Supply chain oriented erp resource dynamic allocation optimization method

Abstract

The application discloses an ERP resource dynamic allocation optimization method for a supply chain, comprising the following steps: S1, collecting production data, inventory data, order data, logistics data and external environment data of each node under the whole chain of the supply chain in real time, thereby, a time series prediction model constructed based on a long short-term memory network and an attention mechanism can effectively capture long-term dependence and complex fluctuation characteristics in the supply chain data, greatly reduce the deviation between the prediction result and the actual demand, construct a multi-objective optimization model, take the minimization of the total cost of the supply chain and the maximization of the resource utilization efficiency as core optimization targets, combine comprehensive constraint conditions and priority parameters, realize the optimal overall benefit of the supply chain, set a deviation monitoring and local rescheduling mechanism, and when a sudden situation occurs in the supply chain, the local rescheduling can be triggered in time, so that the continuity and stability of the overall deployment rhythm of the supply chain are ensured.

Description

Technical Field

[0001] This application relates to the technical field of supply chain, and in particular to a method for dynamic allocation and optimization of ERP resources for supply chain. Background Technology

[0002] Against the backdrop of increasing collaboration and complexity in global supply chains, Enterprise Resource Planning (ERP) systems have become the core carrier for resource management across the entire supply chain. Their resource allocation capabilities determine the operational efficiency, cost control level, and market responsiveness of the supply chain. Currently, with the increasing personalization of market demands and the growing urgency for collaboration among various nodes in the supply chain, traditional ERP resource allocation methods are no longer adequate for the management needs of modern supply chains. Various shortcomings are becoming increasingly apparent, severely hindering the overall competitiveness of the supply chain.

[0003] First, existing ERP systems mostly employ traditional time-series forecasting methods for demand forecasting. These methods struggle to capture long-term dependencies and complex fluctuations in supply chain data, making them ineffective in adapting to cyclical fluctuations and sudden changes in market demand. Furthermore, existing forecasting models are largely static, with fixed parameters after training. They cannot dynamically integrate the latest collected external environmental data for adaptive updates, resulting in significant discrepancies between forecast results and actual demand.

[0004] Secondly, the optimization goals of existing ERP resource allocation are mostly single objectives, either pursuing only the minimization of total supply chain costs or simply emphasizing the maximization of resource utilization efficiency. They fail to achieve synergistic optimization of multiple objectives, which can easily lead to contradictions such as optimal cost but idle resources or highest efficiency but uncontrolled costs.

[0005] Furthermore, existing ERP resource allocation strategies are mostly static, meaning they are executed fixedly once generated, lacking effective deviation monitoring and local rescheduling mechanisms. When unexpected situations occur during supply chain operations, such as excessive deviations in node resource consumption, abnormal resource consumption rates, or the insertion of new urgent orders, adjustments cannot be triggered in a timely manner, leading to a continuous expansion of deviations and serious consequences such as work stoppages and stockouts.

[0006] Application content

[0007] This application aims to address, at least to some extent, the technical problems in the related art.

[0008] To achieve the above objectives, this application proposes a dynamic allocation and optimization method for ERP resources in the supply chain, including the following steps:

[0009] S1. Real-time collection of production data, inventory data, order data, logistics data, and external environment data from each node in the entire supply chain, and data cleaning and standardization processing of the data to form a standardized data stream with a unified data format;

[0010] S2. Input the standardized data stream into the pre-built demand forecasting model for forecasting processing, and output the predicted material demand and resource load of each node within a preset time window.

[0011] S3. Input the predicted material demand and resource load into a multi-objective optimization model with the optimization objectives of minimizing total supply chain cost and maximizing resource utilization efficiency. Combine the current resource reserves, resource constraints and priority parameters of each node, and solve the problem through a multi-objective optimization algorithm to generate a dynamic resource allocation strategy. The dynamic resource allocation strategy includes the resource allocation amount, allocation path and allocation sequence of each node.

[0012] S4. The dynamic resource allocation strategy is distributed to each execution terminal of the ERP system, and each execution terminal performs physical allocation of resources and synchronous update of resources within the system according to the allocation strategy.

[0013] S5. During execution, resource consumption data and cumulative deviation of each node are collected in real time. When the cumulative deviation exceeds the preset deviation threshold, local rescheduling is triggered, the dynamic resource allocation strategy is updated and implemented.

