Power grid material logistics layout and influence factor evaluation method

Abstract

The invention relates to a power grid material logistics layout and influence factor evaluation method, and belongs to the technical field of electric power material distribution optimization. According to the method, for a complex logistics scene of multiple parking lots, multiple tasks and multiple vehicle types, a unified network model including the parking lots, a distribution center and customer demand points is constructed, and dynamic planning and risk assessment of a distribution path are realized by integrating a time window punishment mechanism, multi-vehicle-type cost optimization and a multi-scene simulation technology. The method specifically comprises the following steps: coding a parking lot, a distribution center and a customer point into network nodes, and designing a path rule of closed-loop distribution after a vehicle starts from the parking lot and is loaded by the distribution center; time window deviation punishment cost is defined, and a total cost minimization target is established in combination with transportation cost and empty driving cost; generating a distribution scheme by adopting a hybrid optimization algorithm combining a genetic algorithm and tabu search; according to the invention, the logistics cost can be effectively reduced, the distribution timeliness is improved, the power grid material emergency response capability is enhanced, and scientific decision support is provided for power enterprises.

Description

Technical Field

[0001] This invention belongs to the field of power grid material transportation and logistics, and more specifically relates to a method for evaluating the layout and influencing factors of power grid material logistics. Background Technology

[0002] With the continuous development and expansion of power systems, the demand for transportation and distribution of materials during power grid maintenance and construction is increasing daily. Efficient management of power grid material logistics is crucial not only for the stable operation of the power system but also for the maintenance efficiency of power equipment and the reliability of power supply. Therefore, optimizing the logistics layout and transportation routes of power grid materials has become an important research topic.

[0003] Traditional logistics management methods often employ a single vehicle depot and single vehicle type transportation model, which is suitable for small-scale or simple logistics needs. However, with the expansion of power grid coverage and the increase in the types of goods, the single vehicle depot and single vehicle type model can no longer meet the complex and ever-changing logistics demands. A multi-depot, multi-task, and multi-vehicle logistics model can more flexibly address the distribution needs of different regions and different types of goods, but it also brings more complex route optimization problems.

[0004] The Vehicle Routing Problem (VRP), a classic model in logistics optimization, aims to minimize transportation costs, time, and resource consumption by rationally arranging vehicle delivery routes. Currently, VRP research has developed various variations, such as VRP with time windows, multi-depot VRP, and multi-vehicle type VRP. These models can solve complex problems in real-world logistics to some extent. However, existing VRP models still have some shortcomings when dealing with the logistics layout of power grid materials.

[0005] First, existing models often neglect the coordinated scheduling problem among multiple depots during power grid material delivery. In practice, power grid materials may be distributed across multiple depots. A single-depot model cannot fully utilize the resources of each depot, leading to uneven resource allocation and increased overall transportation costs. Second, power grid materials are diverse, and different vehicle types have different adaptability and cost-effectiveness when transporting different materials. Existing multi-vehicle VRP models often fail to fully consider the synergistic effects and dynamic scheduling between vehicle types. Furthermore, power grid material delivery often has strict time window requirements; any delay or advancement in delivery may lead to production stagnation or resource waste. The flexibility and accuracy of existing models in handling time window constraints still need improvement.

[0006] Furthermore, power grid material logistics also faces supply chain uncertainties, such as demand fluctuations and changes in transportation conditions. These uncertainties significantly impact logistics layout and transportation routes. Traditional deterministic models struggle to effectively address these uncertainties, resulting in insufficient robustness of the logistics system and an inability to maintain efficient operation in dynamic environments. Therefore, constructing a highly adaptable logistics layout model that optimizes transportation routes in complex environments with multiple depots, tasks, and vehicle types has become a pressing technical challenge.

[0007] In summary, existing power grid material logistics management methods suffer from low resource utilization, high transportation costs, and insufficient system robustness when facing demands such as multiple vehicle depots, multiple vehicle types, and complex time windows. Therefore, there is an urgent need for a method that can comprehensively consider multiple influencing factors and optimize the layout and transportation routes of power grid material logistics to improve the overall efficiency and reliability of the logistics system. This leads to a research direction of significant practical importance for this invention: a method for evaluating the layout and influencing factors of power grid material logistics. Summary of the Invention

[0008] Power grid material logistics management faces complex demands involving multiple depots, tasks, and vehicle types. Traditional logistics optimization methods have significant shortcomings in resource allocation, route optimization, time management, and uncertainty management. Therefore, there is an urgent need for an optimization method that comprehensively considers multi-depot coordination, multi-vehicle adaptability, strict time window constraints, and supply chain uncertainties to achieve efficient power grid material logistics layout and minimize system costs. Solving these technical problems will not only improve the efficiency and reliability of power grid material distribution but also significantly reduce operating costs, possessing significant practical and economic implications.

