Multi-target collaborative waste household appliance disassembly resource optimal configuration method and system

Abstract

The invention relates to a multi-objective collaborative waste household appliance disassembly resource optimal configuration method and system. The configuration method comprises the following steps of: obtaining a multi-objective collaborative waste household appliance disassembly resource; s1, predicting the number of recycled waste household appliances; s2, predicting a disassembly product; s3, making a disassembling plan; s4, establishing a resource optimization configuration model; s5, solving by a hybrid genetic simulated annealing algorithm; and S6, monitoring and adjusting the whole process.

Description

Technical Field

[0001] The present invention relates to a multi-objective collaborative waste household appliance dismantling resource optimization configuration method and system, and the present invention relates to the technical field of resource optimization configuration. Background Art

[0002] With the popularization and accelerated replacement of household appliances, the number of waste household appliances has increased year by year. These waste household appliances not only take up a lot of space, but also contain a lot of harmful substances. If they are not properly disposed of, they will cause serious pollution to the environment. At the same time, waste household appliances also contain many recyclable resources, such as metals and plastics. Therefore, it is of great significance to reasonably dismantle and recycle waste household appliances.

[0003] At present, the treatment of waste home appliances mainly includes the following links: recycling, disassembly, and reuse. Among them, the disassembly link is the most critical step. By disassembling waste home appliances, useful resources can be separated for subsequent reuse. However, in actual operation, the disassembly of waste home appliances faces many challenges:

[0004] 1. Unreasonable resource allocation: Since different types of household appliances have different disassembly products and different requirements for resource allocation, traditional disassembly methods make it difficult to reasonably allocate resources, resulting in low disassembly efficiency.

[0005] 2. The disassembly process is complex: There are many types of home appliances with complex structures. The disassembly process requires professional technology and equipment, and there are differences in the disassembly methods and steps for various types of home appliances.

[0006] 3. Environmental pollution risk: Used household appliances contain a large amount of harmful substances. If they are not disassembled properly, these harmful substances may leak and cause environmental pollution.

[0007] 4. Low economic benefits: Due to the low resource utilization rate during the dismantling process, many useful materials cannot be fully recycled. In addition, the dismantling cost is high, and the overall economic benefits are low.

[0008] The present invention analyzes the whole process of recycling, disassembling and reusing household appliances and finds that the recycling quantity and recycling rate of household appliances play a key role in the management of the whole process. These data directly affect the planning and operation of the recycling center, including layout design, capital planning, storage construction, personnel deployment, transportation arrangement, household appliance sorting and disassembly, etc. Therefore, it is necessary to predict the recycling volume of household appliances in the recycling center in each period in advance through big data analysis and modeling in order to optimize process management. Since the types and quantities of disassembled products have certain uncertainties in the disassembly process, it is very important to establish a prediction model for the number of disassembled products. This not only helps to verify the quantity of the disassembly process, but also provides a decision-making basis for sales planning, disassembly scheduling, remanufacturing and logistics scheduling. Summary of the invention

[0009] In order to solve the above problems, a multi-objective collaborative method for optimizing the allocation of waste household appliance dismantling resources is proposed. This method combines big data analysis, deep learning models and optimization algorithms to scientifically predict and optimize the recycling quantity and dismantling plan of waste household appliances, aiming to improve resource utilization efficiency, reduce dismantling costs, and reduce environmental pollution. By predicting the recycling quantity of household appliances and formulating dismantling plans in the recycling stage, the allocation of dismantling resources is optimized, and the economic benefits, logistics management efficiency and environmental protection effects of the recycling and dismantling process of waste household appliances are improved. Advanced deep learning models (such as long short-term memory networks (LSTM)) are used to predict the recycling and dismantling products of household appliances, thereby achieving full-process optimization of waste household appliance dismantling resources and promoting the recycling of resources and the sustainable development of the industry.

[0010] To solve the above problems, the technical solution adopted by the present invention is:

[0011] A multi-objective collaborative method for optimizing the allocation of waste household appliance dismantling resources, the allocation method comprising the following steps;

[0012] S1, forecast of the number of used household appliances recycled;

[0013] S2, prediction of disassembly products;

[0014] S3, dismantling plan formulation;

[0015] S4, establishment of resource optimization allocation model;

[0016] S5, solution using hybrid genetic simulated annealing algorithm.

[0017] A multi-objective collaborative waste household appliance dismantling resource optimization configuration system is provided, which has an LSTM model and is used to execute the above method.

[0018] Long Short-Term Memory Network, or LSTM for short, is a variant of a recurrent neural network. The LSTM model contains three main gating units: input gate, forget gate, and output gate, which help the model build long-term dependencies by regulating the flow of information. The input gate regulates the information entering the cell state and determines how much input information is stored; the forget gate filters the information in the cell state and determines which information to keep or discard; the output gate controls the output of information from the cell state to the hidden state and determines which information will be output from the cell state to the next time step. Through this gating mechanism, the LSTM model can efficiently learn and memorize long-term dependencies, enabling it to perform well in processing complex time series data. For a description of the structure of the LSTM model, please refer to Figure 2 , the figure shows the specific composition and function of the input gate, forget gate and output gate in each LSTM unit.

[0019] Forget gate: The function of the forget gate is to decide which information needs to be discarded from the cell state and which information needs to be retained. This step is achieved by selectively forgetting the information in the cell state of the previous step. The forget gate takes the hidden state of the previous step and the input data of the current time step as input, and assigns a value between 0 and 1 to each information. This value is called the output of the forget gate (denoted as), where 1 means that the information is completely retained and 0 means that the information is completely discarded. The specific calculation formula is as follows:

[0020] f t =σ(x t W xf +h t-1 W hf +b f ) (1–1);

[0021] Where W xf , W hf represents the weight matrix, b f is the deviation parameter.

[0022] Input gate: To update the cell state, an input gate is required. The input gate mainly consists of two parts:

[0023] Sigmoid layer: determines which values ​​need to be updated and represents their probability as I t , calculation formula:

[0024] I t =σ(x t W xi +h t-1 W hi +b i ) (1–2);

[0025] Where W xi , Whi represents the weight matrix, b i is the deviation parameter.

[0026] Tanh layer: responsible for candidate memory unit C t ' is created and added to the unit state. Calculation formula:

[0027] C′ t =tanh(x t W xc +h t-1 W hc +b c ) (1–3);

[0028] Where W xc , W hc represents the weight matrix, b c is the deviation parameter.

