Multi-objective task scheduling method for internet laundry center

Abstract

The application discloses a multi-target task scheduling method for an Internet laundry center, and aims at solving the problems of high energy consumption, great risk of delay and cross contamination caused by order dynamics and equipment fluctuation, coupling of peak-valley electricity price and heat load, and disinfection and high-temperature process constraints. The application generates equipment control instructions through hard constraint batch and soft constraint multi-target decomposition scheduling, ant colony optimization combined with variable neighborhood search, and dynamic rescheduling with incremental memory and local freezing, so that the technical effects of reducing electricity charges and energy consumption, reducing delay rate, balancing load, complying with disinfection and cooling window, and controlling disturbance are realized.

Description

Technical Field

[0001] This invention relates to the field of industrial internet production scheduling optimization, and particularly to a multi-objective task scheduling method for internet-connected laundry centers. Background Technology

[0002] With the increasing adoption of the Industrial Internet and smart manufacturing in the laundry and care industry, internet-connected laundry centers access orders via mobile devices, collect equipment status data via the Internet of Things, and deploy manufacturing execution systems or scheduling modules oriented towards batch processing and scheduling. Existing technologies often employ rule-driven methods such as due date priority, first-come-first-served, and shortest processing time, or optimize based on linear programming, integer programming, constrained programming, and general metaheuristics. Some solutions consider time-of-use electricity pricing and energy consumption estimation, but these simplify the handling of process windows and equipment changeover costs. Batch management typically merges tasks based on capacity and basic compatibility rules, and uses single-objective or finite multi-objective strategies to improve on-time performance or reduce costs.

[0003] However, existing technologies still have shortcomings:

[0004] First, the energy consumption and cost modeling is not fully coupled with peak and off-peak electricity prices and heat load, and lacks a detailed characterization of the additional energy consumption of program switching and the cooling window after high temperature, which results in limited cost optimization effect under price and temperature constraints during complex periods.

[0005] Secondly, the batch merging and splitting lacks systematic compatibility and cross-contamination control. The fabric type, color, allergen, stain type, and post-disinfection rinsing and isolation window are not uniformly entered into the mold, which can easily lead to batch merging being infeasible or quality risks.

[0006] Third, rescheduling under dynamic events is mostly a global reordering or simple rollback, lacking incremental memory and local freezing mechanisms. The scheme has a large disturbance ratio and insufficient stability, affecting customer satisfaction and execution controllability.

[0007] Therefore, there is an urgent need for a multi-objective scheduling method that simultaneously addresses energy consumption costs, cross-contamination control, and dynamic stability, and improves overall operational performance through the synergy of hard constraint pre-batching and soft constraint decomposition optimization. Summary of the Invention

[0008] One objective of this invention is to propose a multi-objective task scheduling method for internet-connected laundry centers. Addressing the shortcomings of existing technologies, such as insufficient modeling of peak-valley electricity pricing and heat load coupling, inadequate batch compatibility and cross-contamination control, and excessive rescheduling disturbances and poor stability under dynamic events, this invention proposes a technical solution that combines hard-constraint pre-batching, soft-constraint multi-objective decomposition with ant colony optimization, and variable neighborhood search fine-tuning. Furthermore, it employs incremental memory and local freezing for dynamic rescheduling under event triggering. This invention normalizes batches using constraint programming, constructs a weight vector sequence, and shares and strengthens pheromones among adjacent subproblems. Under constraints such as equipment availability, maintenance window, cooling window, and disinfection isolation, it jointly decides on equipment allocation and start time, and outputs equipment control commands. This invention achieves the technical effects of reducing electricity costs and energy consumption, decreasing delays and rescheduling rates, improving load balancing and customer satisfaction, and maintaining compliance with disinfection and high-temperature process windows.

[0009] A multi-objective task scheduling method for an internet-based laundry center according to an embodiment of the present invention includes:

[0010] S1. Receive user orders and equipment status data, and generate task feature vectors, equipment parameters, time period electricity price weighting function, compatibility matrix, disinfection switching parameters, program switching parameters and cooling time;

[0011] S2. Based on the above data, establish constraints such as non-mixing constraints, capacity constraints, post-disinfection rinsing and isolation window constraints, equipment availability and maintenance window constraints, and cooling window constraints after high-temperature programs. Use constraint programming to solve and normalize it into batches. Each batch includes task index, bound washing program, total weight, number of pieces, and deadline.

[0012] S3. Based on the batch, combined with equipment parameters, time period electricity price weighting function, current temperature, allowable temperature range and program switching parameters, calculate and output the initial pheromone matrix, heuristic factor initial value, weight vector sequence, memory retention coefficient, disturbance ratio threshold and stability threshold. The weight vector sequence is used to decompose the extension target, electricity cost and energy consumption weighted cost, equipment load balancing and customer satisfaction.

[0013] S4. Establish a multi-objective model that includes the delay objective, the weighted cost objective of electricity and energy consumption, the equipment load balancing objective, and the customer satisfaction objective. Decompose the model into sub-problems based on the weight vector sequence. In each sub-problem, use ant colony optimization to make joint decisions on equipment allocation and start time. Construct solutions according to the probability transition rules and maintain the non-dominated solution file using non-dominated sorting. Share and strengthen pheromones in the neighborhood of the weight vector sequence. Output candidate scheduling schemes, non-dominated solution files, and updated pheromone matrix.

[0014] S5. Perform variable neighborhood search with swapping, relocation and insertion around the candidate scheduling scheme, non-dominated solution file and updated pheromone matrix to reduce electricity cost and energy consumption weighted cost or reduce delay, and update the non-dominated solution file and pheromone matrix accordingly, and output the improved scheduling scheme, updated non-dominated solution file and updated pheromone matrix again.

[0015] S6. When a new order arrives, equipment fails, or the time period of the time-based electricity price weighting function is switched, identify the affected tasks and frozen batches, perform local rearrangement only on the affected tasks, and perform incremental pheromone updates by weighting the historical pheromone on the device number and start time of the relevant tasks according to the memory retention coefficient, limit the scheme disturbance ratio to not exceed the disturbance ratio threshold, update the non-dominated solution file at the same time, and output the updated scheduling scheme and the final non-dominated solution file.

[0016] S7. Terminate based on stability threshold and disturbance ratio threshold, select the final scheduling scheme from the updated scheduling scheme, and generate equipment control instructions.

[0017] Optionally, step S1 specifically includes:

[0018] The user order data is validated and encoded. A task feature vector is generated according to the fabric type, color, allergen, stain type, weight, number of pieces, washing program requirements, required completion time and user level. The fabric type, color, allergen, stain type and washing program requirements in the task feature vector are represented by a preset enumeration code, the weight is in kilograms, the number of pieces is in pieces, and the required completion time is represented by a timestamp.

[0019] Collect equipment status and master data to generate equipment parameters. The equipment parameters include capacity, availability, maintenance window, unit power, load rate, current temperature and allowable temperature range. The capacity is derived from the equipment master data, the availability is indicated by online status signals, the maintenance window is represented by a time period, the unit power is measured in kilowatts, the load rate is calculated by dividing the current loaded weight by the capacity, the current temperature is measured in degrees Celsius, and the allowable temperature range is represented by the upper and lower limits.

[0020] Based on the power supplier's time-of-use electricity price table, a time-of-use electricity price weighting function is generated. The electricity price for each time period is normalized to a weight of zero to one according to the highest and lowest electricity price range, which is used to reflect the level of electricity price in different time periods.

[0021] A compatibility matrix is ​​generated based on the task feature vector. The compatibility rules of fabric type, color, allergen, stain type and washing program requirements are used to mark whether the fabrics can be mixed or not. If any dimension is incompatible, the fabrics can be marked as not to be mixed.

[0022] Disinfection switching parameters are generated based on the washing process management rules and equipment instructions. These parameters include the rinsing time, isolation time, and release conditions after the disinfection process is completed, which are used to constrain the timing of equipment reuse after disinfection.

[0023] Based on the equipment instructions and program configuration, program switching parameters are generated. These parameters include program switching time and additional energy consumption for switching, which are used to characterize the switching cost between different washing programs.

[0024] The cooling time is estimated based on the current temperature and the allowable temperature range in the equipment parameters. When the current temperature is higher than the upper limit of the allowable temperature, the cooling time is calculated according to the temperature difference ratio. When the current temperature is not higher than the upper limit of the allowable temperature, the cooling time is recorded as zero. The cooling time is used to characterize the duration of the cooling window after the high temperature program.

[0025] Terminology definition:

[0026] The user order data is a collection of order information including fabric category, color, allergen, stain type, weight, number of pieces, washing program requirements, required completion time, and user level.

[0027] The equipment status data is a set of equipment operation data that includes availability status, current load weight, current temperature, and online status signals.

[0028] The device master data is a collection of static attributes and configuration data of the device, including device number, capacity, power per unit, allowable temperature range and supported washing programs, etc.

[0029] The online status signal is a status identifier used to indicate whether the device is in a network-connected, communicable, and operable state.

[0030] The maintenance window is the non-processable time period occupied by planned equipment maintenance, cleaning, or disinfection.

[0031] The time-of-use electricity price list is a list of time-of-use electricity prices published by the power supplier, which gives the corresponding electricity price level for each time period.

[0032] The compatibility rules are preset judgment criteria for fabric type, color, allergen, stain type and washing program requirements, used to determine whether tasks can be mixed for washing.

[0033] The user level is a level identifier used to represent the priority of order services;

[0034] The program configuration is a set of parameter settings for the washing program, used to determine the program switching time and the additional energy consumption for switching;

[0035] The rinsing time is the shortest duration of the rinsing process that must be performed after the disinfection procedure to remove residual disinfectant.

[0036] The isolation time is the minimum idle time that the equipment must remain in before being put into a non-disinfection task after the disinfection process is completed.