[0014] In addition, the application may also include the following additional technical features:

[0015] Specifically, in step S1, each node in the entire supply chain includes at least a production node, a warehousing node, a distribution node, and a sales node; the production data includes equipment operating efficiency, capacity utilization rate, and current load; the industry data includes safety stock level, current inventory level, and turnover rate; the order data includes order priority, delivery time limit, and order quantity; the logistics data includes transportation vehicle status, quantity of materials in transit, and estimated arrival time; and the external environment data includes market demand fluctuation indicators, supplier supply status, and weather and traffic conditions.

[0016] Specifically, in step S2, the demand forecasting model is a time-series forecasting model built on long short-term memory network and attention mechanism. The demand forecasting model takes historical time-series data in the standardized data stream as input and outputs the predicted material demand and resource load within the next N time steps. The demand forecasting model is trained by rolling update, and the model parameters are incrementally updated every preset period using newly added real-time data.

[0017] Specifically, the objective function of the multi-objective optimization model in step S3 includes a weighted combination of the following sub-objective functions: a function to minimize the resource allocation cost of each node, a function to minimize the total transportation cost of the supply chain, a function to maximize the resource utilization efficiency of each node, and a function to maximize the on-time delivery rate of orders; the resource constraints include the maximum inventory capacity constraint, the minimum safety stock constraint, the upper limit constraint of equipment capacity, and the transportation time window constraint for each node; the priority parameters include the order urgency level, the customer importance level, and the node capability level.

[0018] Specifically, the multi-objective optimization algorithm in step S3 adopts an improved NSGA-III algorithm, which includes: using real number encoding to encode the dynamic resource allocation strategy into chromosomes; using the optimization objective as the fitness function, and using non-dominated sorting and reference point association mechanisms for population evolution; introducing a dynamic adjustment strategy based on adaptive crossover rate and mutation rate during the evolution process; outputting a Pareto front solution set, and selecting the optimal solution that satisfies the resource constraints of each node as the dynamic resource allocation strategy.

[0019] Specifically, in step S4, each execution terminal of the ERP system includes at least: a production execution system terminal, a warehouse management system terminal, a transportation management system terminal, and a supplier collaboration platform terminal; the distribution method includes real-time data transmission and distribution through the API interface of the ERP system, combined with asynchronous event broadcasting and distribution through a message queue, to ensure that any node in the system broadcasts status change information to other related nodes immediately after receiving a resource allocation instruction.

[0020] Specifically, the triggering conditions for local rescheduling in step S5 include one or a combination of the following: the cumulative deviation of any node exceeds a preset deviation threshold, the resource consumption rate of any node deviates from the predicted resource load by more than a preset ratio threshold, and a new emergency order appears within a preset time window with a higher priority than the existing order in the current allocation strategy.

[0021] Specifically, the local rescheduling step includes: taking the node that triggers the rescheduling as the central node, limiting the rescheduling scope to the central node and its upstream and downstream first-level nodes; constrained by not exceeding the overall architecture of the dynamic resource allocation strategy, only performing local re-optimization on the allocation scheme within the limited scope, generating a local update strategy and modifying the corresponding part of the dynamic resource allocation strategy.

[0022] In summary, the beneficial effects of the ERP resource dynamic allocation optimization method for the supply chain proposed in this application are as follows:

[0023] 1. The time-series forecasting model, built on long short-term memory networks and attention mechanisms, can effectively capture long-term dependencies and complex fluctuations in supply chain data, and can adapt to various scenarios such as cyclical fluctuations and sudden changes in market demand. At the same time, the model adopts a rolling update method, which can dynamically integrate the latest collected external environmental data to achieve adaptive updates, significantly reducing the deviation between forecast results and actual demand.

[0024] 2. Construct a multi-objective optimization model with the core optimization objectives of minimizing total supply chain cost and maximizing resource utilization efficiency. At the same time, combine comprehensive constraints and priority parameters to make the generated resource allocation strategy more in line with actual business needs and achieve the best overall supply chain efficiency.

[0025] 3. Set up a deviation monitoring and local rescheduling mechanism. When a sudden situation occurs in the supply chain, local rescheduling can be triggered in time without the need to perform full forecasting and global rescheduling. Deviations can be corrected quickly to avoid shutdowns or stockouts and ensure the continuity and stability of the overall supply chain allocation rhythm. Attached Figure Description

[0026] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0027] Figure 1 This is a flowchart of the ERP resource dynamic allocation and optimization method for the supply chain in this application;

[0028] Figure 2 This is a flowchart illustrating the demand forecasting model of the supply chain-oriented ERP resource dynamic allocation and optimization method proposed in this application.

