electric vehicles having a range of ≤200km and hybrid vehicles having a range of ≤500km. If the range exceeds the range, intermediate charging stations will be planned. When the available capacity of new energy vehicles is ≥60% of the required capacity, all new energy vehicles will be used. When it is less than 60%, new energy vehicles will be prioritized for transportation tasks in carbon emission sensitive areas (carbon emission quota remaining rate <30%). The scheduling priority of new energy vehicles is higher than that of fuel vehicles. Under the same conditions, the order acceptance response time of new energy vehicles is ≤3 minutes, and that of fuel vehicles is ≤5 minutes.

[0038] To ensure the feasibility and robustness of the output solution in the intelligent algorithm solution-solving stage, the virtual-real simulation verification stage constructs a virtual-real linkage digital twin simulation platform for hubs and regional warehouse networks. The optimal reference solution output in the intelligent algorithm solution-solving stage is projected into the virtual twin space to achieve full-scene, multi-dimensional simulation verification. Based on the digital twin data base built in the trusted data standardization stage, the platform restores all scene elements such as warehouse network layout, equipment operating status, cargo flow process, transportation route, and carbon emission monitoring points, realizing real-time data interaction between physical entities and virtual models. Simulation platform operating parameters: Normal scenarios run at 1 minute / step, and emergency scenarios are simulated at 10 seconds / step with finer detail; core indicator data is collected once per simulation step to generate continuous time-series curves; in addition to the core indicators, auxiliary judgment criteria include monitoring resource utilization (warehousing equipment, transportation vehicles), scheduling response time (from instruction issuance to execution completion), and data consistency (deviation between virtual and physical data). All three must meet the following conditions for simulation verification to be considered successful: resource utilization ≤ 90%, scheduling response time ≤ 5 minutes, and data consistency deviation ≤ 2%.

[0039] The consistency verification mechanism is as follows: every 30 minutes, core data such as energy consumption, carbon emissions, and inventory are sampled and verified, with a sampling ratio of 10% to 15%; when the difference between physical data and virtual model data is ≤2%, it is judged to be consistent; when the difference is 2% to 5%, the virtual model is automatically calibrated using physical data; when the difference is >5%, an alarm is triggered and the execution of relevant optimization schemes is suspended, and manual intervention is required for verification; the verification priority of carbon emission data and timeliness-related data is higher than that of cost data to ensure the data reliability of core optimization targets.

[0040] We will build an extreme scenario library that includes typical scenarios such as sudden surges in orders, transportation route congestion, tightening of carbon emission quotas, equipment failure, and extreme weather (heavy rain, high temperature) to simulate the performance of the optimal solution in different scenarios, including cost consumption, total carbon emissions, operational efficiency, and timeliness compliance rate. If the simulation results fail to meet the standards, such as excessive carbon emissions or time delays, the results are fed back to the intelligent algorithm solution stage in real time through the digital twin platform, automatically adjusting the algorithm parameters and resolving the problem. If all the indicators meet the standards, the solution is output to the closed-loop execution iteration stage to ensure the reliability of the implemented solution.

[0041] The closed-loop execution iteration stage inherits the optimal solution verified in the virtual and real simulation verification stage, and constructs a full-process closed-loop execution system with blockchain smart contracts, real-time monitoring, and dynamic evaluation. The optimal solution verified in the simulation is decomposed into specific control instructions, which are sent to automated equipment, transportation vehicle scheduling system, and warehouse management system through industrial Ethernet. At the same time, based on the preset rules of the blockchain smart contract, such as carbon emission exceedance threshold, cost overrun threshold, and time delay threshold, the instructions are automatically triggered and executed. Smart contract execution process: ① After the optimal solution is broken down into control instructions, the smart contract automatically associates them with the corresponding execution nodes. For example, warehousing instructions are associated with the warehousing management system, and transportation instructions are associated with the vehicle dispatching system. ② Triggering conditions: Meet preset thresholds, such as carbon emissions exceeding limits or time delays; or trigger at a set time, such as executing an inventory count instruction at 3:00 AM every day; ③ After the execution node completes the instruction, it uploads the execution status and result data to the blockchain in real time. Once the smart contract verifies that the result is correct, it marks the execution as completed. Hub warehouse administrators have the authority to modify contract rules, which requires confirmation from more than 2 / 3 of the regional warehouse nodes. Regional warehouse administrators only have the authority to view and execute instructions, and transportation companies can only access route planning instructions that are relevant to themselves.