[0037] The release conditions are a set of conditions that allow the equipment to undertake non-disinfection tasks after the rinsing time and isolation time are met and the necessary tests are completed.

[0038] The temperature difference ratio is the ratio of the temperature difference between the current temperature and the upper limit of the allowable temperature to the width of the allowable temperature range, and is used to linearly estimate the cooling time.

[0039] Optionally, step S2 specifically includes:

[0040] Based on the fabric category, color, allergen, stain type and washing procedure requirements in the task feature vector, a non-mixable washing constraint is established according to the compatibility matrix. Tasks marked as non-mixable in the compatibility matrix are prohibited from being included in the same batch.

[0041] Establish capacity constraints based on the capacity in the equipment parameters to ensure that the total weight of each batch does not exceed the capacity.

[0042] Establish post-disinfection rinsing and isolation window constraints based on disinfection switching parameters, so that tasks containing disinfection procedures can meet the post-disinfection rinsing and isolation requirements at the batch level;

[0043] Based on the availability status and maintenance window in the equipment parameters, establish constraints on the availability status and maintenance window of the equipment to ensure that the deadline of the batch is not earlier than the end time of the maintenance window and that the washing program corresponding to the batch is processable when the equipment is available.

[0044] The cooling window constraint after the high-temperature program is established based on the cooling time, so that the batch containing the high-temperature washing program meets the cooling window requirement in subsequent scheduling.

[0045] Under the above constraints, the task feature vector is solved by batching and splitting through constraint programming. Priority is given to merging tasks that are marked as washable in the compatibility matrix and have the same washing program requirements. When capacity constraints or deadline conflicts make it infeasible, batching is performed and the relevant constraints are recalculated.

[0046] The feasible solution is normalized into batches, which include the corresponding task index, the bound washing program, the total weight, the number of pieces, and the deadline.

[0047] Terminology definition:

[0048] The non-washable constraint is a prohibition rule for batching tasks based on the compatibility matrix. If there is incompatibility in any dimension, the tasks are prohibited from being merged into the same batch.

[0049] The capacity constraint is a constraint that limits the total weight of each batch to no more than the target equipment capacity;

[0050] The post-disinfection rinsing and isolation window constraints are as follows: when a batch includes a disinfection procedure, the equipment must first meet the minimum rinsing time and maintain the minimum isolation time, and subsequent non-disinfected batches are only allowed to be scheduled after the release conditions are met.

[0051] The equipment availability status and maintenance window constraints are time constraints that allow batches to be scheduled only when the equipment is available and its planned processing time does not overlap with the maintenance window.

[0052] The cooling window constraint after the high-temperature program means that after the equipment performs the high-temperature washing program, it must undergo a non-processable period of not less than the cooling time. During this period, it is prohibited to schedule batches with an upper temperature limit lower than the current temperature of the equipment.

[0053] The high-temperature washing program is a type of washing program where the process setting temperature is higher than the upper limit of the equipment's allowable temperature range or higher than the upper limit of the temperature of subsequent batches.

[0054] The constraint programming method is a computational method that describes batching and splitting problems using a constraint satisfaction / optimization model, and uses a constraint solver to search for solutions that satisfy all constraints.

[0055] The batching process refers to the process of merging multiple tasks into a single batch while satisfying the compatibility matrix and various constraints.

[0056] The batch splitting process refers to the process of splitting a batch into multiple sub-batches and recalculating the relevant constraints when the batching result violates the capacity constraint or a deadline conflict occurs.

[0057] The deadline is the latest time that the batch must be completed, and it is the earliest value among the required completion times of the tasks contained in the batch.

[0058] The deadline conflict refers to a situation where a batch cannot be completed before its deadline due to constraints of equipment and process windows.

[0059] The feasible solution is the batch / batch combination result that simultaneously satisfies all constraints such as non-mixing, capacity, post-disinfection rinsing and isolation window, equipment availability and maintenance window, and post-high temperature cooling window.

[0060] The batch is a standardized processing unit used for subsequent scheduling, including task index, bound washing program, total weight, number of pieces and deadline;

[0061] The task index is a unique set of identifiers that identify each original task within a batch;

[0062] The bound washing program is a specific washing program determined at the batch level and executed with the batch;

[0063] The processability refers to the conditional state under which the equipment can perform batch processing within its supported washing programs, capacities, and permissible temperature ranges.

[0064] The disinfection procedure is a specific category of procedures that uses chemical or high-temperature methods to disinfect during the washing process, and must comply with rinsing and isolation requirements.

[0065] Optionally, step S3 specifically includes:

[0066] Based on the batch's task index, bound washing program, total weight, number of pieces, and deadline, and combined with the availability status and maintenance window in the equipment parameters, the earliest processing time for each batch on each equipment is determined to estimate the expected completion time.

[0067] Calculate the projected delay cost based on the estimated completion time and the deadline, where the projected delay cost is the non-negative value of the time difference between the estimated completion time and the deadline.

[0068] Based on the unit power and time-period electricity price weighting function in the equipment parameters, the weighted cost of electricity and energy consumption is estimated within the expected processing time window of the batch. The weighted cost of electricity and energy consumption is obtained by multiplying the unit power and processing time and weighting it with the time-period electricity price weighting function.

[0069] The thermal penalty is calculated based on the current temperature, allowable temperature range and cooling time, and the switching penalty is calculated in combination with the program switching time and additional switching energy consumption in the program switching parameters. This is used to characterize the cooling impact and program switching cost after high temperature.

[0070] The load penalty is calculated based on the load rate in the equipment parameters to characterize the impact of the current load on subsequent batches;

[0071] Heuristic factor initial values ​​are generated by weighted aggregation of predicted delay costs, electricity and energy consumption weighted costs, thermal penalties and switching penalties, and load penalties. These heuristic factor initial values ​​are normalized in the form of the inverse of the aggregated cost for each batch and each equipment combination.

[0072] An initial pheromone matrix is ​​generated based on the processability of the batch and the availability of the equipment. The initial pheromone matrix is ​​initially set with a uniform value for each batch and each equipment combination and is fine-tuned according to the relative magnitude of the initial values ​​of the heuristic factors.

[0073] Based on four types of objectives—extended objectives, electricity and energy consumption weighted costs, equipment load balancing, and customer satisfaction—a weighted vector sequence is constructed. The weighted vector sequence consists of several weight vectors, and the sum of the four weights of each weight vector is one. This sequence is used to decompose multiple objectives.

[0074] Set a memory retention coefficient to weight historical pheromones and perform incremental updates in subsequent dynamic rescheduling;

[0075] Set disturbance ratio threshold and stability threshold to limit the change ratio of scheduling scheme and judge the stability of non-dominated solution files in subsequent dynamic rescheduling and termination judgment.

[0076] Output the initial pheromone matrix, initial values ​​of heuristic factors, weight vector sequence, memory retention coefficient, perturbation ratio threshold, and stability threshold.

[0077] Terminology definition:

[0078] The earliest processing time is the earliest point in time when the batch can begin processing on the target equipment under the constraints of equipment availability and maintenance window.

[0079] The estimated completion time is the estimated completion time of this batch, assuming that the process windows such as cooling, disinfection, rinsing and isolation and program switching are met, starting from the earliest processing time.

[0080] The estimated processing time window is the time interval between the estimated start time and the estimated completion time of the batch on the target equipment;

[0081] The predicted delay cost is the non-negative time difference between the expected completion time and the batch deadline.

[0082] The time-of-use electricity price weighting function is a weighting function that normalizes the time-of-use electricity price to the range of zero to one based on the highest and lowest electricity price intervals, and is used to reflect the level of electricity prices in different time periods.

[0083] The electricity cost and energy consumption weighted cost is a cost measure obtained by multiplying the unit power by the processing time within the expected processing time window and weighting it with a time-period electricity price weighting function.

[0084] The heat penalty is a cost metric corresponding to the relationship between the current temperature and the allowable temperature range, and the impact of cooling time on scheduling feasibility. It increases with the increase of temperature difference exceeding the limit and cooling time.

[0085] The switching penalty is a cost metric that maps the program switching time between different washing programs to the additional energy consumption of switching.

[0086] The load penalty is a cost metric corresponding to the impact of the current load rate of the equipment on the processability and delay risk of subsequent batches, and it increases as the load rate increases;

[0087] The aggregated cost is a comprehensive evaluation value obtained by weighting and summing the predicted delay cost, electricity cost and energy consumption weighted cost, thermal penalty, switching penalty and load penalty according to the set weights.

[0088] The initial value of the heuristic factor is a heuristic preference factor obtained by taking the reciprocal of the aggregation cost and normalizing it over the combination of batch and equipment.

[0089] The initial pheromone matrix is ​​a pheromone concentration matrix initialized on the combined link of batch and device, set with a uniform initial value and fine-tuned according to the relative magnitude of the initial value of the heuristic factor;

[0090] The weight vector sequence is a set of weight vectors used to decompose four types of objectives: extension target, electricity cost and energy consumption weighted cost, equipment load balancing and customer satisfaction. The sum of the four weights of each weight vector is one.

[0091] The equipment load balancing is a target indicator used to measure the uniformity of utilization of each piece of equipment. The target performance improves as the difference in utilization between equipment decreases.

[0092] The customer satisfaction is a target indicator used to measure the quality of customer service provided by the scheduling results, taking into account the degree of delay, the reassignment ratio and the weight of user level.

[0093] The memory retention coefficient is a coefficient used to retain and weight historical pheromones during dynamic rescheduling, and is used to control the superposition ratio of historical pheromones and current incremental pheromones.

[0094] The disturbance ratio threshold is an upper limit threshold that restricts the number of batches that change device allocation or start time in a single reordering to the proportion of the number of batches participating in scheduling.