[0029] Figure 3 This is a flowchart of the optimization algorithm for the dynamic allocation and optimization method of ERP resources for the supply chain in this application. Detailed Implementation

[0030] To make the technical means, inventive features, objectives, and effects of this application easier to understand, the application is further described below with reference to specific illustrations. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0031] The present application will now be described in further detail with reference to the accompanying drawings.

[0032] like Figures 1-3 As shown in the figure, the ERP resource dynamic allocation and optimization method for the supply chain in this application includes the following steps:

[0033] S1. Real-time collection of production data, inventory data, order data, logistics data, and external environment data from all nodes in the entire supply chain, and data cleaning and standardization processing to form a standardized data stream with a unified data format.

[0034] It should be noted that the collected raw data undergoes data cleaning and standardization to form a standardized data stream with consistent format, time alignment, and efficient computation. This stream is then stored in a real-time data lake or message queue for use by downstream models.

[0035] S2. Input the standardized data stream into the pre-built demand forecasting model for forecasting processing, and output the material demand forecast and resource load forecast for each node within the preset time window.

[0036] It should be noted that the material demand forecast is: for each node in the supply chain (such as the raw material warehouse of the factory and the distribution inventory of the distribution center), the hourly / daily demand of each material within a future preset time window (e.g., 7 days, 30 days) is predicted, including the safety stock replenishment, order commitment and expected loss.

[0037] Resource load forecast: Predicts the resource consumption level of each node within the same time window, including equipment hours, labor hours, warehouse space occupancy, transportation vehicle load requirements, energy consumption, etc.

[0038] During the training phase, the model utilizes features such as demand patterns, delivery date fluctuations, and resource utilization relationships from historical data, and can dynamically integrate the latest collected external environmental data (such as promotional activity announcements and sudden work stoppage notices) for adaptive updates. The output includes confidence intervals to help the model balance risks and costs in subsequent optimization.

[0039] S3. Input the material demand forecast and resource load forecast into a multi-objective optimization model with the optimization objectives of minimizing the total supply chain cost and maximizing resource utilization efficiency. Combine the current resource reserves, resource constraints and priority parameters of each node, and solve the problem through a multi-objective optimization algorithm to generate a dynamic resource allocation strategy. The dynamic resource allocation strategy includes the resource allocation amount, allocation path and allocation sequence of each node.

[0040] It should be noted that minimizing total supply chain costs includes procurement costs, warehousing and holding costs, transportation costs, stockout losses, overtime and expedited costs, and sunk costs resulting from idle or wasted resources.

[0041] Maximize resource utilization efficiency: Achieve optimal resource input-output ratio by improving indicators such as overall equipment efficiency (OEE), warehouse turnover rate, transportation load factor, and staff hour utilization rate.

[0042] The following constraints and parameters need to be included in the optimization model:

[0043] Current resource availability: Real-time available material inventory, idle machine time, remaining storage capacity, and standby vehicles at each node.

[0044] Resource constraints include hard constraints such as maximum equipment capacity, maximum warehouse volume, load / volume limits for transport vehicles, hazardous materials storage regulations, and timeliness requirements for low-temperature delivery.

[0045] Priority parameters: Weighted coefficients set based on order urgency, customer level, material importance (such as ABC classification), and node criticality (such as bottleneck process nodes), which affect the order of resource allocation.

[0046] A multi-objective optimization algorithm is used to iteratively find the optimal solution within the feasible region, ultimately generating a dynamic resource allocation strategy. This strategy explicitly includes:

[0047] Resource allocation for each node: the quantity and type of materials to be transferred into or out of the node, as well as the scheduling instructions for the equipment and manpower required.

[0048] Allocation route: The specific link of resource flow (e.g., the multimodal transport route from warehouse A to production workshop C via transporter B), and a comparison of multiple options for cost and timeliness of the available routes.

[0049] Allocation sequence: The start and end times, sequential dependencies, and waiting windows of each allocation operation are set to ensure that the overall rhythm is coordinated with the production plan and logistics shifts.

[0050] The model supports rolling optimization: after each allocation, the strategy for the remaining time window can be recalculated based on the latest state.