[0009] To achieve the above objectives, the present invention employs the following technical solution: the method comprises:

[0010] Construct a logistics network model with multiple vehicle depots, multiple tasks, and multiple vehicle types, and uniformly encode vehicle depots, distribution centers, and customer demand points as network nodes to form a weighted graph containing node sets and path arc sets;

[0011] Based on the geographical location of the customer's demand point, time window constraints, and delivery priority, plan a closed path for vehicles to depart from the depot, be loaded at the distribution center, deliver to the customer's point, and return to the original depot.

[0012] Define a time window penalty mechanism, and calculate the time penalty cost by combining the deviation between the actual time the vehicle arrives at the customer point and the time window;

[0013] Based on the transportation costs, empty driving costs, and time penalty costs of multiple vehicle types, an optimization objective of minimizing total cost is established;

[0014] By employing a multi-vehicle-yard scheduling strategy, vehicles of different models are allocated to perform delivery tasks, generating delivery plans that meet capacity, route continuity, and time constraints.

[0015] In one scheme, the classification of network nodes includes:

[0016] Customer node set (N): Composed of power grid material demand points, each node is associated with demand, service time window and priority label;

[0017] Distribution Center Node (O): As the sole cargo transit node, it is responsible for receiving vehicles loaded in the parking lot and distributing goods;

[0018] Vehicle depot node set (S): Each depot is equipped with multiple types of vehicles, including models with different load capacities, ranges and cost characteristics;

[0019] The node coding rule is as follows: customer point number 1 to N, distribution center number N+1, parking lot number N+2 to N+S+1, where S is the total number of parking lots.

[0020] In one embodiment, the goods allocation method of the distribution center includes:

[0021] Real-time monitoring of vehicle arrival times at the distribution center from each depot; dynamic allocation of loading tasks to avoid congestion.

[0022] Optimize cargo loading sequence based on vehicle capacity limitations and customer demand priorities;

[0023] Radio frequency identification (RFID) technology is used to track materials, ensuring accurate loading and keeping inventory updated.

[0024] In one scheme, the time window penalty mechanism includes:

[0025] For vehicles that arrive before the earliest service time, the penalty cost is calculated based on the waiting time and the penalty coefficient per unit time.

[0026] For vehicles arriving after the latest service time, penalty costs are calculated based on a penalty coefficient weighted by the delay time and customer priority.

[0027] The penalty cost is incorporated into the overall optimization objective, and the time deviation and transportation cost are balanced by adjusting the route planning.

[0028] In one scheme, the vehicle routing optimization algorithm includes:

[0029] Initial path generation is based on genetic algorithm, and path combination is iteratively optimized through crossover and mutation operations;

[0030] A tabu search strategy is introduced to avoid local optima and dynamically adjust the relationship between vehicle type and parking lot allocation.

[0031] Urgent orders are handled using an insertion method, which dynamically inserts new tasks into the existing path and recalculates the time window constraints.

[0032] In one scheme, the multi-parking lot scheduling strategy includes:

[0033] Based on the spatial distribution of parking lots and customer locations, service areas are divided and tasks are assigned to the nearest parking lot.

[0034] In cross-regional demand scenarios, enabling collaboration mechanisms allows vehicle resources to be shared between parking lots; load balancing algorithms limit the maximum workload of a single parking lot to reduce operational risks.

[0035] In one approach, the assessment of influencing factors includes:

[0036] Sensitivity analysis: Adjust the number of vehicle models, time window width, or penalty coefficient to quantify the impact on total cost;

[0037] Comparative analysis: Evaluate the cost savings and carbon emission reduction effects of setting up temporary parking lots or introducing electric vehicles;

[0038] A decision matrix is ​​generated to recommend the optimal parking lot layout, vehicle configuration, and flexible time window strategy. The beneficial effects of this invention are:

[0039] First, this invention employs a multi-depot collaborative scheduling mechanism, significantly improving the utilization efficiency of logistics resources by optimizing resource allocation and coordination among multiple depots. Compared to the traditional single-depot model, this invention fully leverages the resource advantages of each depot, avoiding concentrated waste and inefficient use of resources, thereby substantially reducing overall transportation costs. Simultaneously, multi-depot collaborative scheduling enhances the flexibility and responsiveness of the logistics system, enabling faster material distribution to address grid maintenance and emergency needs.

