61 results about "Crowding distance" patented technology

The invention discloses an industrial chain multi-objective optimization solving method and a computer program product, and belongs to the technical field of data processing method.The industrial chain multi-objective optimization solving method comprises the steps that a mathematical model is constructed, and multiple objective functions of an industrial chain are quantized; constructing a swarm intelligence bat optimization algorithm; parameters in the swarm intelligence bat optimization algorithm are initialized; executing a search process, updating the speed and position of each bat individual, and comprehensively evaluating the fitness value of each bat individual by using a plurality of objective functions; a Pareto dominating strategy is adopted to optimize the multiple objective functions, and a Pareto non-dominated solution set is screened out; using a congestion distance algorithm to calculate a congestion distance between solutions in a Pareto non-dominated solution set, and rejecting over-concentrated solutions to form an optimal solution; carrying out back propagation on the optimal solution and the fitness value, and optimizing a next iteration result; and outputting an optimal solution when an iteration termination condition is satisfied. According to the method, the solving precision, the convergence speed and the global search capability of the industrial chain optimization problem can be improved.

The invention provides an unmanned aerial vehicle multi-target path planning method based on a multi-target sheep flock migration algorithm. The method comprises the following steps: 1, constructing a three-dimensional model of a task environment: 1.1, constructing a grid-based three-dimensional task space model; 1.2, determining collision probability distribution; 2, initializing algorithm parameters and populations; 2.1, initializing multi-target sheep flock migration algorithm parameters; 2.2, randomly generating an initial population in the task area range; 3, performing non-dominated sorting on the population, and selecting head sheep; 3.1, calculating an objective function value of each sheep; 3.2, determining a Pareto optimal solution set by adopting non-dominated sorting; 3.3, selecting the sheep with the largest crowding distance as the head sheep; 4, executing a grazing operator; 5, judging and executing a migration operator or a contraction operator; 5.1, if the optimal solution of the group is updated, executing the migration operator; 5.2, if the optimal solution of the group is not updated, executing a contraction operator; and 5, judging whether the maximum number of iterations is reached or not, if so, outputting an optimal path planning solution, and otherwise, returning to the step 3.

The invention relates to the technical field of multi-objective optimization, in particular to a multi-mechanism deep integrated multi-objective optimization framework and an optimization method and application thereof, comprising a reinforcement learning module, a genetic algorithm module, a bidirectional evolutionary learning coupling module, a Pareto experience pool module, a preference management module and a graph neural network module, the bidirectional evolutionary learning coupling module is used for realizing bidirectional information interaction between the reinforcement learning module and the genetic algorithm module, and the reinforcement learning module generates candidate solutions and feeds the candidate solutions back to the genetic algorithm module; the genetic algorithm module screens the elite solution with the crowding distance larger than a set threshold value in the first Pareto layer and updates the strategy of the feedback reinforcement learning module. According to the method, the strategy learning ability of reinforcement learning and the global population search ability of the genetic algorithm are fused, and the overall optimization efficiency and the approximation ability to the complex Pareto frontier are remarkably improved.

The invention discloses a sintering batching optimization scheme, and relates to the technical field of sintering ore batching control, and the technical key points are as follows: a random forest proxy model is used to predict the performance of finished ore, and a nonlinear mapping relationship between an independent variable and a dependent variable is used as a target function; carrying out Pareto optimal solution set search through non-dominated sorting and crowding distance calculation by adopting an NSGA-II (Non-dominated Sorting Genetic Algorithm-II) algorithm; calculating the weight of each target index by using an entropy weight method based on a Pareto solution set; weight is determined from a Pareto solution set generated by an NSGA-II algorithm, an optimal batching scheme closest to an ideal solution and far away from a negative ideal solution is screened out through a TOPSIS method, and multi-target comprehensive weighing is achieved; according to the method, sintering batching optimization based on finished ore performance feedback is realized, dynamic optimization of a batching plan can be effectively guided, the sinter quality control level is improved, the production cost is reduced, and technical support is provided for efficient and stable operation of a blast furnace ironmaking process.

The invention belongs to the technical field of semi-active suspension control methods, and relates to a semi-active suspension control method based on an improved multi-objective particle swarm optimization (MOPSO) algorithm, which comprises the following steps of: performing adaptive grid division on an external file of the MOPSO algorithm to determine global optimum with the most diversity and global optimum with the most convergence; judging the evolution state of the population according to the distribution entropy variation of all optimal solutions in an external file, and selecting a global optimal gBest for guiding evolution; according to the variable quantity of the local crowding distance of the current population evolution, adjusting flight parameters of population evolution to balance the evolution trend of the population; and fusing the two core mechanisms, providing a multi-information fusion multi-target particle swarm optimization algorithm, and applying the multi-information fusion multi-target particle swarm optimization algorithm to determination of a semi-active suspension LQR control strategy weight coefficient. According to the novel multi-objective optimization algorithm, external archive information and current population dynamic information are cooperatively integrated, so that balance between convergence precision and population diversity is effectively realized.