[0051] S4. The dynamic resource allocation strategy is distributed to each execution terminal of the ERP system. Each execution terminal performs physical allocation of resources and synchronizes the resources within the system according to the allocation strategy.

[0052] It should be noted that the Production Execution System (MES) updates production schedules and issues raw material requisition forms, equipment switching instructions, and personnel shift schedules.

[0053] Warehouse management system: generates picking lists, transfer orders, and replenishment tasks, and triggers AGV or automated storage and retrieval system operations.

[0054] Transportation Management System: Assigns vehicles, plans routes, generates waybills, and sends dispatch instructions to the driver's mobile app.

[0055] Purchasing / Sales module: Automatically generates drafts of external adjustment orders for manual review.

[0056] While executing physical allocation, each terminal must update the resource status (inventory accounts, equipment occupancy tables, human resources calendar) in the ERP system in real time to ensure that the digital twin in the system is synchronized with the physical world. To ensure consistency, a two-phase commit or event sourcing mechanism can be adopted to avoid data discrepancies caused by network latency or execution failure.

[0057] S5. During execution, collect resource consumption data and cumulative deviation of each node in real time. When the cumulative deviation exceeds the preset deviation threshold, trigger local rescheduling, update the dynamic resource allocation strategy and issue it for implementation.

[0058] It should be noted that the system has multiple preset deviation thresholds. When any cumulative deviation exceeds the corresponding threshold (for example, the material shortage deviation exceeds 20% of the safety stock, or the equipment load deviation may cause the next bottleneck process to stop), a local rescheduling is immediately triggered.

[0059] Local rescheduling does not re-execute the full prediction in step S2. Instead, based on the current actual state and remaining plans, it calls the fast rescheduling algorithm to correct only the strategies for affected nodes and within the time range. If the corrected strategy is verified by simulation to be better than the original plan and meets the constraints, it automatically overrides the original strategy, and step S4 is re-executed. The rescheduling process retains on-site operation logs and reports them to the central monitoring dashboard for subsequent analysis and optimization by managers.

[0060] In one embodiment of this application, each node in the entire supply chain in step S1 includes at least a production node, a warehousing node, a distribution node, and a sales node; production data includes equipment operating efficiency, capacity utilization rate, and current load; industry data includes safety stock level, current inventory level, and turnover rate; order data includes order priority, delivery time limit, and order quantity; logistics data includes transportation vehicle status, quantity of materials in transit, and estimated arrival time; external environment data includes market demand fluctuation indicators, supplier supply status, and weather and traffic conditions.

[0061] It should be noted that production data is used to assess the real-time operational status of the production end; industry data is used to determine inventory health; order data reflects the time sensitivity and scale of customer demand; logistics data involves the status of transportation vehicles, the quantity of materials in transit, and estimated arrival times (based on GPS and historical road condition calculations); and external environment data collects market demand fluctuation indicators (such as order growth rates or seasonal indices), supplier supply status (normal, delayed, interrupted), and weather and traffic conditions (such as rainfall and road closure information). All of the above multi-source heterogeneous data is collected through a unified interface standard.

[0062] In one embodiment of this application, the demand forecasting model in step S2 is a time-series forecasting model constructed based on a long short-term memory network and an attention mechanism. The demand forecasting model takes historical time-series data in a standardized data stream as input and outputs the predicted material demand and resource load within the next N time steps. The demand forecasting model is trained using a rolling update method, and the model parameters are incrementally updated every preset period using newly added real-time data.

[0063] Specifically, the Long Short-Term Memory Network (LSTM) effectively captures long-term dependencies (such as periodic demand fluctuations) in supply chain data through the structural design of forget gates, input gates, and output gates, while mitigating the gradient vanishing problem. The attention mechanism dynamically assigns weights to features at different time steps in historical time series data, highlighting time segments that have a greater impact on the current prediction task (such as periods of recent order surges).

[0064] The model takes historical time-series data from a standardized data stream as input and outputs predicted material demand (types and quantities of materials required at each node) and resource load (such as projected equipment occupancy rate and projected warehouse inbound / outbound volumes) for the next N time steps (e.g., N=7 days). In terms of training, the model employs a rolling update mechanism: every preset period (e.g., every 24 hours), the model parameters are incrementally updated using newly added real-time data, rather than being completely retrained. This maintains the model's sensitivity to the latest dynamics while reducing computational overhead.