[0095] The stability threshold is an upper limit threshold for determining whether the maximum change in aggregate cost of representative solutions in the final non-dominated solution archive of adjacent rescheduling rounds is acceptable.

[0096] Optionally, step S4 specifically includes:

[0097] Based on the initial pheromone matrix, heuristic factor initial values, and weight vector sequence, and combined with batch and equipment parameters, time period electricity price weight function, current temperature, allowable temperature range, cooling time, and program switching parameters, a multi-objective optimization model is established, which includes minimizing delay, minimizing the weighted cost of electricity and energy consumption, balancing equipment load, and maximizing customer satisfaction. This model is used to evaluate the comprehensive performance of batch start time and sequence decision-making on each device.

[0098] The multi-objective optimization model is decomposed into multiple sub-problems based on the weight vector sequence, and ant colony optimization is run in each sub-problem. A candidate set is generated according to the combination of batch and equipment. The candidate set only includes the combination of equipment and start time that satisfies the constraints of equipment availability and maintenance window, earliest processing time, cooling time and program switching parameters. At the same time, the batches containing the disinfection program follow the requirements of rinsing and isolation after disinfection.

[0099] During the solution construction process, the probability transfer rule is constructed based on the weighted initial value of the heuristic factor and the pheromone. Equipment allocation and start time are jointly decided in batches. Feasibility repair is performed for cases where time conflicts or windows are not met, including postponing the start time or replacing available equipment to maintain constraint satisfaction.

[0100] After completing the scheduling construction of a subproblem, calculate four types of indicators: delay, weighted cost of electricity and energy consumption, equipment load balancing and customer satisfaction. Perform non-dominated sorting and update the non-dominated solution file to retain the representative solution of the current iteration.

[0101] Local pheromone updates are performed on the selected equipment and start time connection to enhance exploration; global pheromone updates are performed on representative solutions in the non-dominated solution archive to enhance good decision-making; and pheromone is shared and enhanced in the neighborhood of the weight vector sequence to improve search synergy between adjacent subproblems.

[0102] In each iteration, the heuristic factor is recalculated based on the time period electricity price weighting function, the current temperature and the allowable temperature range, the cooling time and program switching parameters, and the load rate in the equipment parameters. This is used to dynamically reflect the changes in forecast delay costs, electricity and energy consumption weighted costs, thermal penalties, switching penalties, and load penalties.

[0103] Output candidate scheduling schemes, non-dominated solution files, and an updated pheromone matrix. The candidate scheduling schemes consist of a mapping from batch to device number and start time.

[0104] Terminology definition:

[0105] The multi-objective optimization model is an optimization model that simultaneously includes minimizing delay, minimizing the weighted cost of electricity and energy consumption, balancing equipment load, and maximizing customer satisfaction. It is used to evaluate the comprehensive performance of batch start time and sequence decision-making on each device.

[0106] The sub-problem is an independent optimization instance formed by weighted decomposition of multiple objectives based on a single weight vector in the weight vector sequence.

[0107] The ant colony optimization is a metaheuristic optimization algorithm based on constructing probabilistic selections using pheromones and heuristic factors, and iteratively updating pheromones to improve solution quality.

[0108] The candidate set is a set of combinations of equipment and start time that only meet the constraints of equipment availability and maintenance window, earliest processing time, cooling time and program switching in the combination of batch and equipment.

[0109] The probability transition rule is a rule that calculates the probability of the combination of device and start time being selected based on the weighted result of pheromone and heuristic factor and combined with constraint feasibility.

[0110] The equipment allocation is a decision to assign batches to specific equipment to perform corresponding washing programs;

[0111] The joint decision-making process is a decision-making process that simultaneously determines the allocation of equipment and the start time for a batch;

[0112] The time conflict refers to a situation where the processing time intervals of two or more batches on the same equipment overlap or overlap with the maintenance, cooling, or isolation windows.

[0113] The window does not meet the requirements of the process and resource windows, such as the planned start or completion time not meeting the requirements of the cooling window, disinfection and rinsing and isolation window, maintenance window, or temperature allowable range;

[0114] The feasibility repair is a process of making an infeasible solution in the construction satisfy all constraints by operations such as delaying the start time or replacing available equipment.

[0115] The non-dominated sorting is a process of sorting the solution set according to the Pareto dominance relation in order to identify solutions that are not superior to other solutions on all objectives at the same time.

[0116] The non-dominated solution archive is a storage structure used to persistently store the set of non-dominated solutions obtained in the current iteration;

[0117] The representative solution is a representative solution selected from the non-dominated solution archive for evaluation, updating, and pheromone enhancement.

[0118] The device and start time are linked as a batch-device-start time combination unit represented in the pheromone matrix;

[0119] The local pheromone update is to adjust the pheromone concentration on the connection between the device used to construct or improve the solution and the start time to enhance the update operation of local exploration;

[0120] The global pheromone update is an update operation that strengthens pheromones on relevant links based on representative solutions in the non-dominated solution archive to guide the global search.

[0121] The neighborhood of the weight vector sequence is a set of adjacent weight vectors selected in the weight space with a certain weight vector as the center and according to a preset distance or perturbation range;

[0122] The shared and enhanced pheromone refers to the operation of propagating the same or similar connections within the neighborhood of the weight vector sequence and increasing the pheromone level to improve cross-subproblem search collaboration.

[0123] The iteration is one algorithmic loop from solution construction, evaluation, pheromone update to archive maintenance;

[0124] The candidate scheduling scheme is a list of alternative scheduling schemes consisting of the correspondence between batch, equipment number, and start time.

[0125] The mapping is a set of correspondences between batch identifiers and device numbers and start times;

[0126] The sequence decision is a decision-making process that determines the processing order of batches on the same equipment and determines the start time of each batch accordingly.

[0127] Optionally, step S5 specifically includes:

[0128] Based on the candidate scheduling scheme, the non-dominated solution file, and the updated pheromone matrix, three types of neighborhoods are constructed around the candidate scheduling scheme and applied sequentially. The three types of neighborhoods include swap operation, relocation operation, and insertion operation. The swap operation is to swap the start time of two batches on the same device or to swap the device allocation and start time combination of two batches on different devices. The relocation operation is to move a batch to a new position in the sequence of the same device or to move the batch to another device and reset the start time under the premise of satisfying the constraints. The insertion operation is to insert a batch in the idle time window of the device and postpone or fine-tune the start time of the adjacent batches.

[0129] After each neighborhood construction, a feasibility check is performed to ensure that the equipment availability status and maintenance window constraints, the earliest processing time constraint, the cooling time constraint, the program switching parameter constraint, and the rinsing and isolation requirements after disinfection are simultaneously met. For cases where the constraints are not met, a feasibility repair is performed, including extending the start time or replacing available equipment to maintain the constraint satisfaction.

[0130] The aggregate cost is calculated for the neighborhood solutions that pass the feasibility check. The aggregate cost is represented as a weighted sum of predicted delay costs, electricity and energy consumption weighted costs, thermal penalties and switching penalties, and load penalties, and is used to evaluate the overall performance of the neighborhood solutions.

[0131] When the aggregation cost of a neighborhood solution is lower than that of the current solution or is not dominated by any solution in the current non-dominated solution file during the non-dominated sorting, the neighborhood solution is accepted as an improved and replaced candidate scheduling scheme. At the same time, the non-dominated sorting is performed to update the non-dominated solution file, the dominated solution is removed and the representative solution is retained.

[0132] Pheromones are enhanced on the improved devices and the connection to the start time. On the unimproved connections, historical pheromones are weighted according to the memory retention coefficient and the improvement increment is added to form an updated pheromone matrix. In the above neighborhood loop, when no improvement is achieved in a single neighborhood operation, the process moves to the next neighborhood. Fine-tuning ends when no improvement is achieved within a preset number of attempts.

[0133] Output an improved scheduling scheme, an updated non-dominated solution file, and a further updated pheromone matrix.

[0134] Terminology definition:

[0135] The neighborhood is a set of candidate solutions formed around the candidate scheduling scheme through limited local structural adjustments;

[0136] The swapping operation is an operation that swaps two batches in a neighborhood, including swapping the start time of two batches on the same device, or swapping the device allocation and start time combination of two batches on different devices.

[0137] The relocation operation is an operation that adjusts the position of a single batch in the neighborhood, including moving the batch to a new position in the sequence of the same device, or moving the batch to another available device and resetting the start time, provided that the constraints are met.

[0138] The insertion operation is to insert a batch within the device's idle time window and to postpone or fine-tune the start time of adjacent batches.

[0139] The device idle time window is the time interval during which no batches are assigned in the current device sequence;

[0140] The equipment sequence is a batch sequence arranged on the same equipment according to the order of processing;

[0141] The start time is the planned start time for the batch to begin processing on the target equipment;

[0142] The feasibility check is a process of verifying the overall satisfaction of the neighboring solution with constraints such as equipment availability and maintenance window, earliest processing time, cooling time, program switching parameters, and disinfection, rinsing and isolation.

[0143] The neighborhood solution is a scheduling scheme generated from candidate scheduling schemes through exchange, relocation, or insertion operations.

[0144] The pheromone enhancement is an update operation that increases the pheromone concentration on the corresponding device and start time connection to enhance the probability of the connection being selected.

[0145] The improvement increment is an increment value calculated based on the overall performance improvement of the neighborhood solution relative to the current solution and used for pheromone enhancement;

[0146] The updated pheromone matrix is ​​the latest pheromone concentration matrix formed after neighborhood fine-tuning based on the incremental improvement of pheromone enhancement and memory retention coefficients;

[0147] The neighborhood loop is an iterative process that sequentially applies three types of neighborhoods: swap, relocation, and insertion, and evaluates and updates them after each operation.

[0148] The preset number of attempts is the maximum number of attempts allowed to run without improvement under a single neighborhood category;

[0149] The improved scheduling scheme is the current optimal or non-dominated scheduling scheme that is accepted and replaces the original candidate scheduling scheme after neighborhood operation and feasibility check.