[0040] This invention significantly improves the optimization level of power grid material logistics layout and transportation routes through multi-vehicle collaborative scheduling, multi-vehicle adaptive optimization, strict time management, and robust uncertainty handling mechanisms. These beneficial effects not only improve the overall efficiency and reliability of logistics management, but also significantly reduce operating costs, enhance system stability and sustainable development capabilities, and have broad application prospects and significant practical economic value. Attached Figure Description

[0041] Figure 1 This is a flowchart of the method of the present invention;

[0042] Figure 2 For vehicle delivery route maps;

[0043] Figure 3 Sensitivity analysis;

[0044] Figure 4 Sensitivity analysis. Detailed Implementation

[0045] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Typical embodiments of the invention are shown in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete.

[0046] Unless otherwise defined, all technical and scientific terms used in this invention have the same meaning as understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. To facilitate understanding, the invention will now be described more fully with reference to the accompanying drawings. Typical embodiments of the invention are shown in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to make the disclosure of the invention more thorough and complete. Figure 1 As shown, a method for assessing the layout and influencing factors of power grid material logistics is presented.

[0047] Step 1: Construct a VRP model with multiple vehicle sites, multiple tasks, and multiple vehicle types.

[0048] Treating the parking lot, distribution center, and delivery point as nodes in the same network, urban freight delivery begins with a vehicle of model K corresponding to parking lot S, which first loads goods at distribution center O, then delivers the goods to customer point N. After completing the delivery task, the vehicle returns to its original parking lot. The delivery task can be represented by a weighted graph G(V, E), as follows: Figure 2 :

[0049] Wherein: S i (i = 1, 2, ..., s) represents the parking lot, N i (i = 1, 2, ..., n) represent demand points, O represents a distribution center, and V represents a distribution center. i (i = 1, 2, ..., n) represents the demand point N. i Corresponding node V n+i V n+i V represents the node corresponding to the distribution center. i (i = n+2, n+3, ..., n+s+1) represents the parking lot S. i The corresponding nodes are V = (v1, v2, ..., v n v n+1 v n+2 v n+3 , ..., v n+m+1 Let E be a set of nodes, and E = {v} i vj ):v i v j ∈V, i≠j} is an arc set. The set V consists of three subsets: N={v1, v2, ..., v...} n} represents the set of client nodes, O = {v n+1 Let S be the set of distribution center nodes, where S = {v} n+2 v n+3 , ..., v n+s+1} represents the set of nodes in the parking lot. For each node v i ∈N, indicating that there is customer demand at this node; for each node vi∈S, it means that this node has several vehicles of different models, and K={1,2,…,k} represents the set of vehicle models. Each customer point v i There is a service time window [ET] i ,LTi],ET i Indicates the earliest start time, LT i This indicates the latest start time for service. A time penalty cost will be incurred if a vehicle arrives at the customer's location before the earliest start time or after the latest start time.

[0050] To simplify the mathematical model of this problem, the sets, parameters, and variables used in the mathematical model are defined as follows:

[0051] Customer locations are numbered 1, 2, ..., N; distribution centers are numbered N+1; parking lots are numbered N+2, N+3, ..., N+S+1; vehicle models are numbered 1, 2, ..., K; r sk ij = (1, 0) means that if a type k car in parking lot s travels from node i to node j, then r sk ij =1, otherwise r sk ij =0; Dij represents the distance from node i to node j; N represents the total number of customers; S represents the total number of parking lots; Ck represents the transportation cost per unit distance for type k vehicles; E k V represents the empty driving cost per unit distance for vehicle type K; k Q represents the distance traveled per unit time by the type K vehicle; k Indicates the loading capacity of the K-type vehicle; W i T represents the quantity of goods demanded by customer i; i t represents the time it takes for the vehicle to arrive at node i; ij ET represents the time it takes for a vehicle to travel from node i to node j; i Indicates the earliest start time for customer i; LT iLet represent the latest service start time for customer i; 'a' represents the early arrival penalty coefficient, i.e., the time cost of arriving each unit of time earlier; and 'b' represents the late arrival penalty coefficient, i.e., the time cost of arriving each unit of time later. The goal of our mathematical model is to minimize the sum of transportation costs, empty-run costs, and time costs incurred during vehicle dispatching tasks in urban delivery. The minimum cost objective function is as follows:

[0052]

[0053] The meaning of the relevant expressions is as follows:

[0054] (1) means minimizing the sum of the vehicle's transportation cost, empty driving cost, and time cost.

[0055] (2) indicates that the delivery volume of each vehicle type cannot exceed its capacity limit.

[0056] (3) indicates that each task can only be served by the vehicle once.

[0057] (4) and (5) indicate that on the sub-path of each vehicle type, the vehicle starts from the parking lot, reaches a node and then leaves, eventually returning to the original parking lot.

[0058] (6) indicates the time when the vehicle arrives at task point j.