[0029] Output gate: determines the cell state C through the sigmoid layer t The output part of the unit state C is t Processed and multiplied with the output of the sigmoid layer to obtain the final output hidden state h t . Output gate calculation formula:

[0030] O t =σ(x t W xo +h t-1 W ho +b o )

[0031] h t =O t *tanh(C t )(1–4);

[0032] Where W xo , W ho represents the weight matrix, b o is the deviation parameter, b o is the output of the sigmoid layer.

[0033] Specifically, this method mainly includes the following innovations:

[0034] 1. Recycling quantity prediction: Use the long short-term memory network (LSTM) to predict the recycling quantity of used household appliances, thus providing a basis for the formulation of the dismantling plan.

[0035] 2. Optimal allocation of disassembly resources: Through a multi-objective optimization model, resources in the disassembly process are rationally allocated to maximize economic benefits while minimizing storage costs and energy consumption.

[0036] 3. Hybrid optimization algorithm: Use hybrid genetic simulated annealing algorithm to solve the optimization model to improve the speed and accuracy of model solution.

[0037] 4. Full-process management and control model: Build a deep learning model for the full-process management and control of home appliance products, comprehensively monitor and optimize the recycling, disassembly, and reuse links, and ensure the coordinated and efficient operation of each link.

[0038] Through the above-mentioned technical means, it is possible to achieve efficient and optimal allocation of waste home appliance dismantling resources, improve resource utilization, reduce processing costs, reduce environmental pollution, and promote the sustainable development of the waste home appliance recycling industry.

[0039] In the recycling of home appliances, accurately predicting the number of recycling orders for different types of products is the key to optimizing the recycling process and improving resource utilization efficiency. The recycling volume of home appliances is affected by many factors, such as policies and regulations, recycling methods, geographical location, seasonal changes, and people's environmental awareness. These factors jointly affect the recycling volume of each recycling center in different time periods. The recycling volume is directly related to the operating efficiency of the storage, sorting, and disassembly of home appliances.

[0040] Traditional prediction methods are mostly based on linear models, which are difficult to process complex time series data. Therefore, in order to quickly and accurately predict the recycling quantity of each recycling center at different times and better manage the recycling process, the present invention adopts a long short-term memory network (LSTM) model. With its powerful processing ability for time series data, the LSTM model can accurately capture the dynamic changes in the home appliance recycling market, thereby realizing the prediction of the number of recycling orders for multiple categories of products such as refrigerators, air conditioners, washing machines, televisions, freezers, computers, and water heaters. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a technical flow chart. Figure 2 This is the LSTM network structure diagram. Figure 3 It is a recycling quantity forecast chart for all industries and trades. Figure 4 This is a forecast chart of refrigerator recycling quantity. Figure 5 This is a forecast chart of the number of air conditioner recycling. Figure 6 This is a forecast chart of the number of washing machines recycled. Figure 7 This is a forecast chart for the number of color TVs to be recycled. Figure 8 This is a freezer recycling quantity prediction chart. Fig. 9 This is a forecast chart of computer recycling quantity. Fig.10 It is a prediction chart of the weight of broken iron of dismantling products.

[0042] Fig.11 It is a prediction chart of the foam weight of the disassembled product. Fig.12 It is a prediction chart of the weight of crushed plastics as disassembly products.

[0043] Fig.13 It is the loss function curve. Fig.14 It is a non-dominated hierarchy diagram. Fig.15 It is a non-dominated solution to the congestion graph.

[0044] Fig.16 It is a diagram of the steps for executing the elite strategy of the present invention. Fig.17 It is the optimal fitness curve obtained by each algorithm. Fig.18 It is the Pareto solution diagram of the disassembly plan. Fig.19 It is a disassembly data table diagram of a refrigerator disassembly line of the present invention. Fig. 20 It is a chromosome design chart of the structural product I of the present invention. Fig.21 It is a cross-data table diagram of the present invention. Fig. 22 It is a disassembly schedule diagram of the present invention. Fig.23 It is a design flow chart of the hybrid genetic simulated annealing algorithm of the present invention. DETAILED DESCRIPTION

[0045] like Figure 1-23 ,This embodiment is as follows, the multi-objective collaborative method for optimizing and configuring resources for dismantling of old household appliances includes the following steps;

[0046] S1, forecast of the number of used household appliances recycled;

[0047] The long short-term memory network (LSTM) is used to predict the recycling quantity of different types of used household appliances. Combined with the recycling market demand, the time series data is analyzed to generate a recycling quantity prediction model.

[0048] In the recycling stage, by collecting historical waste home appliance recycling data and combining it with the long short-term memory network (LSTM) model, the recycling quantity of different home appliances (such as refrigerators, air conditioners, washing machines, etc.) is predicted. In the prediction process, various factors affecting home appliance recycling (such as national policies, market demand, seasonal changes, etc.) are taken into account, and a recycling quantity prediction model is generated by modeling time series data. The use of the LSTM model enables the dynamic changes of the home appliance recycling market to be accurately captured.

[0049] S2, dismantling plan formulation;

[0050] Based on historical disassembly data, the types and quantities of disassembly products of different types of home appliances are predicted, including iron, plastic, foam and other products, and a disassembly product prediction model is constructed.

[0051] Based on the recycling quantity prediction results in step S1, a dismantling plan for used home appliances is formulated. The plan arranges the dismantling sequence and schedule reasonably according to the recycling quantity of different home appliance products to ensure the effective use of resources. This step optimizes the dismantling process by matching the dismantling requirements of each home appliance with the dismantling capacity of the enterprise.

[0052] S3, prediction of disassembly products;

[0053] According to the predicted recycling quantity, a dismantling plan for used household appliances is formulated to determine the dismantling order and output volume of different household appliances, providing a basis for subsequent resource optimization.

[0054] During the dismantling phase, the output of different products (such as broken iron, foam, plastic, etc.) during the dismantling of waste home appliances is predicted through the analysis of historical dismantling data and dismantling products. The types and quantities of these products are modeled using deep learning models to ensure that companies have sufficient data support in the sales and reuse of dismantled products and improve resource utilization efficiency.

[0055] S4, establishment of multi-objective resource optimization allocation model;

[0056] Combining the data in S1 and S3, a multi-objective collaborative resource optimization allocation model is constructed. The model takes the maximization of economic benefits, minimization of storage costs and minimization of energy consumption of dismantling enterprises as the main goals, and considers the equipment, personnel, storage capacity and dismantling characteristics of various household appliances of enterprises to make reasonable resource allocation. The model provides enterprises with optimized dismantling scheduling solutions by setting different constraints (such as dismantling capacity, storage capacity, etc.).