[0150] Optionally, step S6 specifically includes:

[0151] Set event trigger conditions to monitor new order arrivals, equipment failures, and time period switching of the time period electricity price weighting function. When any of the conditions are met, dynamic rescheduling will be initiated.

[0152] In the improved scheduling scheme, affected tasks are identified, and batches that conflict with the resources of the faulty equipment in the batches allocated to the faulty equipment and their subsequent sequences are marked as affected tasks. Batches located within the time period switching window are also marked as affected tasks. New orders are processed into batches based on the compatibility matrix, capacity constraints and disinfection switching parameters and then incorporated into the affected tasks.

[0153] The batches to be frozen are defined as follows: batches that have been started, batches that are close to the deadline, and batches that contain disinfection procedures. Among them, batches that are close to the deadline are those whose required completion time and the current time do not exceed the preset time. Frozen batches are not included in this reordering.

[0154] Local rearrangement is performed only for the affected tasks, taking into account the constraints of equipment availability and maintenance window, earliest processing time, cooling time, program switching parameters, and post-disinfection rinsing and isolation requirements. Local solutions are constructed on the connection between the equipment number and start time of the affected tasks according to the probability transition rules defined in S4. Feasibility repairs are performed for cases where time conflicts or windows are not met, including postponing the start time or replacing available equipment to keep the constraints satisfied. The overall performance of the local solution is evaluated by predicting the delay cost, the weighted cost of electricity and energy consumption, the thermal penalty and switching penalty, and the load penalty.

[0155] Incremental pheromone updates are performed on the adjusted device number and start time link, historical pheromone is weighted according to the memory retention coefficient and enhanced with the current iteration increment, existing pheromone levels are retained for unadjusted links, and the increment size is determined by the overall performance improvement of the affected task.

[0156] Calculate the scheme disturbance ratio, which is the ratio of the number of batches that change equipment allocation or start time in this reordering to the number of batches participating in scheduling. When the scheme disturbance ratio reaches the disturbance ratio threshold, stop further reordering and roll back the last modification to ensure that the disturbance is under control.

[0157] The solution obtained from the local rearrangement is sorted non-dominated and the non-dominated solution file is updated. Dominated solutions are removed and representative solutions are retained until no further overall performance improvement is achieved within the scope of the affected task or the perturbation ratio of the scheme reaches the perturbation ratio threshold. The updated scheduling scheme and the final non-dominated solution file are then generated.

[0158] Terminology definition:

[0159] The event triggering conditions are the judgment rules for monitoring the arrival of new orders, equipment failures, and time period switching of the electricity price weight function, and initiating dynamic rescheduling accordingly.

[0160] The dynamic rescheduling is a scheduling process that, when the event triggering conditions are met, only makes local scheduling adjustments to the affected tasks and updates the pheromone and non-dominated solution files.

[0161] The arrival of a new order refers to an event in which order information that has been added within the current scheduling cycle and has not yet been batched enters the system.

[0162] The equipment failure refers to an event in which equipment becomes unavailable or its performance degrades, making it impossible to execute the original processing plan according to the constraints.

[0163] The event that the time period switching to time-of-use pricing causes a boundary crossing that results in a change in the weight range of the time period pricing weight function;

[0164] The time period switching window is a time interval set around the time period switching boundary, used to identify batches affected by the switching;

[0165] The affected tasks are the set of batches that need to be rescheduled due to equipment failure, time period switching, or the incorporation of new orders;

[0166] The resource conflict refers to a situation where there is overlap or mutual exclusion in the time, capacity, or process window of the target equipment, making it impossible to satisfy them simultaneously.

[0167] The frozen batch is the set of batches that do not participate in the current rescheduling due to status or time limit restrictions during dynamic rescheduling;

[0168] The initiated batch refers to the batch that has already started processing on the equipment and cannot be interrupted or rolled back;

[0169] The batches nearing their deadline are those whose required completion time and the current time difference do not exceed a preset duration.

[0170] The partial rescheduling is a scheduling operation that adjusts only the equipment allocation and start time of the affected tasks while keeping the frozen batch unchanged.

[0171] The local solution is a feasible scheduling scheme constructed for the affected task subset based on the connection between device number and start time.

[0172] The incremental pheromone update is an update operation that weights historical pheromones based on the memory retention coefficient and adds the current iteration increment to the connection between the adjusted device number and the start time.

[0173] The historical pheromone refers to the pheromone concentration on the device number and start time that was retained in the previous or earlier iteration;

[0174] The disturbance ratio of the scheme is the proportion of the number of batches that change the equipment allocation or start time in this rescheduling to the total number of batches participating in the scheduling.

[0175] The rollback is a process that cancels the most recent modification to the equipment allocation or start time to bring the disturbance under control when the disturbance ratio of the scheme reaches the disturbance ratio threshold.

[0176] The scope of the affected tasks is the boundary of the set of batches marked as affected tasks in the current dynamic rescheduling;

[0177] The overall performance improvement is a measure of the improvement in the quality of local solutions of the affected task based on the decrease in aggregation cost or the degree of improvement in non-dominated ranking, and is used to determine the magnitude of the pheromone increment.

[0178] The updated scheduling scheme is the currently accepted scheduling scheme after completing local rearrangement, pheromone update and non-dominated sorting;

[0179] The final non-dominated solution file is the non-dominated solution file retained at the end of this round of dynamic rescheduling.

[0180] Optionally, step S7 specifically includes:

[0181] The stability of the final non-dominated solution file is assessed. The maximum change in the aggregate cost of the representative solutions in the final non-dominated solution file between the previous round of rescheduling and the current round of rescheduling is used as the stability measure. When the maximum change is not greater than the stability threshold, the file is considered stable.

[0182] The disturbance ratio of the updated scheduling scheme is calculated by dividing the number of reassigned batches in the updated scheduling scheme by the total number of batches. When the disturbance ratio is not greater than the disturbance ratio threshold, the disturbance is considered acceptable.

[0183] Under the condition that both the stability threshold and the disturbance ratio threshold are met, the scheduling scheme with the lowest aggregation cost is selected from the updated scheduling schemes as the final scheduling scheme, and the final scheduling scheme is converted into equipment control instructions. The equipment control instructions include equipment number, washing program, start time and batch allocation, and are used as new data output for this step for issuance and execution.

[0184] Terminology definition:

[0185] The stability judgment is a process of determining whether the changes in the final non-dominated solution file are within an acceptable range based on stability metrics and stability thresholds.

[0186] The stability metric is the maximum change in aggregate cost of representative solutions in the final non-dominated solution archives of the previous and current rounds of rescheduling.

[0187] The maximum change is the result of calculating the absolute value of the aggregate cost difference between the representative solution sets of the two rounds of final non-dominated solution files and taking the maximum value of the result.

[0188] The reassignment batch refers to the batch whose equipment allocation or start time has changed in the updated scheduling scheme compared to the previous round of scheduling;

[0189] The disturbance can be accepted as a state determination that the disturbance ratio of the scheme is not greater than the disturbance ratio threshold;

[0190] The final scheduling scheme is to select the scheduling scheme with the minimum aggregation cost from the updated scheduling schemes, while simultaneously satisfying the stability threshold and the disturbance ratio threshold.

[0191] The equipment control command is a set of control information used to drive the equipment to execute according to the final scheduling plan, including equipment number, washing program, start time and batch allocation;

[0192] The batch allocation is a set of allocation relationships that associate each batch with the target device and its planned start time.

[0193] The beneficial effects of this invention are:

[0194] 1. Based on multi-objective decomposition and ant colony optimization considering peak and valley electricity price weights, unit power, thermal penalties and switching penalties, combined with variable neighborhood search fine-tuning, we can achieve a significant reduction in electricity cost and energy consumption weighted cost, a decrease in the proportion of delayed batches, and improve equipment load balancing and customer satisfaction.

[0195] 2. By using a compatibility matrix and disinfection switching parameters to pre-set the batching, and overlaying capacity constraints, cooling windows and equipment availability windows, it is ensured that only washable tasks are batched, effectively avoiding cross-contamination, and that rinsing and isolation after disinfection and cooling after high temperature all meet process requirements.

[0196] 3. When triggered by new orders, equipment failures, or electricity price time shifts, dynamic rescheduling with incremental memory and partial freezing is adopted, and the adjustment range is constrained by the scheme disturbance ratio threshold and stability threshold. This results in faster rescheduling, controlled disturbances, and improved scheduling scheme stability. Attached Figure Description

[0197] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0198] Figure 1 This is a flowchart of a multi-objective task scheduling method for internet-based laundry centers proposed in this invention.

[0199] Figure 2 This is a schematic diagram of the "Multi-device Time Window Compliance and Peak-Valley Electricity Price Coordinated Scheduling" of the present invention. The bottom of the diagram represents the horizontal time axis 20; the top represents the electricity price band, with the peak price zone 31 (represented by dense diagonal lines) on the left and the valley price zone 32 (represented by sparse dots) on the right; the middle section, from top to bottom, represents multiple equipment lanes 11-16. Rectangles 41a, 41b, and 41c within each lane represent batch blocks after batch standardization, all arranged within the valley price zone 32; in the peak price zone 31, only dotted lines indicate avoided positions that are not used for production scheduling; vertical... The direct mask represents the process and resource windows: cooling window 51 (diagonal line), rinsing window 52 (dot matrix), isolation window 53 (grid), and maintenance window 54 (solid black bar). Batch blocks avoid maintenance and prohibition windows and meet the corresponding timing constraints. The program switching position between adjacent batches is indicated by symbol 55. This figure intuitively shows that, under the premise of meeting the process window constraints such as cooling, rinsing, isolation, and maintenance, batches are preferentially scheduled in the valley price zone 32, achieving the comprehensive technical effect of peak avoidance and valley production and energy cost reduction. Detailed Implementation

[0200] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0201] refer to Figure 1 A multi-objective task scheduling method for internet-based laundry centers, comprising:

[0202] S1. Receive user orders and equipment status data, and generate task feature vectors, equipment parameters, time period electricity price weighting function, compatibility matrix, disinfection switching parameters, program switching parameters and cooling time;

[0203] S2. Based on the above data, establish constraints such as non-mixing constraints, capacity constraints, post-disinfection rinsing and isolation window constraints, equipment availability and maintenance window constraints, and cooling window constraints after high-temperature programs. Use constraint programming to solve and normalize it into batches. Each batch includes task index, bound washing program, total weight, number of pieces, and deadline.