[0065] In one embodiment of this application, the objective function of the multi-objective optimization model in step S3 includes a weighted combination of the following sub-objective functions: a function to minimize the resource allocation cost of each node, a function to minimize the total transportation cost of the supply chain, a function to maximize the resource utilization efficiency of each node, and a function to maximize the on-time delivery rate of orders; resource constraints include the maximum inventory capacity constraint, the minimum safety stock constraint, the upper limit constraint of equipment capacity, and the transportation time window constraint of each node; priority parameters include the order urgency level, the customer importance level, and the node capability level.

[0066] Specifically, the resource allocation cost minimization function at each node measures the direct costs (such as overtime pay and cross-warehouse transfer freight) incurred by the transfer of manpower, equipment, and materials between different nodes.

[0067] The total transportation cost minimization function for the supply chain covers trunk transportation, feeder delivery, and last-mile costs throughout the entire process from production to sales;

[0068] The function for maximizing resource utilization efficiency at each node is expressed as the inverse of a combination of indicators such as equipment hour utilization and warehouse location utilization, encouraging balanced load.

[0069] The on-time delivery rate maximization function is calculated based on the degree of conformity between the planned completion time and the actual completion time of the order, and can introduce a penalty term for delayed delivery.

[0070] Priority parameters are embedded as soft constraints in the optimization process, including order urgency levels (e.g., express, expedited, regular), customer importance levels (e.g., strategic customers, regular customers), and node capability levels (reflecting the upper limit of node processing capacity). These parameters collectively determine the evaluation criteria and direction of feasible solutions.

[0071] In one embodiment of this application, the multi-objective optimization algorithm in step S3 adopts an improved NSGA-Ⅲ algorithm, specifically including: using real number encoding to encode the dynamic resource allocation strategy into chromosomes; using the optimization objective as the fitness function, and using non-dominated sorting and reference point association mechanisms for population evolution; introducing a dynamic adjustment strategy based on adaptive crossover rate and mutation rate during the evolution process; outputting the Pareto front solution set, and selecting the optimal solution that satisfies the resource constraints of each node as the dynamic resource allocation strategy.

[0072] It should be noted that the improved third-generation non-dominated sorting genetic algorithm (NSGA-Ⅲ) is used to solve the multi-objective optimization problem.

[0073] Encoding method: Real number encoding is used to represent the dynamic resource allocation strategy as chromosomes. Each gene position of each chromosome corresponds to a decision variable, such as the capacity allocation ratio of the production node, the replenishment trigger threshold of the storage node, and the vehicle route selection coefficient of the delivery node.

[0074] Fitness function: The four optimization objectives mentioned above are directly used as the fitness function vector. The algorithm divides the population into different levels based on non-dominated sorting, and introduces a reference point association mechanism to maintain the diversity of the solution set in the high-dimensional objective space.

[0075] Adaptive evolution strategy: In crossover and mutation operations, a dynamic adjustment strategy is adopted. When the population convergence speed is lower than the preset threshold, the crossover rate and mutation rate are adaptively increased to enhance the global search capability; when trapped in a local optimum, the mutation rate is reduced to stabilize the superior genes.

[0076] Solution Set Output and Optimization: The algorithm ultimately outputs a Pareto front solution set, which is a set of non-dominated solutions that cannot improve other objectives without compromising any of the objectives. Subsequently, based on the resource constraints of each node and business rules, an optimal solution is selected from this set as the dynamic resource allocation strategy to be implemented.

[0077] In one embodiment of this application, each execution terminal of the ERP system in step S4 includes at least: a production execution system terminal, a warehouse management system terminal, a transportation management system terminal, and a supplier collaboration platform terminal; the distribution method includes real-time data transmission and distribution through the API interface of the ERP system, combined with asynchronous event broadcasting and distribution through a message queue, to ensure that any node in the system broadcasts status change information to other related nodes immediately after receiving a resource allocation instruction.

[0078] It should be noted that the API interface transmits and distributes data in real time: the ERP central server pushes resource allocation instructions as structured data directly to each terminal through RESTful API or gRPC protocol to ensure the real-time nature of critical instructions;

[0079] Asynchronous broadcasting via message queue: Commands are simultaneously published to message middleware such as Kafka, enabling asynchronous event broadcasting in conjunction with the publish mode. Upon receiving a resource allocation command, any node (terminal or service) automatically triggers a state change event and broadcasts this change information to other related nodes via the message queue.