[0057] S5, hybrid genetic simulated annealing algorithm solution;

[0058] In order to solve the above multi-objective optimization model, a hybrid genetic simulated annealing algorithm is used. This algorithm combines the global search capability of the genetic algorithm and the local optimization capability of the simulated annealing algorithm to ensure that the optimal solution can be found quickly under complex constraints. Through multiple iterative optimizations of the model, the algorithm can achieve rational resource allocation, reduce storage costs and energy consumption, and improve the overall economic benefits of dismantling enterprises.

[0059] S6, full process monitoring and real-time adjustment: Full process monitoring and adjustment, based on the real-time data of each step of the recycling, disassembly and reuse of used household appliances, dynamically optimize and adjust the resource allocation and recycling and disassembly progress of the whole process to ensure the coordinated operation of each link;

[0060] Monitoring methods include cameras and sensors.

[0061] In actual operation, a full-process monitoring system is built to monitor all aspects of waste home appliances from recycling, dismantling to reuse in real time. The system can dynamically adjust the dismantling plan and resource allocation according to the actual dismantling progress and market demand, ensuring the coordinated operation of all links and maximizing benefits.

[0062] In step S1, the following steps are included:

[0063] First, the home appliance recycling order dataset is preprocessed; before building the LSTM model, the collected home appliance recycling order dataset is preprocessed in detail, including data cleaning, standardization, and missing value filling. This preprocessing process ensures the stability of the data and the accuracy of the model. The dataset is segmented according to different home appliance product categories, recycling sources, and other dimensions, so that the model can make accurate predictions for different categories of home appliance products.

[0064] Then, in the model training phase, an LSTM network was built, and the training cycle and learning rate were set; the LSTM network had three layers, 2,000 training cycles, and the learning rate was set to 0.002. In order to enhance the generalization ability of the model, cross-validation optimization technology was used. These measures enable the model to accurately identify seasonal fluctuations and long-term trends in the home appliance recycling market, providing companies with reliable decision-making basis and data support.

[0065] The efficient disassembly of household appliances is a core step in the entire household appliance recycling and reuse process. The present invention helps to realize the recycling of resources and effectively reduce the pollution of waste to the environment. In the process of efficient disassembly, different materials and parts can be separated in an orderly manner for reprocessing or reuse. This method can not only extend the service life of household appliances and reduce the demand for the exploitation of natural resources, but also reduce energy consumption and waste emissions, which helps to promote the development of the circular economy and also contributes to the construction of a green and low-carbon society.

[0066] In actual disassembly operations, there are many types of products from disassembly of household appliances, and the weights of disassembly products corresponding to household appliances of different models, weights and quantities vary greatly.

[0067] In step S1, before building the model, the existing disassembly data needs to be preprocessed. The data preprocessing process includes cleaning the data, processing missing values, outliers and duplicate data to ensure the quality and integrity of the data. This process is crucial to improving the accuracy and stability of the model. After processing, three disassembly products of this type of refrigerator within 24 days were finally selected as research objects: broken iron, foam and broken plastic.

[0068] In step S3, due to the small amount of data used, in order to avoid the overfitting problem of the model, the model training was set for 100 training cycles and the learning rate was set to 0.001. At the same time, the Adam optimizer was selected to speed up the convergence of the model and improve its stability. Through this optimization process, the output weight of different disassembly products can be predicted more accurately.

[0069] In the following analysis, the output per unit input weight was predicted for the three main dismantling products of 221L to 500L Freon-free refrigerators (EMR) - broken iron, foam and broken plastic. The prediction results reflect the output rules of different dismantling materials in the processing process through the output of the model, providing a decision-making basis for enterprises to optimize the dismantling process and resource allocation. The specific results can be viewed in the module of Figure 19. Figure 10-12 The output predictions for various types of materials under different training conditions are shown.

[0070] The cross entropy loss function (CrossEntropy Loss) is used in this model. The following formula is the calculation formula of the cross entropy loss function.

[0071]

[0072] The cross entropy loss function has become an effective tool for optimizing model parameters and accelerating the convergence of the training process due to its simple and efficient gradient calculation method. Compared with other loss functions, such as mean square error loss (usually used for regression tasks) or Hinge loss (mostly used for support vector machines, etc.), the cross entropy loss function performs particularly well in classification tasks. It can better distinguish different categories during model training, significantly improve classification accuracy, and enhance the generalization ability of the model.

[0073] For information about the changes in loss during training, refer to the chart module. Fig.13 , which shows the change curve of the loss function during the entire training process, thus reflecting the training convergence effect and performance optimization process of the model. By observing this loss curve, we can intuitively understand the learning effect of the model at each training stage. The evaluation indicators of the model are Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), Coefficient of Determination (R square) and so on. 2 ).

[0074] The mean square error is the average of the squares of the differences between the predicted values ​​and the true values, and is calculated as:

[0075]

[0076] Where n is the number of samples, yi is the true value, is the model prediction value.

[0077] The root mean square error is the square root of the mean square error. It measures the average difference between the predicted value and the true value. Its formula is:

[0078]

[0079] The mean absolute error is the average of the absolute differences between the predicted value and the true value. It can be used to measure the average degree to which the predicted value deviates from the true value. The formula is as follows:

[0080]

[0081] R-squared is a commonly used indicator to measure the goodness of fit of a model. The calculation formula is as follows:

[0082]

[0083] in is the mean of the true values.

[0084] From the experimental results, the mean square error (MSE) of the model is 0.25, the root mean square error (RMSE) is 0.5, the mean absolute error (MAE) is 0.03, and the determination coefficient (R 2 ) is 0.96. These indicators show that the LSTM model can accurately predict the unit weight of each type of household appliance disassembly product. Although the model has achieved good prediction results, it is still necessary to continuously optimize the model structure and parameter settings in the future to further improve its prediction accuracy and generalization ability.

[0085] Next, the model's prediction capability is further enhanced by introducing features and data sources related to the dismantling of home appliances (such as production process parameters, environmental factors, etc.). In addition, deep learning is combined with traditional time series analysis methods, or external data sources are introduced for joint modeling to improve the accuracy of predictions. In order to ensure the long-term stability and accuracy of the model, a continuous monitoring and feedback mechanism is established to adjust the model structure and prediction strategy in a timely manner.

[0086] In summary, by collecting a large amount of historical data and using deep learning technology, we have developed a predictive model for the recycling volume of home appliances and the output rate of dismantled products. This has laid a solid foundation for the optimal allocation of resources and closed-loop control of the entire process in the future.