[0204] S3. Based on the batch, combined with equipment parameters, time period electricity price weighting function, current temperature, allowable temperature range and program switching parameters, calculate and output the initial pheromone matrix, heuristic factor initial value, weight vector sequence, memory retention coefficient, disturbance ratio threshold and stability threshold. The weight vector sequence is used to decompose the extension target, electricity cost and energy consumption weighted cost, equipment load balancing and customer satisfaction.

[0205] S4. Establish a multi-objective model that includes the delay objective, the weighted cost objective of electricity and energy consumption, the equipment load balancing objective, and the customer satisfaction objective. Decompose the model into sub-problems based on the weight vector sequence. In each sub-problem, use ant colony optimization to make joint decisions on equipment allocation and start time. Construct solutions according to the probability transition rules and maintain the non-dominated solution file using non-dominated sorting. Share and strengthen pheromones in the neighborhood of the weight vector sequence. Output candidate scheduling schemes, non-dominated solution files, and updated pheromone matrix.

[0206] S5. Perform variable neighborhood search with swapping, relocation and insertion around the candidate scheduling scheme, non-dominated solution file and updated pheromone matrix to reduce electricity cost and energy consumption weighted cost or reduce delay, and update the non-dominated solution file and pheromone matrix accordingly, and output the improved scheduling scheme, updated non-dominated solution file and updated pheromone matrix again.

[0207] S6. When a new order arrives, equipment fails, or the time period of the time-based electricity price weighting function is switched, identify the affected tasks and frozen batches, perform local rearrangement only on the affected tasks, and perform incremental pheromone updates by weighting the historical pheromone on the device number and start time of the relevant tasks according to the memory retention coefficient, limit the scheme disturbance ratio to not exceed the disturbance ratio threshold, update the non-dominated solution file at the same time, and output the updated scheduling scheme and the final non-dominated solution file.

[0208] S7. Terminate based on stability threshold and disturbance ratio threshold, select the final scheduling scheme from the updated scheduling scheme, and generate equipment control instructions.

[0209] In this specific embodiment, S1 specifically refers to:

[0210] The user order data is validated and encoded. Based on the preset validation rules, the integrity and value range of fabric category, color, allergen, stain type, weight, number of pieces, washing procedure requirements, required completion time and user level are checked. The fabric category, color, allergen, stain type and washing procedure requirements are expressed by preset enumeration encoding, the weight is in kilograms, the number of pieces is in pieces, and the required completion time is represented by timestamp. Then, a task feature vector is generated for subsequent scheduling and constraint modeling.

[0211] Simultaneously, the device status and master data are collected to generate device parameters. The device parameters include capacity, availability, maintenance window, unit power, load rate, current temperature and allowable temperature range. The capacity comes from the device master data, the availability is indicated by the online status signal, the maintenance window is represented by a time period, the unit power is measured in kilowatts, the load rate is calculated by dividing the current loaded weight by the capacity, the current temperature is measured in degrees Celsius, and the allowable temperature range is represented by the upper and lower limits.

[0212] Subsequently, a time-of-use price weighting function is constructed based on the power supplier's time-of-use price schedule, normalizing the price of each time period to zero or one to reflect peak-valley differences, using the following formula:

[0213] Perform calculations;

[0214] in For a moment The time-period electricity price weighting function value is used as a weight in subsequent electricity cost and energy consumption weighted cost estimation. For a moment The electricity price in the time-of-use electricity price table, The lowest electricity price within the statistical period. The highest electricity price within the statistical period. For time indexing, when When Set all values ​​to zero to avoid the denominator being zero;

[0215] When generating the compatibility matrix, based on the task feature vector, the compatibility rules of fabric type, color, allergen, stain type and washing program requirements are used to mark whether the fabric can be mixed or not. If any dimension is incompatible, it is determined that the fabric cannot be mixed.

[0216] Disinfection switching parameters are generated based on the washing process management rules and equipment instructions. These parameters include the rinsing time, isolation time, and release conditions after the disinfection process is completed, which are used to constrain the timing of equipment reuse after disinfection.

[0217] Based on the equipment instructions and program configuration, program switching parameters are generated. These parameters include program switching time and additional energy consumption for switching, which are used to characterize the switching cost between different washing programs.

[0218] The cooling time is estimated based on the current temperature and allowable temperature range in the equipment parameters. When the current temperature is higher than the upper limit of the allowable temperature, the cooling time is calculated according to the temperature difference ratio. When the current temperature is not higher than the upper limit of the allowable temperature, the cooling time is recorded as zero. The above task feature vector, equipment parameters, time period electricity price weight function, compatibility matrix, disinfection switching parameters, program switching parameters and cooling time serve as the input data basis for subsequent steps.

[0219] In this specific embodiment, S2 specifically refers to:

[0220] Using task feature vectors, equipment parameters, compatibility matrix, disinfection switching parameters and cooling time as inputs, first establish non-mixable washing constraints. Based on the compatibility matrix, tasks marked as non-mixable in any dimension are prohibited from being merged into the same batch. In the batching strategy, tasks marked as mixable in the compatibility matrix and with the same washing program requirements are prioritized for merging.

[0221] Secondly, capacity constraints are established, using the capacity specified in the equipment parameters as the upper limit to control the total batch weight, employing a formula... Make a judgment, among which This is a collection of tasks awaiting batch processing. For task indexing, For batch indexing, For the task The weight is in kilograms. For the task Should it be included in the batch? Indicator variables, The reference capacity is in kilograms. The minimum capacity of the equipment that is in a usable state and supports the washing program bound to this batch is taken to ensure subsequent processing capability.

[0222] Again, establish post-disinfection rinsing and isolation window constraints at the batch level, requiring tasks with disinfection procedures to meet the minimum rinsing time and minimum isolation time in the batch results and follow release conditions to avoid cross-contamination;

[0223] At the same time, constraints on equipment availability and maintenance window are established, requiring that the washing program corresponding to the batch be arranged only when the equipment is available, and that its planned processing time does not overlap with the maintenance window, and ensuring that the batch's deadline is not earlier than the maintenance window's end time.

[0224] Establish cooling window constraints for batches that include high-temperature washing programs, so that the batch follows the cooling time estimated from the current temperature and the allowable temperature range in subsequent scheduling, and subsequent batches that do not meet the upper temperature limit shall not be scheduled within the cooling window.

[0225] Under the above constraints, the task feature vector is solved by batching and splitting through constraint programming. When capacity constraints or deadline conflicts make it infeasible, batch splitting is performed and the relevant constraints are recalculated. Finally, the feasible solution is normalized into batches. Each batch records the task index, bound washing program, total weight, number of pieces and deadline. The deadline is taken as the earliest value among the required completion times of the tasks contained in the batch, and it is used as the input basis for subsequent steps.

[0226] In this specific embodiment, S3 specifically refers to:

[0227] Based on the batch's task index, bound washing program, total weight, number of pieces, and deadline, and combined with the availability status and maintenance window in the equipment parameters, the earliest processing time for each batch on each equipment is determined to estimate the expected completion time. Based on the expected completion time and deadline, the predicted delay cost is calculated. Using the unit power and time-period electricity price weighting function in the equipment parameters, the weighted cost of electricity and energy consumption is estimated within the batch's expected processing time window. The thermal penalty is calculated based on the current temperature, allowable temperature range, and cooling time. The switching penalty is calculated by combining the program switching time and additional switching energy consumption in the program switching parameters. Simultaneously, the load penalty is calculated based on the load rate in the equipment parameters. Based on this, an initial heuristic factor is generated using the inverse of the aggregate cost, using the formula:

[0228] ;

[0229] Perform calculations, where For batch With equipment The initial values ​​of the heuristic factors on the combination are the heuristic preference quantities used to construct the probability transition rules. This is a batch index, taken from a batch set. For device indexes, taken from the device collection. For batch In the equipment The projected delay cost is expressed as the non-negative time difference between the projected completion time and the deadline. For batch In the equipment The electricity cost and energy consumption weighted average cost are obtained by multiplying the unit power and processing time and weighting by a time-period electricity price weighting function. For batch In the equipment The thermal penalty is a measure of the impact of current temperature, allowable temperature range, and cooling time on high temperatures. For batch In the equipment The switching penalty is based on a measurement of the program's switching time and the additional energy consumption during switching. For equipment The load penalty measures the impact of the current load on subsequent batches based on the load rate in the device parameters. For heuristic aggregation weight coefficients, take non-negative values ​​and satisfy the condition that the sum is one to achieve weighted aggregation of each penalty term and cost term under a unified scale;

[0230] Subsequently, an initial pheromone matrix is ​​generated based on the batch's processability and equipment availability. A unified initial value is set and fine-tuned according to the relative magnitude of the heuristic factor initial values. At the same time, a weight vector sequence for multi-objective decomposition is constructed to cover four types of objectives: delay objectives, electricity and energy consumption weighted costs, equipment load balancing, and customer satisfaction. This ensures that the sum of the four weights of each weight vector is one. A memory retention coefficient is set to weight historical pheromones and perform incremental updates in subsequent dynamic rescheduling. Furthermore, a disturbance ratio threshold and a stability threshold are set to limit the change ratio of the scheduling scheme and to judge the stability of the non-dominated solution file.