[0059] (7) indicates that each user point is guaranteed to be served. In addition, the initial time is counted from the parking lot and the initial time is 0.

[0060] Based on the TSP ant colony algorithm, and fully considering the specific requirements of VRP, the algorithm's selection, update, and coordination mechanisms are further improved by introducing an adaptive transfer strategy. Simultaneously, taking into account the diversity of social division of labor within the ant colony, tasks with many constraints are assigned to scout ants, while the search ants complete the task of finding feasible solutions to the objective function. Through the division of labor and cooperation among different ant species, the pheromone adaptive strategy and the polymorphic ant colony algorithm are finally integrated, effectively overcoming the shortcomings of the basic ant colony algorithm, such as long computation time and susceptibility to premature convergence. The solution steps are as follows:

[0061] (1) Initialize Q, C and the maximum number of generations N. c , max.

[0062] (2) Place n scout ants in n cities. Each scout ant takes its city i as the center, scouts the other n-1 cities, calculates the total scout table, and assigns the result to sij.

[0063] (3) Design the amount of information on each path at the initial moment.

[0064] (4) Set the number of generations Nc The initial value is 0.

[0065] (5) Randomly select the initial position of each search ant and put that position into the tabuk table corresponding to each search ant.

[0066] (6) Calculate the position that each search ant k will move to, let it be the position j to be moved to, and let the previous position be i. Put j into the tabuk diagram corresponding to the search ant k until each search ant completes a cycle and a solution is obtained.

[0067] (7) Calculate the objective function value f(Zk) (k = 1, 2, ..., n) for each search ant and record the best solution at present.

[0068] (8) Perform 2-opt local optimization. If the new solution is better than the original solution, replace it; otherwise, retain the original solution.

[0069] (9) If the maximum number of generations N is reached c ,max, go to (12), if the solution obtained has not improved significantly in the most recent iterations before reaching the maximum number of evolutions, then adjust the volatility coefficient ρ, go to (11), otherwise go to (10).

[0070] (10) Modify the pheromone concentration on each path, set τij = 0, and set the tabuk table to empty.

[0071] (11)N c ←N c +1, turn (5).

[0072] (12) Output the optimal solution.

[0073] Simulation results: Consider 3 parking lots, 1 distribution center, and 16 customer points, all distributed within a 100×100 square area. Each parking lot has two vehicle types (model 1 and model 2). Customer data is shown in the table below, where Wi represents the cargo demand of customer i. The table lists the demand of all 16 customers, along with their earliest and latest service start times. The parking lot and distribution center data tables show the number of vehicle types 1 and 2 available at each location. The vehicle type data tables also provide the unit distance transportation cost, empty-run cost, travel distance, and loading capacity for both type 1 and type 2 vehicles.

[0074]

[0075]

[0076]

[0077] Based on the above data, the following parameter values ​​were used: α = 1, β = 2, ρ(t0) = 1, n = 100, Q = 100, C = 3, Max(Pc) = 10, and the number of iterations was 50. The minimum total cost obtained by the model solution was 425.6, and the vehicle scheduling situation is shown in the table below. Considering the total cost under the condition of random vehicle scheduling, the total cost is calculated to be 440.92. The results show that the total cost optimized using the adaptive polymorphic ant colony algorithm is 425.6, which is less than the unoptimized total cost of 440.92 under the random case. This indicates that it is necessary to optimize the VRP model for multiple vehicle depots, multiple vehicle types, and multiple tasks, and also shows that the adaptive polymorphic ant colony algorithm has achieved a satisfactory optimization effect.

[0078]

[0079] Step 2: Deterministic Supply Chain Model [DM (Deterministic Model)]

[0080] (1) Set.

[0081] I: A collection of potential suppliers with signed agreements in their inventory;

[0082] J: A set of alternative warehouses (transit warehouses or terminal warehouses);

[0083] T|T|: Planning period set (containing several periods);

[0084] K: Set of demand points.

[0085] (2) Index.

[0086] i: Supplier index;

[0087] j: Repository index;

[0088] t: Periodic index;

[0089] k: Demand point index.

[0090] (3) Parameters.

[0091] α i Fixed costs associated with collaborating with supplier i;

[0092] c t jj : The unit transfer cost from warehouse j to warehouse j at the end of period t, or the inventory holding cost of warehouse j at the end of period t;

[0093] o t ik The unit transportation cost from supplier i to demand point k in period t;

[0094] o t jk The unit transportation cost from warehouse j to demand point k in period t;

[0095] h t j The unit inventory holding cost of warehouse j in period t;

[0096] I i0 Supplier i's initial inventory at the start of the planning period;

[0097] I 0j The initial inventory of warehouse j at the start of the planning period;

[0098] m i : The maximum contractable inventory level of supplier i;

[0099] β j : The maximum inventory level of warehouse j;

[0100] d t k Demand at point k in period t (assuming it is random and discrete).