[0080] The combination of real-time and asynchronous design ensures that even in the event of network fluctuations or temporary offline status of a terminal, other nodes can still eventually receive a complete copy of the instructions through the message queue, thus maintaining the eventual consistency of the entire supply chain system. The decentralized nature of the broadcast link also avoids the information silo problem caused by single points of failure.

[0081] In one embodiment of this application, the triggering conditions for local rescheduling in step S5 include one or a combination of the following: the cumulative deviation of any node exceeds a preset deviation threshold, the resource consumption rate of any node deviates from the predicted resource load by more than a preset ratio threshold, and a new emergency order appears within a preset time window with a higher priority than the priority of existing orders in the current allocation strategy.

[0082] It should be noted that the cumulative deviation exceeds the limit: the cumulative deviation between the actual operating parameters of any node (such as actual inventory or actual capacity consumption) and the predicted values ​​in the dynamic resource allocation strategy exceeds a preset deviation threshold (e.g., the cumulative deviation exceeds 15% of the predicted value). This deviation is calculated using a sliding window (e.g., the past 6 hours) to avoid frequent rescheduling caused by a single small fluctuation.

[0083] Resource consumption rate deviates from the proportion: If the resource consumption rate of any node (such as the actual outbound rate of materials or the actual working hour consumption rate of equipment) deviates from the resource load forecast by more than the preset proportion threshold (for example, the actual consumption rate is higher than the forecast value by 30% and lasts for more than 10 minutes), it indicates a surge in demand or abnormal efficiency.

[0084] Urgent order insertion: When a new urgent order appears within a preset time window (such as a delivery window of the next 2 hours), and the priority of this order (such as the expedited level) is higher than the priority of all existing incomplete orders in the current allocation strategy, the original strategy's sorting and resource reservation will be disrupted, requiring local intervention.

[0085] In one embodiment of this application, the local rescheduling step specifically includes: taking the node that triggers the rescheduling as the central node, limiting the rescheduling scope to the central node and its upstream and downstream first-level nodes; constrained by the overall architecture of the dynamic resource allocation strategy, only the allocation scheme within the limited scope is locally re-optimized, generating a local update strategy and correcting the corresponding part in the dynamic resource allocation strategy.

[0086] It should be noted that the rescheduling scope is limited to the central node and its upstream first-level nodes (such as production nodes supplying the warehouse) and downstream first-level nodes (such as distribution centers or sales outlets supplied by the warehouse). This scope is selected based on the material flow diagram of the supply chain, where the disturbance transmission effect of first-level nodes is the most significant, while nodes further away are less affected by the delay of local disturbances and can be handled in the next round of global optimization.

[0087] Constraints: Local rescheduling is constrained to not exceed the overall architecture of the original dynamic resource allocation strategy. That is, it does not change the decision variables outside the scope of the strategy (e.g., external transportation routes that are not involved remain as planned), and only allows local re-optimization of allocation schemes within the scope of the strategy (e.g., the replenishment volume of the warehouse, the delivery rhythm of upstream and downstream nodes).

[0088] Generation and Correction: The subproblem is solved using a heuristic algorithm (such as local search or constrained genetic operators) to generate a local update strategy. This strategy is then used to replace the corresponding data fields in the original dynamic resource allocation strategy, while the rest of the global strategy remains unchanged.

[0089] This design responds to local disturbances while avoiding the high computational cost of global rescheduling, and ensures the continuity and stability of the overall supply chain strategy.

[0090] It should be noted that, in this document, 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.

[0091] The present application and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present application. The actual structure is not limited to this. In conclusion, if a person skilled in the art is inspired by this description and designs a similar structure and embodiment without departing from the spirit of the present application, such design should fall within the protection scope of the present application.