[0087] In step S4, dismantling companies play a vital role in the entire process of recycling, dismantling and reuse of waste household appliances. When making dismantling decisions, companies need to seek a balance between economic benefits, material management and environmental protection to improve resource recovery rates and reduce processing costs. The basic principle of resource optimization allocation of dismantling companies is to determine the amount of household appliances to be dismantled in different periods based on their dismantling capabilities, and maximize economic profits through reasonable production arrangements.

[0088] It is known that dismantling companies can dismantle a variety of household appliances, such as refrigerators, air conditioners, washing machines, computers and televisions. For each household appliance dismantled, the company can obtain corresponding government subsidies. Each product can only be operated on a specific dismantling line, and the recycling quantity of different household appliances in the recycling market varies significantly with the seasons. Each household appliance will produce several different dismantling products when dismantled. The types and quantities of these products are different, and the values ​​of the products are different, and the benefits they bring are also different. At the same time, the number of dismantled products needs to meet the needs of subsequent sales and remanufacturing, but it is limited by the company's storage capacity and cannot exceed the storage limit. In addition, the scheduling restrictions of personnel and equipment must also be considered during the dismantling process.

[0089] Aiming at the resource optimization allocation problem of dismantling enterprises in the dismantling process, in order to formulate a reasonable and balanced household appliance dismantling scheduling plan, the present invention comprehensively considers the recycling quantity of the front-end recycling link, the sales plan of the back-end dismantling products and the impact of the reused materials, intermediate storage and transportation costs, aiming to achieve the goals of maximizing dismantling benefits, minimizing storage costs and minimizing energy consumption. To this end, a multi-objective collaborative household appliance dismantling resource optimization allocation model is established, and a hybrid genetic simulated annealing algorithm is used to solve it, ensuring that the optimal resource allocation solution is found under complex constraints.

[0090] S4.1, the multi-objective collaborative household appliance disassembly resource optimization configuration model has the following assumptions:

[0091] The demand for dismantling products in each planning period can be predicted by analyzing historical data;

[0092] The recycling volume of used household appliances in each planning period can be predicted by analyzing historical data;

[0093] The production time and quantity of various remanufactured products must meet the customer's order requirements;

[0094] The cost of out-of-stock is not considered;

[0095] Labor costs are included in production costs, that is, labor costs are not considered separately.

[0096] According to the needs of the model to be established, there are the following symbols:

[0097] E D,ni : The energy consumption of a single appliance of the i-th type of recycled household appliance disassembled by the n-th disassembly line;

[0098] C O,j : Sales price of disassembled products;

[0099] G O,tj : The number of sales orders for disassembled products in month t;

[0100] P i: Maximum disassembly capacity of disassembly line ( / day);

[0101] P R,i : The upper limit of the inventory of recycled household appliances;

[0102] P R,j : Dismantling product inventory limit;

[0103] t: Disassembly and production planning cycle of household appliances, t∈{1,2,...,12}

[0104] i: type of household appliance to be disassembled, i∈{1,2,...I}

[0105] j: type of disassembly product, j∈{1,2,...,J}

[0106] n: disassembly line, n∈{1,2,...,N}

[0107] Q: Dismantle the production profit of the enterprise;

[0108] Z: dismantle the production cost of the enterprise;

[0109] E: Dismantle the production energy consumption of enterprises;

[0110] Q i : Dismantle individual home appliances to obtain government subsidies;

[0111] C R,ti :Recycling cost of a single household appliance product;

[0112] G R,ti :The number of recycled household appliances;

[0113] C D,ti : Disassembly cost of a single household appliance product;

[0114] G D,tni :Number of household appliances disassembled;

[0115] C I,ti :Recover the inventory cost of a single piece of home appliance products;

[0116] G I,ti : Inventory quantity of recycled home appliances;

[0117] C P,ti : Inventory cost of a single piece of disassembled product;

[0118] G P,tj : The number of dismantled products of household appliances;

[0119] The production costs of dismantling enterprises include home appliance recycling costs (Z1), dismantling costs (Z2), recycled home appliance storage costs (Z3), and dismantling product storage costs (Z4). The income of dismantling enterprises is government subsidies (Q1) and sales income of dismantling products (Q2).

[0120]

[0121]

[0122] S4.2, a multi-objective optimization model is established with the goals of maximizing disassembly revenue (Q), minimizing storage costs (Z), and minimizing energy consumption (E). The purpose is to solve the number of household appliances disassembled in each time period G D,tni :

[0123]

[0124] In step S5, S5A, a multi-objective collaborative resource optimization and configuration algorithm for disassembly of household appliances is performed;

[0125] S5A.1, Chromosome encoding design

[0126] The number of household appliances disassembled in each time period is used as the decision variable. At the same time, the decision variables are all positive integers, which is conducive to the encoding of genetic algorithm chromosomes. Therefore, the encoding method of positive integers is adopted. Let CH represent a chromosome, where the chromosome CH uses an I×T matrix to represent the number of disassembled products in the period CH it , a gene on the sub-chromosome represents the disassembly amount of product i in the tth cycle CH it The following is a chromosome with a planning period of T and a product type of I.

[0127] S5A.2, Generation of initial population

[0128] The generation of the initial population is an important process in the genetic algorithm, and its main purpose is to randomly generate the initial solution in the feasible domain. In order to improve the quality of the initial solution, the initial solution must meet the actual situation, such as not exceeding the maximum dismantling capacity of the dismantling enterprise in each cycle, and not exceeding the sum of the expected recycling quantity and the current inventory quantity of each product.

[0129] CH it ≤P i

[0130] CH it ≤C R,it +G I,ti

[0131] Therefore, during the initial population generation process, the number of disassembly needs to be randomly generated between the above two values.

[0132] S5A.3, Selection

[0133] The selection operation imitates the law of "survival of the fittest" in nature. If the individual's fitness is high, it has a greater probability of being inherited to the next generation, otherwise the probability is small. There are many ways to perform the selection operation, such as roulette selection, sorting selection, optimal individual preservation, random league selection, etc.