[0231] The final output includes the initial pheromone matrix, initial values ​​of the heuristic factors, weight vector sequence, memory retention coefficient, perturbation ratio threshold, and stability threshold.

[0232] In this specific embodiment, S4 specifically refers to:

[0233] Based on the initial pheromone matrix, heuristic factor initial values, and weight vector sequence, and combined with batch and equipment parameters, time period electricity price weight function, current temperature, allowable temperature range, cooling time, and program switching parameters, a multi-objective optimization model is established, which includes minimizing delay, minimizing the weighted cost of electricity and energy consumption, balancing equipment load, and maximizing customer satisfaction. This model is used to evaluate the comprehensive performance of batch start time and sequence decision-making on each device.

[0234] The multi-objective optimization model is decomposed into multiple sub-problems based on the weight vector sequence, and ant colony optimization is run in each sub-problem. First, for each batch, a candidate set is generated that only contains constraints that satisfy the equipment availability and maintenance window, the earliest processing time, the cooling time, and the program switching parameter constraints, and that follow the requirements for rinsing and isolation after disinfection. The candidate set is represented by a feasible triplet of batch, equipment, and start time.

[0235] During the deconstruction phase, a probabilistic transition rule is used to make joint decisions on "equipment allocation and start time". When there is a time conflict or the window is not met, a feasibility repair is performed, including postponing the start time or replacing available equipment to keep the constraints satisfied.

[0236] To maintain terminology consistency with step S3, this embodiment uses heuristic factors from the batch-device level. Break it down into batch-device-start time levels Candidate links are selected using the following probability transition formula:

[0237] ;

[0238] in To perform batch construction in the current construction step Select equipment With start time The selection probability is a normalized probability value. This is a batch index, taken from a batch set. For device indexes, taken from the device collection. The start time is taken from the set of time points that satisfy the earliest processable time constraint and the constraints of each process window. For batch In the equipment With start time The pheromone intensity linked to the device and the start time. For batch In the equipment With start time The heuristic factor is used to measure the aggregated advantages and disadvantages of forecast delay costs, electricity and energy consumption weighted costs, thermal penalties, switching penalties, and load penalties, and is dynamically updated with iteration. and These are non-negative control parameters, used to adjust the relative effects of pheromones and heuristic factors, respectively. For batch The candidate set includes all devices that meet the constraints of equipment availability and maintenance window, earliest processing time, cooling time, program switching parameters, and post-disinfection rinsing and isolation requirements. combination, and The device and start time index are used for summation;

[0239] After completing the scheduling construction of a single subproblem, four types of indicators are calculated: delay, electricity and energy consumption weighted cost, equipment load balancing and customer satisfaction. Non-dominated sorting is performed and the non-dominated solution file is updated to retain representative solutions. At the same time, local pheromone updates are performed on selected links to enhance exploration, global pheromone updates are performed on representative solution links in the non-dominated solution file to strengthen good decisions, and pheromone is shared and strengthened in the neighborhood of the weight vector sequence to improve collaborative search across subproblems.

[0240] In each iteration, the calculation is recalculated based on the time-period electricity price weighting function, the current temperature and the allowable temperature range, the cooling time and program switching parameters, and the equipment load rate. To dynamically reflect changes in forecast delay costs, electricity and energy-weighted costs, thermal penalties, switching penalties, and load penalties;

[0241] The final output includes candidate scheduling schemes, non-dominated solution files, and an updated pheromone matrix. The candidate scheduling schemes consist of a mapping from batch to device number and start time.

[0242] In this specific embodiment, S5 specifically includes:

[0243] Based on the candidate scheduling scheme, the non-dominated solution file, and the updated pheromone matrix, three types of neighborhoods are constructed around the candidate scheduling scheme and applied sequentially. The three types of neighborhoods include swap operation, relocation operation, and insertion operation. The swap operation is to swap the start time of two batches on the same device or to swap the device allocation and start time combination of two batches on different devices. The relocation operation is to move a batch to a new position in the same device sequence, or to move the batch to another device and reset the start time under the premise of satisfying the constraints. The insertion operation is to insert a batch in the device idle time window and postpone or fine-tune the start time of the adjacent batches.

[0244] After each neighborhood construction, a feasibility check is performed to ensure that the equipment availability status and maintenance window constraints, the earliest processing time constraint, the cooling time constraint, the program switching parameter constraint, and the rinsing and isolation requirements after disinfection are simultaneously met. For cases where the constraints are not met, a feasibility repair is performed, including extending the start time or replacing available equipment to maintain the constraint satisfaction.

[0245] The aggregation cost is calculated for neighborhood solutions that pass the feasibility check, and the aggregation cost and non-dominated ranking are used as acceptance criteria. The aggregation cost is calculated using the following formula:

[0246] ;

[0247] in For neighborhood solutions The aggregation cost is a scalar used to evaluate the comprehensive performance of neighborhood solutions. To construct and pass the feasibility check of a scheduling scheme under the current neighborhood operations, the solution set is taken from the neighborhood of the candidate scheduling scheme. This is a batch set containing the indexes of all batches involved in the scheduling. For batch indexing, This is a collection of devices, containing indexes of all devices involved in the scheduling. For device indexing, In the scheduling scheme Next batch The projected delay cost is expressed as the non-negative time difference between the projected completion time and the deadline. In the scheduling scheme Next batch The electricity cost and energy consumption weighted average cost is obtained by multiplying the unit power and processing time and weighting it using a time-period electricity price weighting function. In the scheduling scheme Next batch Thermal penalty is a measure of the impact of current temperature, allowable temperature range, and cooling time on high temperatures. In the scheduling scheme Next batch The switching penalty is based on a measurement of the program switching time and the additional energy consumption during switching. In the scheduling scheme Lower device The load penalty is used to measure the impact of the current load on subsequent batches. The aggregation weight coefficient is a non-negative value that sums to one to weight each cost and penalty on a uniform scale.

[0248] When the aggregation cost of a certain neighborhood solution is lower than the aggregation cost of the current candidate scheduling scheme or is not dominated by any representative solution in the non-dominated solution file in the non-dominated sorting, the neighborhood solution is accepted and used as the improved scheduling scheme. At the same time, the non-dominated sorting is performed to update the non-dominated solution file to remove dominated solutions and retain representative solutions.

[0249] Pheromones are enhanced on the device and start time connection corresponding to the accepted improvement. Historical pheromones are weighted according to the memory retention coefficient and the improvement increment is added to the unimproved device and start time connection to form an updated pheromone matrix. In the neighborhood loop, if no improvement is achieved in a single neighborhood operation, the process moves to the next neighborhood. Fine-tuning ends when no improvement is achieved within a preset number of attempts.

[0250] The final output includes an improved scheduling scheme, an updated non-dominated solution file, and a further updated pheromone matrix.

[0251] In this specific embodiment, S6 specifically refers to:

[0252] Set event trigger conditions to monitor new order arrivals, equipment failures, and time period switching of the time period electricity price weighting function. When any of the conditions are met, dynamic rescheduling will be initiated.

[0253] In the improved scheduling scheme, affected tasks are identified, and batches that conflict with the resources of the faulty equipment in the batches allocated to the faulty equipment and their subsequent sequences are marked as affected tasks. Batches located within the time period switching window are also marked as affected tasks. New orders are processed into batches based on the compatibility matrix, capacity constraints and disinfection switching parameters and then incorporated into the affected tasks.

[0254] The batches to be frozen are defined as follows: batches that have been started, batches that are close to the deadline, and batches that contain disinfection procedures. Among them, batches that are close to the deadline are those whose required completion time and the current time do not exceed the preset time. Frozen batches are not included in this reordering.

[0255] Local rearrangement is performed only for the affected tasks, taking into account the constraints of equipment availability and maintenance window, earliest processing time, cooling time, program switching parameters, and post-disinfection rinsing and isolation requirements. Local solutions are constructed on the connection between the equipment number and start time of the affected tasks according to the probability transfer rules defined in step S4. Feasibility repairs are performed for cases where time conflicts or windows are not met, including postponing the start time or replacing available equipment to keep the constraints satisfied. The comprehensive performance of the local solution is evaluated by predicting the delay cost, the weighted cost of electricity and energy consumption, the thermal penalty and switching penalty, and the load penalty.

[0256] Incremental pheromone updates are performed on the adjusted device number and start time link, historical pheromone is weighted according to the memory retention coefficient and enhanced with the current iteration increment, existing pheromone levels are retained for unadjusted links, and the increment size is determined by the overall performance improvement of the affected task.

[0257] The disturbance ratio of the scheme is calculated using the following formula:

[0258] Perform measurement;

[0259] in The disturbance ratio is the proportion of batches whose equipment allocation or start time is changed in this rescheduling to the total number of batches participating in the scheduling. To determine the number of batches for which device allocation or start time will be changed in this reordering. The number of batches participating in the scheduling is set simultaneously. As the perturbation ratio threshold, when achieve Stop further rearrangement and roll back the last modification to ensure the disturbance is under control;

[0260] The solution obtained from the local rearrangement is sorted non-dominated and the non-dominated solution file is updated. Dominated solutions are removed and representative solutions are retained until no further overall performance improvement is achieved within the scope of the affected task or the perturbation ratio of the scheme reaches the perturbation ratio threshold. The updated scheduling scheme and the final non-dominated solution file are then generated.

[0261] In this specific embodiment, S7 specifically refers to:

[0262] The stability and disturbance ratio of the final non-dominated solution file are judged, and the final scheduling scheme is selected and equipment control commands are generated when both meet the threshold conditions.