[0042] The trusted sensing network, built through the trusted data standardization phase, collects real-time data on equipment operation status, cargo flow progress, cost consumption and carbon emissions, and compares it with the preset indicators of the optimization plan in real time. When the deviation exceeds the preset threshold, the smart contract automatically triggers local adjustment instructions, such as switching to new energy transportation routes, adjusting the priority of warehousing operations, and allocating emergency inventory in regional warehouses. Deviation threshold setting rules: ① In the economic dimension, the total cost overrun threshold is ≤5%, and the unit cost fluctuation threshold is ≤3%; ② Environmental dimension: actual carbon emissions exceeding the quota threshold by ≤3%, and carbon intensity fluctuation threshold by ≤2%; ③ From an operational perspective, the delivery delay threshold is ≤2 hours, and the order fulfillment rate is ≥90% below the threshold; ④ In terms of security, the threshold for path security event occurrence rate is ≤1%, and the threshold for equipment failure rate is ≤3%; The threshold can be finely adjusted within the range of ±1% to ±2% depending on industry characteristics (e.g., cold chain, fast-moving consumer goods).

[0043] Establish a multi-dimensional dynamic evaluation system that includes economic, environmental, operational, and safety dimensions. The economic dimension includes total cost reduction rate, unit cargo cost, and return on investment; the environmental dimension includes carbon emission reduction rate, carbon intensity, and carbon quota utilization rate; the operational dimension includes order fulfillment rate, delivery timeliness compliance rate, and equipment utilization rate; and the safety dimension includes route safety rate and equipment failure rate. The evaluation cycle (e.g., daily / weekly) is set according to the characteristics of the operational scenario. The optimization effect is periodically evaluated by using the analytic hierarchy process combined with the feedback results of reinforcement learning. The evaluation results are fed back to the global collaborative modeling stage, and the objective function weight coefficients and constraint parameters are updated in real time to achieve adaptive iteration of the optimization model. At the same time, it provides normalized operational data support for the collaborative emergency dispatch stage.

[0044] Specific evaluation process: ① Summarize the four core data categories of cost, carbon emissions, timeliness, and safety within the evaluation period, and remove outliers; ② Score according to four levels: "Excellent (90-100 points), Good (80-89 points), Pass (70-79 points), and Unsatisfactory (<70 points)," with the scoring criteria linked to preset thresholds; ③ Calculate the comprehensive score based on the weights of 35% for the economic dimension, 30% for the environmental dimension, 25% for the operational dimension, and 10% for the safety dimension; ④ A comprehensive score ≥80 points indicates good fit, maintaining the current model parameters; 60-79 points indicates basic fit, with fine-tuning of weight coefficients; <60 points indicates poor fit, triggering model reconstruction and a full parameter update.

[0045] In response to potential emergencies during routine operations and the cross-node collaboration needs of hub and regional warehouse networks, the collaborative emergency dispatch phase constructs a distributed collaborative emergency dispatch system based on a blockchain consortium chain and multiple intelligent agents. This system enables data sharing and trusted collaboration among multiple entities, including hub warehouses, regional warehouses, transportation companies, and carbon trading platforms, ensuring real-time synchronization and reliable interaction of cargo flow information, inventory information, carbon emission information, and emergency resource information. Deploy a multi-agent model to simulate the operational status and emergency needs of each warehouse network node. When emergencies occur, such as regional warehouse stockouts, equipment failures, transportation disruptions caused by extreme weather, or sudden adjustments to carbon emission quotas, the multi-agent model collaboratively assesses the emergency needs and automatically calls the hybrid intelligent algorithm in the intelligent algorithm solution stage to generate emergency optimization solutions, such as emergency replenishment route planning for hub warehouses, cross-regional warehouse resource allocation, new energy vehicle substitution scheduling, and temporary carbon emission quota adjustment schemes (adapted to the dynamic constraints of regional carbon emission quotas). The implementation priority of the plan is as follows: ① Emergency medicines, cold chain vaccines and other life-saving goods take priority over ordinary goods and have the highest implementation priority; ② Emergency plans for carbon emission exceeding the limit, such as sudden quota reduction, take priority over plans for cost overruns; ③ Plans for route interruption take priority over plans for equipment failure; ④ Cross-regional allocation plans take priority over single-node internal adjustment plans; the priority is ranked from "1 to 4", with level 1 to be implemented immediately and level 4 to be implemented as appropriate without affecting the core indicators.