[0134] S5A.4, Crossover

[0135] For individuals in the population, some chromosomes are selected from the parent generation according to the crossover probability. The following two strategies can be used for crossover. One of them is randomly selected during the crossover operation. The first crossover method is column crossover, which randomly selects a planning period t1∈[1,T], swaps the products of the two planning periods, and forms a new chromosome. The second crossover method is row crossover, which randomly selects a disassembled product and swaps the quantities of the two products in the corresponding planning periods. Taking column crossover as an example, the first two parent chromosomes CH1 and CH2 are crossed;

[0136] In the column crossover method, a planning period t1 is randomly selected; where t1∈[1,T], T is the total number of planning periods, and t1 and t2 are time points in the total number of planning periods; the product data of this planning period are swapped to form a new chromosome;

[0137] When the optimization period is 1 year, T is 12, and t1 and t2 are months from 1 to 12;

[0138] S5A.5, genetic variation

[0139] The mutation operation mainly generates offspring by randomly modifying the genes of the parent generation, which helps to maintain the diversity of the population and escape from the local optimal solution. m Select some chromosomes from the parent generation. For chromosome CH, randomly select a product output in any planning period and regenerate it. It should be noted that if the new chromosome CH' after mutation is not feasible, the chromosome needs to be repaired. The specific method is: check whether the mutated gene position and the subsequent planning period product meet the above conditions. If not, regenerate the unsatisfactory gene position, that is, randomly generate a CH it As a gene locus of subchromosome CH'.

[0140] S5A.6, Fast Nondominated Sorting

[0141] Fast non-dominated sorting is a concept proposed based on Pareto domination. Suppose there are k objective functions denoted as fi(x), where i is any integer among 1, 2, …, k, and j is also any integer among 1, 2, …, k, but i ≠ j. If for any objective function, fi(x1) < fi(x2), then individual x1 is said to dominate x2; if for any objective function, fi(x1) ≤ fi(x2) and there is at least one objective function such that fi(x1) < fi(x2) holds, then x1 is said to weakly dominate x2; if there exists an objective function such that fi(x1) < fi(x2) holds and there also exists an objective function satisfying fj(x1) < fj(x2), then individuals x1 and x2 are said to be non-dominant to each other.

[0142] Each individual in the population has two parameters ni and Si. ni is the number of individuals in the population that dominate individual i, and Si is the set of individuals dominated by individual i. The steps of fast non-dominated sorting are as follows:

[0143] First, by loop comparison, find all individuals in the population with ni = 0, assign them a non-dominated rank of 1, and store these individuals in the non-dominated set rank1.

[0144] Second, for each individual in the set rank1, subtract 1 from the ni of each individual in the set of individuals it dominates. If nj - 1 = 0, then add individual j to the set rank2 and assign the individuals in it a non-dominated rank of 2.

[0145] Third, repeat the above operations for the individuals in rank2 until all individuals are assigned a non-dominated rank.

[0146] The non-dominated rank is also called the Pareto rank. In the Pareto rank, individuals with a rank of 1 are not dominated by other individuals and are called non-dominated solutions, or Pareto optimal solutions. And the solution set formed by these Pareto optimal solutions forms a curve, which is the Pareto front.

[0147] For example, suppose there are two objective functions f1 and f2. After fast non-dominated sorting, all individuals are divided into three Pareto ranks. The specific situation is as shown in Fig.14 the figure. The individuals marked with circles in the figure are the individuals with a Pareto rank of 1, and they form the Pareto optimal solution set of the current problem. The distribution of these individuals forms a curve, which is the Pareto front of this problem, showing various solutions that reach the optimal solution under these two objective functions.

[0148] S5A.7, Crowding Degree

[0149] Crowding represents the density of individuals in space, which can be intuitively represented by a rectangle around individual xi that does not include other individuals. Fig.15 shown.

[0150] S5A.8, Elite Strategy

[0151] The elite strategy forms a new group by mixing the parent and offspring individuals, which expands the selection range when generating the next generation of individuals, achieving the purpose of retaining excellent individuals and eliminating inferior individuals. Fig.16 The example shown in the figure is analyzed. In the figure, P represents the parent population, the number of individuals in it is n, and Q represents the offspring population. The specific steps are as follows:

[0152] First, the parent population and the child population are merged to form a new population. Then the new population is sorted non-dominatedly. In this example, the population is divided into 6 Pareto levels.

[0153] Secondly, the new parent generation is generated. First, the non-dominated individuals with a Pareto rank of 1 are placed into the new parent set, and then the individuals with a Pareto rank of 2 are placed into the new parent population, and so on.

[0154] Again, if after all individuals with level k are placed in the new parent set, the number of individuals in the set is less than n, and after all individuals with level k+1 are placed in the new parent set, the number of individuals in the set is greater than n, then the crowding degree of all individuals at level k+1 is calculated and all individuals are sorted in descending order according to the crowding degree, and then all individuals with a level greater than k+1 are eliminated. In this example, it can be seen that k is 2, so the crowding degree of individuals with Pareto level 3 is calculated and sorted in descending order, and individuals with levels 4 to 6 are all eliminated.

[0155] Afterwards, the individuals in level k+1 are placed into the new parent set one by one in the order arranged in step 2 until the number of individuals in the parent set is equal to n, and the remaining individuals are eliminated.

[0156] S5B, multi-objective collaborative optimization and allocation of household appliance product disassembly resources;

[0157] This algorithm uses simulated annealing algorithm as the framework, combines the idea of ​​genetic algorithm, and combines local search capability into global search algorithm to improve the ability to obtain the optimal solution, making it have the advantages of fast solution speed, strong search capability, and good robustness. The specific implementation steps are as follows: (as shown in the chart module Fig.17 shown):

[0158] Step 5B.1: Initial temperature T max , termination temperature T min, cooling coefficient α, number of iterations N I , population size N, crossover probability P c , mutation probability P m etc. to set parameters:

[0159] Step5B.2: Generate an initial solution set according to the setting rules;

[0160] Step5B.3: If there are N solutions in the solution set obtained during the iteration process, go to Step5B.5, otherwise go to Step5B.4;

[0161] Step5B.4: Generate new solutions for the current solution set through neighborhood search and crossover mutation behavior. If the current solution set meets the set conditions, record the results and go to Step5B.3;

[0162] Step5B.5: Calculate the fitness of all individuals and the crowding degree of each individual, sort all individuals according to the crowding distance, select non-inferior individuals to form a new population and record them, and go to Step5B.6;

[0163] Step5B.6: Compare the new and old individuals. If the fitness function value of the old individual is not as good as that of the new individual, accept the new individual, that is, replace the old individual with the new individual. Otherwise, go to Step5B.7.

[0164] Step 7: Generate a random number R∈[0, 1]. If R is less than exp(-ΔE / kT), accept the new individual. Otherwise, accept the new individual according to the Metropolis criterion and go to Step 5B.8.

[0165] Step5B.8: If the algorithm reaches the set number of iterations in the inner loop, go to Step5B.9, otherwise go to Step3;

[0166] Step5B.9: If the current temperature of the algorithm in the outer loop reaches the set termination temperature, the process ends and the loop is exited. Otherwise, the temperature is reduced according to the set strategy, the inner loop is reset, and the process goes to Step5B.3.