[0263] First, the final non-dominated solution file of the previous round of rescheduling is denoted as... The final non-dominated solution file of the current round of rescheduling is recorded as follows: Let the representative solution index set jointly maintained by the two rounds of archives be denoted as The aggregation cost function is denoted as Furthermore, its calculation follows the aggregation weight coefficient set in step S5. A stability metric is constructed as the maximum change in the aggregation cost of a representative solution, and the formula is used:

[0264] ;

[0265] in As a stability metric, it measures the maximum change in aggregate cost of representative solutions in the final nondominated solution archives over two rounds. To provide a representative solution index, it is taken from the index set. , This is a representative set of solution indexes jointly maintained by the final non-dominated solution archives of both rounds. For representative solution index In the current round of final non-dominated solution archives The aggregation cost in For representative solution index The previous round of final non-dominated solution archives The aggregation cost in This is the final non-dominated solution file for the current round of rescheduling. For the final non-dominated solution file from the previous round of rescheduling, a stability threshold is set as denoted as . and with Determined to be stable;

[0266] Secondly, the disturbance ratio of the updated scheduling scheme is calculated, and the disturbance ratio is denoted as... It is defined as the ratio of the number of batches reassigned in the updated scheduling scheme to the total number of batches, where the number of reassigned batches is denoted as... The total quantity of all batches is recorded as Set the disturbance ratio threshold as and with The disturbance was deemed acceptable.

[0267] Simultaneously satisfy and Under the given conditions, the scheduling scheme with the lowest aggregation cost is selected from the updated scheduling schemes as the final scheduling scheme, and the final scheduling scheme is converted into equipment control instructions. The equipment control instructions include equipment number, washing program, start time and batch allocation, which are used as the new data output for this step for issuance and execution.

[0268] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

[0269] This invention directly impacts the technical problem by coordinating hard-constraint pre-batch merging with soft-constraint multi-objective decomposition scheduling. Steps S1 and S2 construct non-mixing and process window constraints using a compatibility matrix, disinfection switching parameters, capacity, and cooling windows, and normalize them into batches using constraint programming, avoiding cross-contamination and reducing the search space. Step S3 constructs heuristic factors and initial pheromones using a time-based electricity price weight function, unit power, thermal penalty, switching penalty, and load penalty, forming a weight vector sequence that decomposes delay, electricity cost and energy consumption, load balancing, and customer satisfaction into multiple sub-problems. Step S4 uses ant colony optimization for joint decision-making on equipment allocation and start time, maintains files in a non-dominated sorting manner, and shares and strengthens pheromones in the neighborhood of the weight vector sequence. Step S5 further reduces costs and delays through variable neighborhood search involving exchange, relocation, and insertion. This achieves reduced electricity cost and energy consumption, lower delay rates, more balanced load, and full compliance with disinfection and cooling windows.

[0270] This invention makes several improvements to the algorithm structure to address technical issues: First, the design of separating hard and soft constraints and batch processing ensures that the feasible region is strictly filtered before entering the heuristic stage. The candidate set retains only connections that meet the requirements of equipment availability, maintenance window, earliest processing time, cooling, and program switching, significantly reducing infeasible repairs and computational overhead. Second, the multi-objective weight vector sequence and neighborhood pheromone sharing mechanism improve the collaborative search efficiency of adjacent subproblems, balancing cost and satisfaction. Third, when triggered by new orders, equipment failures, or electricity price time-sharing, incremental pheromone updates with memory retention coefficients and local rearrangements of frozen batches are introduced. The adjustment range is constrained by perturbation ratio thresholds and stability thresholds, making rescheduling both fast and stable, maintaining the continuity of historically excellent decisions, and ultimately maintaining a comprehensive performance of low cost, low latency, and controlled perturbation in dynamic environments.

Claims

1. A multi-objective task scheduling method for an Internet laundry center, characterized in that, Comprise: S1, receiving user order and equipment state data, generating task feature vector, equipment parameter, time period electricity price weight function, compatibility matrix, disinfection switching parameter, program switching parameter and cooling time; S2, based on the above data, establish non-mixing constraint, capacity constraint, disinfection after flushing and isolation window constraint, equipment available state and maintenance window constraint and cooling window constraint after high temperature program, solve by constraint programming and normalize to batch, batch contains task index, binding washing program, total weight, piece number and deadline; S3, on the basis of batch, combined with equipment parameter, time period electricity price weight function, current temperature, allowed temperature range and program switching parameter, calculate and output initial pheromone matrix, heuristic factor initial value, weight vector sequence, memory retention coefficient, disturbance proportion threshold and stability threshold, wherein the weight vector sequence is used to decompose the delay target, electricity cost and energy consumption weighted cost, equipment load balancing and customer satisfaction; S4, establish a multi-objective model containing delay target, electricity cost and energy consumption weighted cost target, equipment load balancing target and customer satisfaction target, decompose into sub-problems according to weight vector sequence, adopt ant colony optimization to make joint decision of equipment allocation and starting time in each sub-problem, construct solution according to probability transfer rule and maintain non-dominated solution archive by non-dominated sorting, and share and strengthen pheromone in the neighborhood of weight vector sequence, output candidate scheduling scheme, non-dominated solution archive and updated pheromone matrix; S5, around the candidate scheduling scheme, non-dominated solution archive and updated pheromone matrix, perform variable neighborhood search of exchange, relocation and insertion to reduce electricity cost and energy consumption weighted cost or reduce delay, and update non-dominated solution archive and pheromone matrix accordingly, output improved scheduling scheme, updated non-dominated solution archive and again updated pheromone matrix; S6, when detecting new order arrival, equipment failure or time period electricity price weight function period switching, identify affected tasks and frozen batches, only rearrange affected tasks locally, and perform incremental pheromone update by weighting historical pheromone according to memory retention coefficient on the connection of device number and starting time of related tasks, limit scheme disturbance proportion to no more than disturbance proportion threshold, update non-dominated solution archive, output updated scheduling scheme and final non-dominated solution archive; S7, determine termination according to stability threshold and disturbance proportion threshold, select final scheduling scheme in updated scheduling scheme, and generate equipment control instruction.

2. The multi-target task scheduling method for an Internet laundry center according to claim 1, wherein, S1 specifically: Field verification and coding are performed on user order data, and task feature vectors are generated according to fabric category, color, allergen, stain type, weight, piece number, washing program demand, required completion time and user level, wherein fabric category, color, allergen, stain type and washing program demand in the task feature vector are represented by preset enumeration coding, weight is measured in kilograms, piece number is measured in pieces, and required completion time is represented by timestamp; Collecting device status and main data, generating device parameters, the device parameters including capacity, available state, maintenance window, unit power, load rate, current temperature and allowed temperature range, wherein the capacity is derived from the device main data, the available state is indicated by the online state signal, the maintenance window is expressed in time period, the unit power is measured in kilowatt, the load rate is calculated by dividing the current loading weight by the capacity, the current temperature is measured in Celsius, and the allowed temperature range is expressed by upper and lower limits; Generating a time-of-use electricity price weight function according to the time-of-use electricity price table of the power supply party, normalizing each time-of-use electricity price to a weight of zero to one according to the highest and lowest electricity price interval, for reflecting the electricity price level at different time periods; Generating a compatibility matrix based on the task feature vector, and labeling washable and non-washable according to the compatibility rules of fabric category, color, allergen source, stain type and washing procedure demand, wherein any dimension with incompatibility is labeled as non-washable; Generating disinfection switching parameters according to washing process management rules and device instructions, the disinfection switching parameters including flushing time, isolation time and release conditions after completion of the disinfection procedure, for constraining the reuse time of the device after disinfection; Generating procedure switching parameters according to device instructions and procedure configuration, the procedure switching parameters including procedure switching time and switching additional energy consumption, for representing the switching cost between different washing procedures; Estimating cooling time according to the current temperature and the allowed temperature range in the device parameters, calculating the cooling time in proportion to the temperature difference when the current temperature is higher than the upper limit of the allowed temperature, and recording the cooling time as zero when the current temperature is not higher than the upper limit of the allowed temperature, the cooling time being used to represent the length of the cooling window after the high-temperature procedure.

3. The multi-target task scheduling method for an Internet laundry center according to claim 1, wherein, S2 specifically comprises: Establishing non-washable constraints according to the fabric category, color, allergen source, stain type and washing procedure demand in the task feature vector, and prohibiting the tasks labeled as non-washable in the compatibility matrix from being integrated into the same batch according to the compatibility matrix; Establishing capacity constraints according to the capacity in the device parameters, so that the total weight of each batch does not exceed the capacity; Establishing disinfection flushing and isolation window constraints according to the disinfection switching parameters, so that the tasks containing the disinfection procedure meet the disinfection flushing and isolation requirements at the batch level; Establishing device available state and maintenance window constraints according to the available state and maintenance window in the device parameters, so that the batch end time is not earlier than the maintenance window end time and the washing procedure corresponding to the batch has processability under the device available state; Establishing cooling window constraints after the high-temperature procedure according to the cooling time, so that the batch containing the high-temperature washing procedure meets the cooling window requirement in subsequent scheduling; Solving the task feature vector by constraint programming method under the above constraints, preferentially merging the tasks labeled as washable and with the same washing procedure demand in the compatibility matrix, and performing batch splitting and recalculating related constraints when capacity constraints or end time conflicts lead to infeasibility; Standardizing the feasible solution into a batch, the batch including corresponding task index, bound washing procedure, total weight, number of pieces and end time.