[0046] The multi-agent model comprises N+1 agents, where N is the number of regional warehouses and one hub warehouse master agent. The master agent is responsible for summarizing global emergency needs and distributing solutions, while the regional warehouse agents are responsible for collecting node operation data and responding to local emergencies. The communication mechanism uses a blockchain consortium chain P2P communication, with the data transmission protocol in JSON format and a communication frequency of 10 seconds per communication. The input data includes the remaining inventory of warehouse network nodes, equipment operating status, carbon emission quota remaining amount, and available emergency resources. The output data includes the emergency need level (levels 1-5), the quantified value of resource gap, and the priority scheduling direction, providing emergency scenario input parameters for the hybrid intelligent solution algorithm.

[0047] The specific criteria for classifying emergency demand levels are as follows: ① Level 1 (Minor): Localized failure of a single node, affecting less than 5% of orders, no cross-node resource allocation required; ② Level 2 (General): Resource shortage in a single regional warehouse is less than 10%, which can be resolved by short-distance allocation from surrounding regional warehouses; ③ Level 3 (Severe): 1-2 regional warehouses have a resource shortage of 10%-20%, requiring emergency replenishment from hub warehouses, with no route interruption; ④ Level 4 (Severe): Resource shortage of 3 or more regional warehouses > 20%, or interruption of 1 core transportation route, requiring cross-regional resource allocation; ⑤ Level 5 (Extremely Severe): Hub warehouse failure or interruption of multiple core routes, affecting more than 50% of orders, requiring the activation of emergency reserve resources and alternative transportation networks.

[0048] After the emergency response is completed, the system automatically reverts to the regular optimization scheduling mode of the closed-loop execution iteration phase. At the same time, the emergency response data is uploaded to the consortium blockchain for storage and fed back to the dynamic evaluation system of the closed-loop execution iteration phase. This updates the extreme scenario library and model parameters, further improving the emergency adaptability and robustness of the entire optimization system.

[0049] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0050] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for cost and carbon emission co-optimization of hub and regional warehouse networks, characterized in that, It includes the trusted data standardization stage, the global collaborative modeling stage, the intelligent algorithm solving stage, the virtual and real simulation verification stage, the closed-loop execution iteration stage, and the collaborative emergency dispatch stage; Trusted data standardization stage: Based on the integration of IoT, edge computing and consortium blockchain technologies, a full-node sensing network is built to collect multi-dimensional operation and environmental data. After processing, heterogeneous data standardization and dynamic calibration of carbon emission factors are achieved. A digital twin data foundation is built through distributed storage and trusted sharing. Global collaborative modeling stage: Based on reliable data, a three-dimensional global collaborative optimization model of cost, carbon emission and timeliness is constructed. The cost and carbon emission factors of the whole link are integrated to establish a three-flow linkage mechanism. A dynamic weight adaptive mechanism and a multi-scenario constraint system are embedded to form a global optimization model that adapts to the characteristics of different industries. Intelligent algorithm solution stage: A hybrid intelligent solution system is designed for the global optimization model. It integrates actual operation and virtual simulation data to generate an initial population. Through step-by-step optimization, it achieves non-dominated solution screening and global optimality improvement. Combined with multi-index evaluation, it selects the optimal solution that is suitable for different scenarios. Virtual-real simulation verification phase: Construct a virtual-real linkage simulation platform, execute typical scenario solution verification based on digital twin data base, and adjust and optimize parameters through closed-loop feedback mechanism; Closed-loop execution iteration phase: Construct a full-process closed-loop execution system to realize the issuance of instructions for the optimal solution and real-time monitoring, link the sensing network to collect data, optimize the model through multi-dimensional evaluation and feedback, and achieve adaptive iterative updates; Collaborative emergency dispatch phase: Construct a distributed collaborative emergency system to achieve multi-entity data sharing and emergency demand assessment, generate emergency optimization plans, and return to regular dispatch after the disposal is completed.

2. The cost and carbon emission co-optimization method for hub and regional warehouse networks according to claim 1, characterized in that, In the trusted data standardization phase, the sensing network covers warehousing equipment, transport vehicles, loading and unloading nodes, and carbon emission monitoring points in hub warehouses and regional warehouses. The collected data includes warehousing operation energy consumption, transportation energy consumption, cargo loading and unloading volume, warehouse capacity occupancy rate, road condition information, carbon emission quotas, cargo temperature control data, supply and demand forecasts, and historical execution feedback data. Edge nodes perform real-time cleaning, deduplication, and format standardization on the collected heterogeneous data. Carbon emission factors are calibrated in real-time and carbon emission data is quantified through a dynamic factor decomposition model, and data units are unified. Standardized core data is distributed and notarized through a consortium blockchain. Access permissions are controlled through smart contracts and trusted sharing across nodes is achieved. Combined with a traceability mechanism, data reliability is ensured. The final digital twin data base realizes real-time data synchronization between physical entities and virtual models.