[0167] According to the characteristics of the optimal allocation of household appliance disassembly resources, the multi-objective optimization model established above is experimented with simulated annealing algorithm and hybrid algorithm, and the results are analyzed. The data of four disassembled products are used as an example for the experiment. In this experiment, the parameters of each algorithm are set as follows: ① Hybrid algorithm, population size N = 100, crossover probability P c =0.8, mutation probability P m =0.3, initial temperature T max =500℃, termination temperature T min =50℃, temperature drop coefficient α=0 .95;②Simulated annealing algorithm, initial temperature T max =500℃, termination temperature T min =50℃, cooling coefficient α=0.95. The optimal fitness curves obtained by each algorithm are shown in the chart module. Fig.18 As shown, the hybrid algorithm is Fig.18 As shown in a, the simulated annealing algorithm is Fig.18 As shown in b.

[0168] from Fig.18 It can be seen from a that the fitness function value of the hybrid optimization algorithm decreases rapidly when the iteration number is 1-100 generations, and the fast non-dominated sorting algorithm begins to play a role when the iteration number is 100-800 generations. The local search ability of the algorithm is further enhanced, and the fitness function value slowly decreases to convergence, and finally converges when the iteration number is about 800 generations. Fig.18 From b, we can see that when the simulated annealing algorithm is used for calculation, the fitness function value decreases rapidly when the number of iterations is about 1-1500 generations, and slowly decreases when the number of iterations is about 1500-2000 generations, and finally converges when the number of iterations is about 2000 generations. From the comparison results of the two algorithms, we can see that under the same data conditions, the hybrid algorithm is significantly better than the simulated annealing algorithm in the number of iterations of target value calculation, and the convergence is further improved, which has the advantage of easily jumping out of the local optimum.

[0169] Calculate the 6-month disassembly plan and obtain the Pareto optimal solution as shown in the chart module Fig.19 shown.

[0170] Select a group of optimal solutions from the solution set. For example, the total dismantling income is 236 million yuan, and the dismantling energy consumption is 274,500 kWh. The corresponding dismantling plan is shown in the table. Fig. 22 shown.

[0171] The present invention realizes resource allocation optimization of the whole process of recycling and disassembling of used household appliances through scientific prediction and optimization technology. The method first predicts the recycling quantity of used household appliances and the recycling quantity of disassembly products based on the long short-term memory network (LSTM) model, and combines the hybrid genetic simulated annealing algorithm to optimize the allocation of resources in the recycling and disassembly process with the optimization goals of maximizing economic benefits, minimizing storage costs and energy consumption, thereby improving resource utilization efficiency, reducing disassembly and circulation costs, reducing carbon emissions in the recycling and disassembly process, improving the ecological environment, and promoting the sustainable development of the used household appliance recycling industry. The present invention has strong practicality and promotion value, and provides an intelligent, full-process resource optimization management tool for used household appliance disassembly enterprises.

[0172] The optimization method proposed in the present invention can effectively improve the resource utilization efficiency of waste home appliance dismantling enterprises, reduce processing costs, and reduce environmental pollution. Experiments show that after using this method, enterprises have achieved significant optimization effects in terms of dismantling product recovery rate, storage management, energy consumption, etc.

[0173] The present invention combines big data analysis and deep learning models to accurately predict the recycling quantity and dismantling products of waste household appliances, and rationally allocates dismantling resources through a multi-objective optimization model. The main technologies include long short-term memory network (LSTM) for recycling quantity prediction and hybrid genetic simulated annealing algorithm for optimization model solution. Through these technical means, the present invention has achieved remarkable results in improving the dismantling efficiency of waste household appliances, reducing processing costs and reducing environmental pollution.

[0174] Specifically, the prediction model of the number of recycled waste household appliances using the long short-term memory network (LSTM) can handle the complex relationships in time series data and improve the accuracy and reliability of predictions. The optimization model combines the idea of ​​multi-objective collaboration and improves the speed and accuracy of the solution through the hybrid genetic simulated annealing algorithm. It can reasonably allocate dismantling resources under different constraints to maximize economic benefits and resource utilization.

[0175] The experimental results show that the method of the present invention is superior to the traditional method in multiple dimensions. First, in terms of prediction accuracy, the LSTM model significantly improves the prediction accuracy of the recycling quantity and disassembly products. Secondly, in terms of optimization solution efficiency, the hybrid genetic simulated annealing algorithm shows a higher convergence speed and solution quality. Finally, in terms of practical application effects, by rationally allocating disassembly resources, the storage cost and energy consumption are reduced, and the overall economic and environmental benefits of disassembly enterprises are improved.

[0176] The present invention is fully described for a clearer disclosure, and the prior art is not listed one by one.

[0177] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some of the technical features therein with equivalents; it is obvious for those skilled in the art to combine multiple technical solutions of the present invention. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention. The technical contents not described in detail in the present invention are all known technologies.

Claims

1. A multi-objective collaborative method for optimizing resource allocation for dismantling of waste household appliances, characterized in that: The configuration method includes the following steps: S1, forecast of the number of used household appliances recycled; S2, prediction of disassembly products; S3, dismantling plan formulation; S4, establishment of resource optimization configuration model; S5, solution by hybrid genetic simulated annealing algorithm.

2. The multi-objective collaborative resource optimization allocation method for dismantling waste household appliances according to claim 1 is characterized by: S6: Full process monitoring and adjustment. Based on the real-time data of each step of the recycling, disassembly and reuse of used household appliances, the resource allocation and recycling and disassembly progress of the whole process are dynamically optimized to ensure the coordinated operation of each step. Monitoring methods include cameras and sensors.

3. The multi-objective collaborative resource optimization allocation method for dismantling waste household appliances according to claim 1 is characterized by: In S1, during the recycling phase, by collecting historical waste home appliance recycling data, the long short-term memory network LSTM is used to predict the recycling quantity of different types of waste home appliances. Combined with the recycling market demand, the time series data is analyzed to generate an LSTM model for recycling quantity prediction; Use the LSTM model to predict the recycling quantity of different household appliances; In the prediction process, by modeling the time series data; S1 includes the following specific steps: First, the home appliance recycling order data set is preprocessed; the preprocessing steps include data cleaning, processing missing values, outliers and duplicate data; after processing, the research objects are selected; The research objects include broken iron, foam and broken plastic Then, in the model training phase, the LSTM network was built, the training cycle and learning rate were set, and cross-validation was used.