[0101] (4) Decision variables

[0102] xti: Supplier i's contracted inventory level in period t;

[0103] xtj: The amount of inventory that warehouse j has prepared in period t;

[0104] z i : The maximum contracted inventory quantity with supplier i;

[0105] I t i : Supplier i's inventory at the end of period t;

[0106] I t j The inventory level of warehouse j at the end of period t;

[0107] u t ik Supplier i allocates demand d to demand point k at time t. t k The proportion;

[0108] v t jk Warehouse j satisfies demand d at demand point k during period t. t k The proportion;

[0109] f t jj*The quantity transferred from warehouse j to warehouse j* at the end of period t.

[0110] Its objective function is:

[0111]

[0112] The constraints are as follows:

[0113]

[0114] The objective function aims to minimize total cost, integrating supplier-side and warehouse-side costs. The first term represents the fixed cost of signing agreements with potential supplier inventory agreements (reflected in bidding costs); the second term is the sum of the supplier's inventory preparation cost (reflected in the purchase price) and the cost of allocating the supplier's inventory preparation quantity to demand points (reflected in distribution costs) for each period; the third term is the sum of the warehouse's self-stocking cost, inventory holding cost, inventory turnover and allocation cost, and the cost of allocating inventory to demand points (reflected in distribution costs) for each period.

[0115] The first constraint is the ending inventory level of the supplier's contracted inventory. Within each time period, the available allocation should not exceed the ending inventory level of the previous period's contracted inventory plus the current period's replenishment. The second constraint is the upper limit of contracted inventory contracts that cooperative suppliers can sign, and the contracted inventory volume cannot exceed the cooperative supplier's capacity limit. The third constraint ensures that the replenishment volume of contracted inventory suppliers across all periods does not exceed the limit of the signed contracted inventory volume. The fourth constraint is the warehouse's ending inventory constraint for the period, which equals the ending inventory level of the previous period plus the current period's self-held inventory and the amount transferred in from other warehouses, minus the amount transferred out from the warehouse and the amount allocated to various demand points. The fifth constraint is the maximum upper bound constraint on the warehouse's ending inventory level for each period. The sixth constraint ensures that the demand of each demand point can be met in each period, supplied jointly by the warehouse's self-held inventory and contracted inventory. The seventh constraint is the non-negativity and 0-1 integer constraint for all decision variables.

[0116] In reality, to better reduce their warehousing costs, companies often establish corresponding management systems to prioritize the matching and allocation of physical inventory to demand. Therefore, the usage priority of warehouse-stored materials differs from that of contract-based inventory stored with suppliers. Contract-based inventory has a lower usage priority than physical materials in the warehouse. Thus, the following logical relationship exists when using inventory materials:

[0117] if and but

[0118] This logic means that if both the warehouse and the contracted inventory supplier provide materials to the demand point at the same time during a certain period, the ending inventory of that period must be 0. If the ending inventory is not 0, physical inventory should be allocated first, so there is no need to use the reserved materials in the contracted inventory.

[0119] because and It is automatically satisfied, if but Considering the nonnegativity of I, we can conclude that but

[0120] To express the above equation as a constraint of the optimization model, we introduce |T| auxiliary 0-1 variables to construct the following |T| sets of optimization constraints, where M is a sufficiently large constant.

[0121]

[0122] Step 3, Build a robust model (RM)

[0123] Real-world demand forecasts are often imprecise and can be considered uncertain. Decision-making under uncertainty is receiving increasing attention across various industries. First, a robust optimization framework is used to improve the deterministic model above. Let Ω = {1, 2, ..., S} be the set of possible future scenarios, with each scenario having a probability P. s Satisfying ΣP s =1. To better describe the situational adaptability of the model, a special symbol e needs to be introduced. t s φ represents the deviation between the supplier's and warehouse's efforts to meet demand. s This represents demand that changes with different scenarios, assuming that all other parameters besides demand remain constant in each scenario. Use d t s This represents the minimum total cost given the scenario. Additionally, an extra parameter ω is introduced to represent the penalty cost for unmet demand deviations.

[0124] The objective function is:

[0125]

[0126] The constraints are:

[0127]

[0128] Considering that demand changes with different scenarios, many authors use stochastic optimization methods to model such problems. For strategic decisions, a "Here and Now" approach is used; for tactical and operational decisions, a "Wait and See" compensatory approach is employed, where once the strategic decision is made and demand becomes a reality, tactical and operational decisions are made over time. However, for our real-world material supply patterns, compensatory decision-making is clearly not optimal. Robust models incorporating stochastic demand scenarios can better address the uncertainty of future demand for enterprises and maintain the stability of decision-making patterns.