3. The method for coordinated optimization of cost and carbon emissions for hub and regional warehouse networks according to claim 2, characterized in that, The global collaborative modeling phase takes the lowest end-to-end cost, total carbon emissions not exceeding the quota, and order fulfillment timeliness as the three-dimensional optimization objectives. The dynamic weight adaptive mechanism adjusts the target weight coefficients according to the tightness of regional carbon emission quotas, the urgency of goods, and market supply and demand fluctuations. It integrates cost elements including warehousing, transportation, inventory, and sorting, as well as carbon emission elements including transportation and warehousing energy consumption and loading and unloading equipment, to establish a three-flow linkage equation for carbon flow, goods flow, and capital flow. The constraint system includes warehouse capacity, vehicle carrying capacity, supply and demand balance, dynamic constraints on carbon emission quotas, and special constraints adapted to special goods and extreme working conditions.

4. The cost and carbon emission co-optimization method for hub and regional warehouse networks according to claim 3, characterized in that, The hybrid intelligent solution system in the intelligent algorithm solution stage integrates multi-objective optimization, reinforcement learning, and quantum annealing techniques. The initial population generation process fully matches the characteristics of real operation scenarios. Through step-by-step optimization, the non-dominated solutions are first rapidly screened, and then the global optimality of the solution set is improved through quantum annealing algorithm optimization. The multi-index evaluation mechanism combines comprehensive ranking and reward feedback logic to select the optimal reference solution that includes cargo flow allocation, transportation route planning, and warehouse operation scheduling.

5. The cost and carbon emission co-optimization method for hub and regional warehouse networks according to claim 4, characterized in that, The virtual-real simulation verification stage uses a simulation platform that recreates all elements of the scenario, including warehouse network layout, equipment operation, cargo flow, and transportation routes. The typical scenario library includes typical scenarios such as order surges, route congestion, tightening carbon emission quotas, equipment failures, and extreme weather. The simulation process monitors core indicators such as cost consumption, total carbon emissions, operational efficiency, and timeliness compliance rate. If an indicator fails to meet the standard, it is fed back to the intelligent algorithm solution stage through a closed loop for re-solution. If the standard is met, it is output to the closed-loop execution iteration stage.

6. The cost and carbon emission co-optimization method for hub and regional warehouse networks according to claim 5, characterized in that, The closed-loop execution system in the closed-loop execution iteration phase integrates blockchain smart contracts, real-time monitoring, and dynamic evaluation functions. It breaks down the optimal solution into control commands and sends them to each execution system. The smart contracts enable automatic triggering and execution of these commands. Based on a trusted sensing network, it collects actual data on equipment operating status, cargo flow progress, cost consumption, and carbon emissions, and compares this data with preset indicators in real time. When the deviation exceeds a threshold, it automatically triggers local adjustments. The multi-dimensional evaluation system covers economic, environmental, operational, and safety dimensions. The evaluation results are fed back to the global collaborative modeling phase to achieve real-time updates of the objective function weights and constraint parameters.

7. The method for coordinated optimization of cost and carbon emissions for hub and regional warehouse networks according to claim 6, characterized in that, The distributed collaborative emergency system in the collaborative emergency dispatch phase is based on a blockchain consortium chain to achieve multi-entity data sharing and trust collaboration, ensuring real-time synchronization of information on cargo flow, inventory, carbon emissions, and emergency resources; it simulates the operational status and emergency needs of warehouse network nodes through a multi-agent model, and in the event of an emergency, it calls the hybrid intelligent solution system in the intelligent algorithm solution phase to generate an emergency optimization plan. After the emergency response is completed, the system returns to the regular dispatch mode. Emergency data is uploaded to the alliance blockchain for storage and fed back to the dynamic evaluation system to update the extreme scenario library and model parameters.

8. The cost and carbon emission co-optimization method for hub and regional warehouse networks according to claim 7, characterized in that, The digital twin data base adopts a hierarchical storage architecture, storing different time-sensitive data in categories such as real-time layer, near-real-time layer, and historical layer.

9. The cost and carbon emission co-optimization method for hub and regional warehouse networks according to claim 8, characterized in that, The virtual-real simulation verification stage is equipped with a data consistency verification mechanism, which automatically calibrates the virtual model when the deviation between physical and virtual data exceeds a threshold.

10. The cost and carbon emission co-optimization method for hub and regional warehouse networks according to claim 9, characterized in that, The collaborative emergency dispatch phase is divided into 1-5 levels of emergency demand based on the scope of impact and resource gaps.

Images