4. The multi-objective collaborative resource optimization allocation method for dismantling waste household appliances according to claim 1 is characterized by: In S2, based on the recycling quantity prediction result in step S1, a dismantling plan for used household appliances is formulated; the dismantling plan arranges the dismantling sequence and schedule according to the recycling quantity of different household appliances.

5. The multi-objective collaborative resource optimization allocation method for dismantling waste household appliances according to claim 1 is characterized by: In S3, according to the predicted number of recycling, a disassembly plan for used household appliances is formulated to determine the disassembly order and output of different household appliances. First, the training cycle is set, the learning rate is set, and the Adam optimizer is selected. Then, we conduct analysis and make unit input-output forecasts for the research object; Next, based on the prediction results, the output of the LSTM model is used to reflect the output rules of different disassembled materials during the processing process; In the LSTM model, the cross entropy loss function is used. The following formula is the calculation formula of the cross entropy loss function; Among them, p(x i ) represents x i The true value of q(x i ) represents the model for the xth i Probability distribution of predictions; Secondly, mean square error, root mean square error, mean absolute error, and R square are used to evaluate the LSTM model; among them, The mean square error is the average of the squares of the differences between the predicted values ​​and the true values, and is calculated as: Where n is the number of samples, y i is the true value, is the model prediction value; The root mean square error is the square root of the mean square error, which measures the average difference between the predicted value and the true value. Its formula is: The mean absolute error is the average of the absolute differences between the predicted value and the true value, which measures the average degree to which the predicted value deviates from the true value. The formula is as follows: R-squared is a commonly used indicator to measure the goodness of fit of a model. The calculation formula is as follows: in, is the mean of the true values; Next, in the LSTM model, features and data sources related to the disassembled products of home appliances are introduced; the features and data sources include production process parameters and environmental factors.

6. The multi-objective collaborative resource optimization allocation method for dismantling waste household appliances according to claim 1 is characterized by: In S4, the data in S1 and S3 are combined to build a multi-objective collaborative resource optimization allocation model, which is solved by a hybrid genetic simulated annealing algorithm to ensure that the optimal resource allocation solution is found under complex constraints. The resource optimization allocation model takes the maximization of economic benefits, minimization of storage costs and minimization of energy consumption of the dismantling enterprise as its main goals, taking into account the equipment, personnel, storage capacity of the enterprise and the dismantling characteristics of various home appliances, as well as setting constraints, including dismantling capacity and storage capacity. S4.1, first, the multi-objective collaborative model for optimizing the allocation of household appliance dismantling resources includes the following assumptions: the demand for dismantling products in each planning period is obtained by analyzing and predicting historical data; The recycling volume of used household appliances in each planning period is obtained through analysis and prediction of historical data; The production time and quantity of various remanufactured products can meet the order requirements of customers; Assume that the out-of-stock cost is not considered; the labor cost is included in the production cost, that is, the labor cost is not considered separately; Then, according to the needs of the model to be established, there are the following symbols: E D,ni : The energy consumption of a single appliance of the i-th type of recycled household appliance disassembled by the n-th disassembly line; C O,j : Sales price of disassembled products; G O,tj : The number of sales orders for disassembled products in month t; P i : Maximum disassembly capacity of disassembly line ( / day); P R,i : The upper limit of the inventory of recycled household appliances; P R,j : Dismantling product inventory limit; t: Disassembly and production planning cycle of household appliances, t∈{1,2,...,12} i: type of household appliance to be disassembled, i∈{1,2,...I} j: type of disassembly product, j∈{1,2,...,J} n: disassembly line, n∈{1,2,...,N} Q: Dismantle the production profit of the enterprise; Z: dismantle the production cost of the enterprise; E: Dismantle the production energy consumption of enterprises; Q i : Dismantle individual home appliances to obtain government subsidies; C R,ti :Recycling cost of a single household appliance product; G R,ti :The number of recycled household appliances; C D,ti : Disassembly cost of a single household appliance product; G D,tni :Number of household appliances disassembled; C I,ti :Recover the inventory cost of a single piece of home appliance products; G I,ti : Inventory quantity of recycled home appliances; C P,ti : Inventory cost of a single piece of disassembled product; G P,tj : The number of dismantled products of household appliances; Secondly, the production costs of dismantling enterprises include home appliance recycling costs (Z1), dismantling costs (Z2), recycled home appliance storage costs (Z3), and dismantling product storage costs (Z4). The income obtained by dismantling enterprises is government subsidies (Q1) and sales of dismantled products (Q2). Q=Q1+Q2 Z=Z1+Z2+Z3+Z4 S4.2, a multi-objective optimization model is established with the goals of maximizing disassembly revenue (Q), minimizing storage costs (Z), and minimizing energy consumption (E). The purpose is to solve the number of household appliances disassembled in each time period G D,tni : min Z=Z1+Z2+Z3+Z4 7. The multi-objective collaborative resource optimization allocation method for dismantling waste household appliances according to claim 1 is characterized by: In S5, S5A, a multi-objective collaborative resource optimization and configuration algorithm for disassembly of household appliances is performed; S5B, realizes the optimal allocation of household appliance product disassembly resources with multi-objective collaboration; The simulated annealing algorithm is used as the framework and combined with genetic algorithm.