Claims

1. A method for dynamic allocation and optimization of ERP resources for supply chain management, characterized in that: Includes the following steps: S1. Real-time collection of production data, inventory data, order data, logistics data, and external environment data from each node in the entire supply chain, and data cleaning and standardization processing of the data to form a standardized data stream with a unified data format; S2. Input the standardized data stream into the pre-built demand forecasting model for forecasting processing, and output the predicted material demand and resource load of each node within a preset time window. S3. Input the predicted material demand and resource load into a multi-objective optimization model with the optimization objectives of minimizing total supply chain cost and maximizing resource utilization efficiency. Combine the current resource reserves, resource constraints and priority parameters of each node, and solve the problem through a multi-objective optimization algorithm to generate a dynamic resource allocation strategy. The dynamic resource allocation strategy includes the resource allocation amount, allocation path and allocation sequence of each node. S4. The dynamic resource allocation strategy is distributed to each execution terminal of the ERP system, and each execution terminal performs physical allocation of resources and synchronous update of resources within the system according to the allocation strategy. S5. During execution, resource consumption data and cumulative deviation of each node are collected in real time. When the cumulative deviation exceeds the preset deviation threshold, local rescheduling is triggered, the dynamic resource allocation strategy is updated and implemented.

2. The method for dynamic allocation and optimization of ERP resources for supply chain orientation according to claim 1, characterized in that, In step S1, each node in the entire supply chain includes at least a production node, a warehousing node, a distribution node, and a sales node; the production data includes equipment operating efficiency, capacity utilization rate, and current load; the industry data includes safety stock level, current inventory level, and turnover rate; the order data includes order priority, delivery time limit, and order quantity; the logistics data includes transportation vehicle status, quantity of materials in transit, and estimated arrival time; and the external environment data includes market demand fluctuation indicators, supplier supply status, and weather and traffic conditions.

3. The method for dynamic allocation and optimization of ERP resources for supply chain orientation according to claim 1, characterized in that, In step S2, the demand forecasting model is a time-series forecasting model built on long short-term memory network and attention mechanism. The demand forecasting model takes historical time-series data in the standardized data stream as input and outputs the predicted material demand and resource load within the next N time steps. The demand forecasting model is trained using a rolling update method, and the model parameters are incrementally updated every preset period using newly added real-time data.

4. The method for dynamic allocation and optimization of ERP resources for supply chain orientation according to claim 1, characterized in that, The objective function of the multi-objective optimization model in step S3 includes a weighted combination of the following sub-objective functions: minimizing the resource allocation cost of each node, minimizing the total transportation cost of the supply chain, maximizing the resource utilization efficiency of each node, and maximizing the on-time delivery rate of orders; the resource constraints include the maximum inventory capacity constraint, minimum safety stock constraint, equipment capacity upper limit constraint, and transportation time window constraint of each node; the priority parameters include the order urgency level, customer importance level, and node capability level.

5. The method for dynamic allocation and optimization of ERP resources for supply chain orientation according to claim 1, characterized in that, The multi-objective optimization algorithm in step S3 adopts an improved NSGA-III algorithm, specifically including: using real number encoding to encode the dynamic resource allocation strategy into chromosomes; using the optimization objective as the fitness function, and using non-dominated sorting and reference point association mechanisms for population evolution; introducing a dynamic adjustment strategy based on adaptive crossover rate and mutation rate during the evolution process; outputting a Pareto front solution set, and selecting the optimal solution that satisfies the resource constraints of each node as the dynamic resource allocation strategy.

6. The method for dynamic allocation and optimization of ERP resources for supply chain orientation according to claim 1, characterized in that, In step S4, the execution terminals of the ERP system include at least: the production execution system terminal, the warehouse management system terminal, the transportation management system terminal, and the supplier collaboration platform terminal; the distribution method includes real-time data transmission and distribution through the API interface of the ERP system, combined with asynchronous event broadcasting and distribution through message queues, to ensure that any node in the system broadcasts the status change information to other related nodes immediately after receiving the resource allocation instruction.

7. The method for dynamic allocation and optimization of ERP resources for supply chain orientation according to claim 1, characterized in that, The triggering conditions for local rescheduling in step S5 include one or a combination of the following: the cumulative deviation of any node exceeds a preset deviation threshold; the resource consumption rate of any node deviates from the predicted resource load by more than a preset ratio threshold; a new emergency order appears within a preset time window and its priority is higher than the priority of existing orders in the current allocation strategy.

8. The method for dynamic allocation and optimization of ERP resources for supply chain orientation according to claim 1, characterized in that, The local rescheduling step specifically includes: taking the node that triggers the rescheduling as the central node, limiting the rescheduling scope to the central node and its upstream and downstream first-level nodes; constrained by not exceeding the overall architecture of the dynamic resource allocation strategy, only locally re-optimizing the allocation scheme within the limited scope, generating a local update strategy and modifying the corresponding part of the dynamic resource allocation strategy.

Images