[0129] This invention proposes a comprehensive integrated optimization model for solving enterprise warehouse self-stocking, supplier collaborative stocking, inter-warehouse turnover and allocation, and the matching and delivery of goods and demand. The Bender decomposition method is used to solve this model. The objective function and constraints containing only design variables are incorporated into the main problem. Control variables, deviation variables, and auxiliary variables are used as decision variables for subproblems. Objective terms containing control variables and demand deviation variables are combined into the objective function of the subproblems, and constraints containing control variables, demand deviation variables, and auxiliary variables are used as constraint functions for the subproblems. The Bender decomposition algorithm with a single-cut mode is used to solve the model. The specific solution steps are as follows:

[0130] (1) Set the counter τ = 0; (2) Let θ be the expected value of the two-stage compensation objective function. If τ = 0, ignore θ; otherwise, add θ to the objective function of the main problem; (3) Solve the main problem; (4) Substitute each variable into each subproblem, find the corresponding relaxation for each subproblem under different scenarios, calculate the expected value of the two-stage compensation objective function of the subproblem, and the optimal dual variable value; (5) If If the solution is not found, stop and report the solution to the optimal master problem; otherwise, proceed to step 6 for iteration. (6) Generate the Bender cut and add it to the master problem. Let τ = τ + 1 and execute step 3 again.

[0131] Example:

[0132] With the increasing scale of power grid construction, the demand for material support is also constantly growing, leading to a more prominent supply-demand imbalance. Warehousing and distribution have become crucial links in logistics operations. Under the overall requirement of "intensive use of resources and professionalization of services," some enterprises have fully launched the construction of a modern warehousing and distribution system with provincial central warehouses as hubs and third-party logistics as support. Relying on provincial and municipal material allocation platforms, an intensive warehousing and distribution network with centralized storage, distribution, and allocation of materials has been built, further optimizing resource allocation and improving operational efficiency. Currently, the warehousing and distribution network has been adjusted from the original three-tiered network of "provincial central warehouse - municipal transit warehouse - county terminal warehouse" to a new warehousing network system of "provincial central warehouse - prefecture (city) transit warehouse - county terminal warehouse + specialized warehouse." Currently, there are as many as 75 warehouses of various types within a certain province, making the improvement of efficiency and effectiveness particularly important. Contractual inventory refers to inventory temporarily stored at corresponding suppliers during the agreed agreement period. It is suitable for materials with high standardization, strong versatility, and frequent use. It effectively innovates the traditional material procurement model in terms of ensuring supply, standardizing procurement, reducing costs, and improving efficiency. In recent years, the scope of power contract inventory implementation in a certain province has steadily expanded, with the number of procurement categories and contract suppliers increasing year by year. Contract inventory adopts centralized batch procurement, and its execution is subject to multi-dimensional and multi-level constraints. Therefore, despite the goal of balancing fairness and efficiency, forecast accuracy varies greatly, and the execution rate for many categories is poor, even falling below 10%, resulting in serious resource waste and impacting operational efficiency. A recent statistic from a certain power company (Company B) shows that from July 2022 to June 2023, 263 material subcategories were executed under distribution network contracts, with 192 subcategories showing a forecast execution deviation exceeding 50%. The overall execution rate for project units was 85.93%, with the overall execution rate for municipal companies at 77.14%, and five companies having an execution deviation within 20%. In light of this, we utilize integrated modeling concepts to propose a model for warehouse self-storage and preparation, contract inventory arrangement and preparation, and unified inventory allocation under multi-cycle scenarios. This model aims to achieve centralized planning and management of material resources and improve system efficiency.

[0133] By utilizing the concept of integrated modeling, this paper proposes a model for warehouse self-storage and stock preparation, contractual inventory arrangement and stock preparation, and unified inventory allocation under multi-cycle scenarios. The aim is to achieve centralized planning and management of material resources and improve system efficiency.