4. The multi-target task scheduling method for an Internet laundry center according to claim 1, wherein, S3 specifically comprises: On the basis of the batch task index, the bound washing program, the total weight, the number of pieces, and the deadline, in combination with the available state and the maintenance window in the device parameters, the earliest processable time of each batch on each device is determined, which is used to estimate the predicted completion time; A predicted delay cost is calculated according to the predicted completion time and the deadline, wherein the predicted delay cost is a non-negative value of a time difference between the predicted completion time and the deadline; An electricity and energy consumption weighted cost is estimated in the predicted processing time window of the batch according to the unit power and the time period electricity price weight function in the device parameters, wherein the electricity and energy consumption weighted cost is obtained by multiplying the unit power and the processing time length and weighted by the time period electricity price weight function; A heat penalty is calculated according to the current temperature, the allowed temperature range, and the cooling time, and in combination with the program switching time and the switching additional energy consumption in the program switching parameters, which is used to represent the cooling influence after high temperature and the program switching cost; A load penalty is calculated according to the load rate in the device parameters, which is used to represent the influence of the current loading on the subsequent batches; A heuristic factor initial value is generated according to the weighted aggregation of the predicted delay cost, the electricity and energy consumption weighted cost, the heat penalty and the switching penalty, and the load penalty, wherein the heuristic factor initial value is normalized in the form of the reciprocal of the aggregated cost on the combination of each batch and each device; An initial pheromone matrix is generated according to the processability of the batch and the available state of the device, wherein the initial pheromone matrix is fine-tuned according to the relative size of the heuristic factor initial value on the combination of each batch and each device with a uniform initial value; A weight vector sequence is constructed according to four types of targets of the delay target, the electricity and energy consumption weighted cost, the device load balancing, and the customer satisfaction, wherein the weight vector sequence is composed of a plurality of weight vectors, and the sum of the four weights of each weight vector is one, which is used to decompose the multi-target; A memory retention coefficient is set, which is used to weight and perform incremental update on the historical pheromone in subsequent dynamic rescheduling; A disturbance proportion threshold and a stability threshold are set, which are used to limit the change proportion of the scheduling scheme and judge the stability of the non-dominated solution archive in subsequent dynamic rescheduling and termination judgment; The initial pheromone matrix, the heuristic factor initial value, the weight vector sequence, the memory retention coefficient, the disturbance proportion threshold, and the stability threshold are output.

5. The multi-target task scheduling method for an Internet laundry center according to claim 1, wherein, S4 specifically is: On the basis of the initial pheromone matrix, the heuristic factor initial value, and the weight vector sequence, in combination with the batch and the device parameters, the time period electricity price weight function, the current temperature, the allowed temperature range, the cooling time, and the program switching parameters, a multi-objective optimization model including delay minimization, electricity and energy consumption weighted cost minimization, device load balancing, and customer satisfaction maximization is established, which is used to evaluate the comprehensive performance of the starting time and the sequence decision of the batch on each device; The multi-objective optimization model is decomposed into multiple sub-problems according to the weight vector sequence, and the ant colony optimization is run in each sub-problem, and a candidate set is generated according to the combination of batches and devices, the candidate set only contains device and starting time combinations that meet the device availability state and maintenance window constraints, the earliest processable time constraints, the cooling time constraints and the program switching parameter constraints, and the candidate set also meets the flushing and isolation requirements after disinfection for batches containing disinfection programs; In the process of constructing the solution, the probability transition rule is constructed according to the initial value of the heuristic factor and the weight of the pheromone, the joint decision of the device allocation and the starting time is made batch by batch, and the feasibility repair is performed for the cases where the time conflict or the window does not meet the requirements, including postponing the starting time or replacing the available device to maintain the constraint satisfaction; After completing the scheduling construction of one sub-problem, four types of indexes including the delay, the electricity and energy consumption weighted cost, the device load balancing and the customer satisfaction are calculated, the non-dominated sorting is performed and the non-dominated solution archive is updated, which is used to retain the representative solutions of the current iteration; Local pheromone update is performed on the selected device and starting time connection to enhance exploration, global pheromone update is performed on the representative solutions in the non-dominated solution archive to strengthen good decisions, and pheromone is shared and strengthened in the neighborhood of the weight vector sequence to improve the search coordination between adjacent sub-problems; In each iteration, the heuristic factor is recalculated according to the time period electricity price weight function, the current temperature and the allowed temperature range, the cooling time and the program switching parameter, and the load rate in the device parameter, which is used to dynamically reflect the changes of the predicted delay cost, the electricity and energy consumption weighted cost, the heat penalty, the switching penalty and the load penalty; The candidate scheduling scheme, the non-dominated solution archive and the updated pheromone matrix are output, and the candidate scheduling scheme is composed of the mapping from the batch to the device number and the starting time.

6. The multi-target task scheduling method for an Internet laundry center according to claim 1, wherein, S5 specifically is: According to the candidate scheduling scheme, the non-dominated solution archive and the updated pheromone matrix, three types of neighborhoods are constructed around the candidate scheduling scheme and are applied in turn, the three types of neighborhoods include the exchange operation, the relocation operation and the insertion operation, wherein the exchange operation is to exchange the starting time of two batches on the same device or to exchange the device allocation and the starting time combination of two batches on different devices, the relocation operation is to move a batch to a new position in the same device sequence or to move the batch to another device and reset the starting time on the premise of meeting the constraints, and the insertion operation is to insert a batch into the idle time window of a device and to postpone or fine-tune the starting time of the adjacent batches; After each neighborhood construction, a feasibility check is performed to make the device availability state and the maintenance window constraint, the earliest processable time constraint, the cooling time constraint, the program switching parameter constraint and the flushing and isolation requirements after disinfection meet simultaneously, and the feasibility repair is performed for the cases that do not meet the constraints, including postponing the starting time or replacing the available device to maintain the constraint satisfaction; The aggregate cost is calculated for the neighborhood solutions that pass the feasibility check, and the aggregate cost is represented by the weighted sum of the predicted delay cost, the electricity and energy consumption weighted cost, the heat penalty and the switching penalty, and the load penalty, which is used to evaluate the comprehensive performance of the neighborhood solutions; When the aggregation cost of the neighborhood solution is lower than that of the current solution or is not dominated by any solution in the current non-dominated solution archive in non-dominated sorting, accept the neighborhood solution as an improvement and replace the candidate scheduling scheme, while performing non-dominated sorting to update the non-dominated solution archive, removing dominated solutions and retaining representative solutions; Perform pheromone reinforcement on the improved corresponding device and starting time connection, weight the historical pheromone according to the memory retention coefficient on the non-improved connection, and superimpose the improvement increment to form an updated pheromone matrix; in the above neighborhood loop, when a single neighborhood operation does not achieve improvement, proceed to the next neighborhood, and when improvement is not achieved within a preset number of attempts, end the fine-tuning; Output the improved scheduling scheme, the updated non-dominated solution archive and the updated pheromone matrix.

7. The multi-target task scheduling method for an Internet laundry center according to claim 1, wherein, S6 specifically comprises: Set event trigger conditions, monitor new order arrival, equipment failure and time period price weight function time period switching, when any condition is met, enter dynamic rescheduling; Identify affected tasks in the improved scheduling scheme, mark the batches assigned to the failed equipment and the batches in the subsequent sequence that conflict with the equipment resources as affected tasks, mark the batches located within the time period switching window as affected tasks, and form processable batches for new orders according to the compatibility matrix, capacity constraints and disinfection switching parameters and integrate them into the affected tasks; Determine the frozen batches, which include the started batches, the near deadline batches and the batches containing disinfection procedures, wherein the near deadline batches are the batches whose required completion time and current time difference does not exceed a preset time length, and the frozen batches do not participate in this rearrangement; Only perform local rearrangement on the affected tasks, follow the device availability state and maintenance window constraints, the earliest processable time constraint, the cooling time constraint, the program switching parameter constraint and the disinfection after flushing and isolation requirements, construct a local solution on the device number and starting time connection of the affected tasks according to the probability transfer rule defined in S4, perform feasibility repair for the cases of time conflict or window dissatisfaction, including postponing the starting time or replacing the available equipment to maintain constraint satisfaction, and evaluate the comprehensive performance of the local solution by the predicted delay cost, the electricity cost and the energy consumption weighted cost, the heat penalty and the switching penalty, and the load penalty; Perform incremental pheromone update on the adjusted device number and starting time connection, weight the historical pheromone according to the memory retention coefficient and reinforce it with the current iteration increment, and retain the existing pheromone level on the unadjusted connection, and determine the increment size according to the comprehensive performance improvement of the affected tasks; Calculate the scheme disturbance proportion, which is the proportion of the number of batches whose device allocation or starting time is changed in the number of batches participating in scheduling, when the scheme disturbance proportion reaches the disturbance proportion threshold, stop further rearrangement and roll back the last modification to ensure controlled disturbance; Perform non-dominated sorting on the solution obtained by local rearrangement and update the non-dominated solution archive, remove dominated solutions and retain representative solutions, until no comprehensive performance improvement is achieved within the affected task range or the scheme disturbance proportion reaches the disturbance proportion threshold, generate an updated scheduling scheme and a final non-dominated solution archive.

8. The multi-target task scheduling method for an Internet laundry center according to claim 1, wherein, S7 specifically comprises: Performing stability judgment on the final non-dominated solution archive, taking the maximum change of the aggregate cost of the representative solutions in the final non-dominated solution archive of the last round of rescheduling and the current round of rescheduling as the stability measure, and determining as stable when the maximum change is not greater than the stability threshold value; Calculating the scheme disturbance proportion of the updated scheduling scheme, dividing the number of batches that are reassigned in the updated scheduling scheme by the total number of batches to obtain the scheme disturbance proportion, and determining as acceptable when the scheme disturbance proportion is not greater than the disturbance proportion threshold value; Under the condition of simultaneously satisfying the stability threshold value and the disturbance proportion threshold value, selecting the scheduling scheme with the minimum aggregate cost from the updated scheduling scheme as the final scheduling scheme, and converting the final scheduling scheme into device control instructions, the device control instructions including device number, washing program, start time and batch allocation, and serving as new data output of the present step for issuing and executing.

Images