8. The multi-objective collaborative resource optimization allocation method for dismantling waste household appliances according to claim 1 is characterized by: In step S5A, experiments are conducted using the simulated annealing algorithm and the hybrid algorithm, and the results are analyzed, and the following steps are performed; S5A.1, chromosome coding design; The number of household appliances disassembled in each time period is used as the decision variable. At the same time, all decision variables are positive integers and are coded in a positive integer manner. When encoding positive integers, let CH represent a chromosome, and the chromosome CH uses an I×T matrix to represent the number of period disassembly products CH it , a gene on the sub-chromosome represents the disassembly amount of product i in the tth cycle CH it , where the planning period is T and the product type is a chromosome I; S5A.2, generation of the initial population; first, the initial solution must meet the actual situation, including that the initial solution cannot exceed the maximum dismantling capacity of the dismantling enterprise in each cycle, and cannot exceed the sum of the expected recycling quantity and the current inventory quantity of each product; CH it ≤P i CH it ≤C R,it +G I,ti Therefore, during the initial population generation process, the number of disassembly is randomly generated based on the formula; S5A.3, perform selection operation; The selection operation is based on the law of survival of the fittest in nature. If the fitness of an individual is high, it has a greater probability of being inherited by the next generation, otherwise the probability is small; the method of performing the selection operation includes roulette selection, sorting selection, optimal individual preservation or random league selection; S5A.4, perform crossover operation; For individuals in the population, some chromosomes are selected from the parent generation according to the crossover probability, and crossover is performed according to the following two strategies. One of the methods is randomly selected during the crossover operation; The first crossover method is column crossover, which randomly selects the planning period t1∈[1,T], In the column crossover method, a planning period t1 is randomly selected; where t1∈[1,T], T is the total number of planning periods, and t1 and t2 are time points in the total number of planning periods; the product data of this planning period are swapped to form a new chromosome; In the column crossover method, two different planning periods (t1 and t2) are randomly selected for data exchange to form a new chromosome; The second crossover method is row crossover, which randomly selects two disassembled products and swaps the quantities of the two disassembled products in the corresponding planning periods; When selecting column crossover, set the first two parent chromosomes CH1 and CH2 before crossover; S5A.5, Gene mutation manipulation The offspring are generated by randomly modifying the genes of the parents, and the individuals in the population are mutated according to the probability P. m Select several chromosomes from the parent generation. For chromosome CH, randomly select a product output in any planning period and regenerate it. If the mutated chromosome CH' violates the constraints of storage capacity or disassembly capacity, it is considered an infeasible chromosome and needs to be repaired; Constraints include the number of dismantled products exceeding the storage limit; When repairing chromosomes, check whether the mutated gene position and the subsequent planned product of the gene position meet the above constraints; if not, regenerate the unsatisfactory gene position, that is, randomly generate a CH it As a gene locus of subchromosome CH'; S5A.6, fast nondominated sort; Fast non-dominated sorting is proposed based on Pareto dominance; assuming there are k objective functions denoted as f i (x), where i is any integer in 1, 2, ..., k, and j is also any integer in 1, 2, ..., k, but i ≠ j; If individuals x1 and x2 have f for any objective function i (x1) <f i (x2), then individual x1 is said to dominate x2; if for any objective function, f i (x1)≤f i (x2) and there is at least one objective function such that f i (x1) <f i (x2) holds true, then x1 is said to weakly dominate x2; if there exists an objective function such that f i (x1) <f i (x2) holds and there exists an objective function that satisfies f j (x1) <f j( x2), then individuals x1 and x2 are said to not dominate each other; Each individual in the population has two parameters n i and S i , n i is the number of individuals that dominate individual i in the population, S i is the set of individuals dominated by individual i. The steps of fast non-dominated sorting are as follows: First, find all n in the population through circular comparison i = 0, assign a non-dominated rank of 1 and store these individuals in the non-dominated set rank1; Secondly, for each individual in the set rank1, the n of each individual in the set of individuals it dominates j Subtract 1 from each. If n j -1=0, then individual j is stored in the set rank2 and the individual non-dominated rank 2 is assigned; Again, repeat the above operation for individuals in rank2 until all individuals are assigned a non-dominated rank; The non-dominated level is called the Pareto level; In the Pareto hierarchy, individuals with a level of 1 are called non-dominated solutions, or Pareto optimal solutions, because they are not dominated by other individuals. The solution set composed of these Pareto optimal solutions forms a curve called the Pareto frontier. Assuming there are two objective functions f1 and f2, all individuals are divided into three Pareto levels after fast non-dominated sorting. The individuals marked with circles are individuals with a Pareto level of 1, which constitute the Pareto optimal solution set of the current problem. The distribution of these individuals forms a curve, which is the Pareto frontier of the problem, showing various solutions to achieve the optimal solution under these two objective functions. S5A.7, Determine the degree of crowding; Crowding represents the density of individuals in space, which can be intuitively expressed by individual x i The surrounding rectangle does not include other individuals; S5A.8, elite strategy; By mixing the parent and offspring individuals to form a new group, the screening range for the next generation of individuals is expanded, so as to achieve the purpose of retaining excellent individuals and eliminating inferior individuals; let the parent population be P, the number of individuals be n, and the offspring population be Q, the specific steps are as follows: First, the parent population and the offspring population are merged to form a new population, and then the new population is non-dominated sorted; Secondly, the new parent generation is generated. First, the non-dominated individuals with a Pareto rank of 1 are placed in the new parent generation set, and then the individuals with a Pareto rank of 2 are placed in the new parent population, and so on. Thirdly, if after all individuals of level k are placed in the new parent set, the number of individuals in the set is less than n, and after all individuals of level k+1 are placed in the new parent set, the number of individuals in the set is greater than n, then the crowding degree of all individuals of level k+1 is calculated and all individuals are sorted in descending order according to the crowding degree, and then all individuals with a level greater than k+1 are eliminated; Afterwards, the individuals in level k+1 are put into the new parent set one by one in sorted order until the number of individuals in the parent set is equal to n, and the remaining individuals are eliminated.

9. The multi-objective collaborative method for optimizing resource allocation for dismantling of waste household appliances according to claim 1; characterized in that: In step S5B, the following steps are performed: Step 5B.1: Initial temperature T max , termination temperature T min , cooling coefficient α, number of iterations N I , population size N, crossover probability P c , mutation probability P m To set the parameters: Step5B.2: Generate an initial solution set according to the setting rules; Step5B.3: If there are N solutions in the solution set obtained during the iteration process, go to Step5B.5, otherwise go to Step5B.4; Step5B.4: Generate new solutions for the current solution set through neighborhood search and crossover mutation behavior. If the current solution set meets the set conditions, record the results and go to Step5B.3; Step5B.5: Calculate the fitness of all individuals and the crowding degree of each individual, sort all individuals according to the crowding distance, select non-inferior individuals to form a new population and record them, and go to Step5B.6; Step5B.6: Compare the new and old individuals. If the fitness function value of the old individual is not as good as that of the new individual, accept the new individual, that is, replace the old individual with the new individual. Otherwise, go to Step5B.

7. Step 7: Generate a random number R∈[0, 1]. If R is less than exp(-ΔE / kT), accept the new individual. Otherwise, accept the new individual according to the Metropolis criterion and go to Step 5B.

8. ΔE is the difference between the new and old individual objective functions; Step5B.8: If the algorithm reaches the set number of iterations in the inner loop, go to Step5B.9, otherwise go to Step3; Step5B.9: If the current temperature of the algorithm in the outer loop reaches the set termination temperature, the process ends and the loop is exited. Otherwise, the temperature is reduced according to the set strategy, the inner loop is reset, and the process goes to Step5B.

3.

10. A multi-objective collaborative waste household appliance dismantling resource optimization configuration system, characterized by: Comprising an LSTM model; used to execute the system described in claim 1.

Images