[0134] To reduce computational load, two warehouses were selected in each prefecture-level city: one as a transit warehouse and the other as a terminal warehouse. The provincial central warehouse was also retained, resulting in a total of |J| = 27 warehouses participating in the model calculations. Insulators were chosen as the material to consider supply and demand issues because they can be used in various transmission lines, are susceptible to environmental factors leading to performance degradation, have a high replacement frequency, and exhibit significant demand uncertainty. For example, from July 2022 to June 2023, the distribution network agreement inventory of insulators was used for both non-business expansion demand and business expansion demand scenarios. The execution rate of the non-business expansion demand agreement inventory was 69.91%, while the execution rate of the business expansion demand agreement inventory was 128.71%. The total amount of the two types of demand was 158 million yuan, and the overall agreement inventory execution rate was 79.18%. Regarding the insulator inventory suppliers, there are currently four suppliers in the system, located in different cities. Therefore, the number of potential suppliers participating in the model calculation is |I| = 4. We plan and design on an annual basis, dividing the entire planning period into 12 cycles, i.e., |T| = 12. It is assumed that the inventory suppliers are determined at the beginning of the year. At the beginning of the year, the values ​​of all design variables also need to be determined. In each cycle, after demand occurs, the quantity of insulators allocated and delivered by each warehouse and supplier to each demand point, as well as the turnover between warehouses, are determined based on the demand. Using districts and counties as the aggregation unit, the insulator demand for each unit for each of the 72 months (6 years) from 2018 to 2023 is statistically analyzed from the information system. This data is used to characterize uncertain demand; that is, each unit has a total of 6 data values ​​and corresponding statistical values ​​for the demand in each cycle (month), such as the maximum, minimum, and average values. Assuming these 6 data values ​​represent the demand under 6 different scenarios, that is, assuming each year from 2018 to 2023 is a scenario, then the number of scenarios |S| = 6. Encode the districts and counties of each city, and randomly select 4 districts and counties as the demand points of each city, then the total number of demand points |K| = 13 × 4 = 52.

[0135] The experimental design includes the following:

[0136] (1) Find the optimal solution for the three types of models proposed above based on the data and parameters of the above cases; (2) Study the computational efficiency of the Bender decomposition algorithm we use and compare it with the results obtained directly from the Gurobi solver language; (3) Under different inventory allocation and distribution modes, compare the impact of the penalty cost parameter ω of the unmet demand deviation on the total cost; (4) Under different inventory allocation and distribution modes, compare the impact of the budget uncertainty parameter γ on the total cost.

[0137] The optimal solution and algorithm efficiency were determined based on the preceding models, solution algorithms, parameters, and data. Two types of models (DM and RM) were solved using the Gurobi solver language and the Bender decomposition algorithm, respectively. In the deterministic model, the demand was calculated using the average of all scenarios; in the robust model, actual demand data from 72 cycles across 6 scenarios (6 years) were used. The main numerical settings for the warehouse are shown in the table below:

[0138]

[0139]

[0140] The main parameter settings for the contracted inventory supplier are as follows:

[0141]

[0142]

[0143] The experimental results are shown below:

[0144]

[0145] In comparison, the deterministic model yields the best objective value and has the lowest cost, with a calculation time of 3,960,587 seconds using the Gurobi solver language and 3,851,169 seconds using the Bender decomposition algorithm. The robust model, on the other hand, has the worst objective value and the highest cost, with a calculation time of 4,869,152 seconds using the Gurobi solver language and 4,660,252 seconds using the Bender decomposition algorithm. This is mainly because the robust model tolerates uncertainty, thus incurring higher costs, while the deterministic model can only meet specific requirements (mean point) and lacks flexibility. Similarly, considering the robustness of models across various scenarios, the objective value is moderate due to the balance between understanding robustness and model robustness. Overall, the optimal value obtained using the Bender decomposition algorithm is better than that obtained using the Gurobi solver language. The efficiency of the two different solution methods can be directly obtained from Table 3. The solution time for the deterministic model and the robust model is about the same; however, the robust model takes relatively more computation time because it includes 6 scenarios; but overall, in the above case scale, the computation time of the Bender decomposition algorithm is better than that of the Gurobi solver language programming.

[0146] To further analyze the penalty cost parameter ω and budget uncertainty parameter γ for unmet demand deviations. tThe impact of t = 1, 2, ... |T| on the total system cost is discussed in conjunction with real-world considerations of inventory priority. For inventory allocation and distribution, companies typically prioritize physical goods stored in their own warehouses, followed by supplier-mandated inventory. We investigate the sensitivity of total cost to the penalty cost parameter ω (for unmet demand) and the budget uncertainty parameter γ under two scenarios: prioritizing physical inventory allocation and eliminating physical inventory priority allocation. The results are as follows. Figure 3 As shown:

[0147] Depend on Figure 4 It is evident that regardless of whether physical inventory allocation is prioritized, the total cost consistently exhibits a changing trend with the penalty cost parameter ω for unmet demand. As the system increasingly focuses on penalizing unmet demand—that is, as the penalty becomes larger—the total cost initially decreases and then increases, reaching a minimum between approximately 45 and 50. This is because when ω is small, the system can reduce costs by ignoring some demand; however, when ω is large, the excessively high penalty cost means that even some unmet demand will still lead to an increase in total cost. Figure 4 One constant rule is that, regardless of how the penalty cost parameter ω for unmet demand changes, prioritizing the allocation of physical inventory will always lead to a reduction in total cost. This is because the system objective function takes into account the inventory holding cost of physical inventory, and factors such as the uneven distribution of contracted inventory suppliers in the system will cause significant differences in total cost.

[0148] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0149] It should be understood that the above detailed description of the technical solutions of the present invention with reference to preferred embodiments is illustrative and not restrictive. Those skilled in the art can modify the technical solutions described in the embodiments or make equivalent substitutions for some of the technical features based on reading this specification; however, these modifications or substitutions do not cause the essence of the corresponding technical solutions to depart from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for assessing the layout and influencing factors of power grid material logistics, characterized in that: The method includes: Construct a logistics network model with multiple vehicle depots, multiple tasks, and multiple vehicle types, and uniformly encode vehicle depots, distribution centers, and customer demand points as network nodes to form a weighted graph containing node sets and path arc sets; Based on the geographical location of the customer's demand point, time window constraints, and delivery priority, plan a closed path for vehicles to depart from the depot, be loaded at the distribution center, deliver to the customer's point, and return to the original depot. Define a time window penalty mechanism, and calculate the time penalty cost by combining the deviation between the actual time the vehicle arrives at the customer point and the time window; Based on the transportation costs, empty driving costs, and time penalty costs of multiple vehicle types, an optimization objective of minimizing total cost is established; By employing a multi-vehicle-yard scheduling strategy, vehicles of different models are allocated to perform delivery tasks, generating delivery plans that meet capacity, route continuity, and time constraints.

2. The method for assessing the layout and influencing factors of power grid material logistics according to claim 1, characterized in that: The network nodes are classified as follows: Customer node set (N): Composed of power grid material demand points, each node is associated with demand, service time window and priority label; Distribution Center Node (O): As the sole cargo transit node, it is responsible for receiving vehicles loaded in the parking lot and distributing goods; Vehicle depot node set (S): Each depot is equipped with multiple types of vehicles, including models with different load capacities, ranges and cost characteristics; The node coding rule is as follows: customer point number 1 to N, distribution center number N+1, parking lot number N+2 to N+S+1, where S is the total number of parking lots.

3. The method for assessing the layout and influencing factors of power grid material logistics according to claim 1, characterized in that: The goods allocation method of the distribution center includes: Real-time monitoring of vehicle arrival times at the distribution center from each depot; dynamic allocation of loading tasks to avoid congestion. Optimize cargo loading sequence based on vehicle capacity limitations and customer demand priorities; Radio frequency identification (RFID) technology is used to track materials, ensuring accurate loading and keeping inventory updated.

4. The method for assessing the layout and influencing factors of power grid material logistics according to claim 1, characterized in that: The time window penalty mechanism includes: For vehicles that arrive before the earliest service time, the penalty cost is calculated based on the waiting time and the penalty coefficient per unit time. For vehicles arriving after the latest service time, penalty costs are calculated based on a penalty coefficient weighted by the delay time and customer priority. The penalty cost is incorporated into the overall optimization objective, and the time deviation and transportation cost are balanced by adjusting the route planning.

5. The method for assessing the layout and influencing factors of power grid material logistics according to claim 1, characterized in that: The vehicle routing optimization algorithm includes: Initial path generation is based on genetic algorithm, and path combination is iteratively optimized through crossover and mutation operations; A tabu search strategy is introduced to avoid local optima and dynamically adjust the relationship between vehicle type and parking lot allocation. Urgent orders are handled using an insertion method, which dynamically inserts new tasks into the existing path and recalculates the time window constraints.

6. The method for assessing the layout and influencing factors of power grid material logistics according to claim 1, characterized in that: The multi-parking lot scheduling strategy includes: Based on the spatial distribution of parking lots and customer locations, service areas are divided and tasks are assigned to the nearest parking lot. In cross-regional demand scenarios, enabling collaboration mechanisms allows vehicle resources to be shared between parking lots; By using load balancing algorithms, the maximum workload of a single parking lot is limited to reduce operational risks.

7. The method for assessing the layout and influencing factors of power grid material logistics according to claim 1, characterized in that: The assessment of the influencing factors includes: Sensitivity analysis: Adjust the number of vehicle models, time window width, or penalty coefficient to quantify the impact on total cost; Comparative analysis: Evaluate the cost savings and carbon emission reduction effects of setting up temporary parking lots or introducing electric vehicles; Generate a decision matrix and recommend the optimal parking lot layout, vehicle configuration, and flexible time window strategy.

Images