The invention discloses an ecological flow process multi-objective optimization method based on an NSGA-II algorithm, and relates to the technical field of water conservancy projects. Comprising the steps of defining discrete decision variables, setting an objective function, constructing constraint conditions and a mixed penalty function, and configuring algorithm parameters; generating an initial solution meeting the ecological flow constraint, and repairing the solution which does not meet the constraint after genetic manipulation to ensure the feasibility of the solution; performing non-dominated sorting on the population to divide Pareto frontier layers, calculating a congestion distance of a solution in each frontier layer, realizing genetic manipulation through tournament selection, binary crossover simulation and polynomial variation, and updating the population in combination with environment selection; and extracting non-dominated solutions, screening optimization schemes from the non-dominated solutions, and outputting monthly scheduling schemes and water transfer and ecological flow shortage statistics of each reservoir under each scheme. According to the invention, automatic distribution of the ecological base flow, the basic ecological flow and the target ecological flow in the interannual period is realized.

The invention provides an axial flux motor multi-objective optimization method based on an improved homing religious algorithm. Firstly, key structure parameters are screened through sensitivity analysis, and the optimization range of the key structure parameters is determined; then, sample data is generated by adopting improved Latin hypercube sampling, and a data set is constructed through finite element simulation; then, performing automatic optimization on smooth factors of the generalized regression neural network based on a He-horse optimization algorithm, and establishing a high-precision agent model; improving a horizon algorithm by adopting Tent chaotic mapping, non-dominated sorting and congestion degree distance, and combining the horizon algorithm with the proxy model to carry out multi-objective optimization to generate a Pareto solution set; and finally, selecting an optimal motor parameter according to the target weight. According to the method, parameterized finite element analysis, an optimization agent model and an improved multi-target algorithm are deeply fused, so that the global optimization capability is ensured, the optimization efficiency and precision are remarkably improved, and an effective solution is provided for rapid and accurate design of the axial flux motor.

The invention relates to an underwater rotary body appearance multi-target transition delay optimization method based on NSGA-II. The method comprises the following steps: establishing a multi-target optimization model; population initialization: selecting a plurality of known underwater revolving body appearances as a primary population; at multiple navigational speeds, predicting a natural transition position of a boundary layer at the head part of each appearance in the current population; performing non-dominated sorting; crowding degree calculation: the crowding distance is calculated according to the target value of each individual, and the individuals with the larger crowding distance are ranked in the front; generating a filial generation population: executing genetic operations of selection, crossover and variation on all shapes in the current generation population so as to generate the filial generation population; using an elitist strategy to mix the filial generations and the parent generations, performing non-dominated sorting and crowding degree calculation, and taking the first N individuals as the next generation of individuals; and performing iterative evolution to obtain a shape Pareto solution set, namely a shape solution set with uniformly distributed transition position delay.

The invention discloses an intensive storage network vehicle path crowd target optimization method based on an NSGA-II algorithm, and belongs to the technical field of vehicle path optimization, and the method comprises the steps: firstly, building a crowd target optimization model which comprises a plurality of targets such as transportation cost, time cost and path length; thirdly, performing population initialization and selection through an NSGA-II algorithm, and adopting non-dominated sorting and crowding distance calculation to ensure that a Pareto optimal solution is searched in a crowd target space; then, the diversity of solutions is improved through crossover and mutation operation, and therefore the convergence process of the algorithm is accelerated; and finally, continuously iterating and updating the population to ensure that the optimal solution of the global scheme is found. According to the intensive storage network vehicle path crowd target optimization method based on the NSGA-II algorithm provided by the invention, an effective solution is provided for vehicle path optimization of an intensive storage network, and decision makers can be supported to make more reasonable transportation arrangements in a complex logistics environment.

The invention discloses a multi-target multi-bar truss optimization method based on single-target sequence search assistance. The method comprises the following steps: preliminarily locking an angular solution region by using an angular solution search mechanism; introducing a crossover operator to communicate with a population history optimal individual to generate an excellent individual; a mutation operator is introduced, so that the population is more diversified; an adaptive strategy is introduced, so that the offspring population is more excellent; external archives are introduced to improve the error-tolerant rate during individual selection; the angular solution quality is improved by adopting an angular solution detection mechanism; using the obtained angular solution to guide a multi-target algorithm to carry out further diversity search; selecting a plurality of non-dominated solutions using a non-dominated sorting method and the congestion distance; and evaluating the result on a plurality of test problem sets. A two-stage search mode is adopted, in the first stage, targets are searched in sequence on the whole, and an angular solution and a non-dominated solution are obtained; in the second stage, the solution obtained in the first stage serves as an initial population, evolution is conducted through a multi-objective optimization algorithm, and therefore better convergence and non-dominated solution diversity are achieved.

The invention discloses a mobilization resource allocation method based on a multi-objective optimization genetic algorithm, which relates to the technical field of neural networks and resource allocation, and comprises the following steps: defining variables, objective functions and constraint conditions of a mobilization resource allocation scheme, and constructing a mobilization resource allocation model; designing a multi-target genetic algorithm for mobilization resource allocation scheme optimization; generating an initial population; performing genetic manipulation by taking the initial population as a parent population Pt to generate a child population Qt; combining the parent population Pt and the child population Qt to obtain a combined population Rt = Pt union Qt; performing non-inferior classification on the population Rt to obtain each front end and a non-dominated sequence value of each individual in each front end; calculating the crowding distance in each front end of the population Rt to obtain the crowding distance of each individual in each front end; selecting individuals according to the non-dominated sequence value and the crowding distance of each individual in each front end of the population Rt to obtain an updated new population Pt + 1; and judging whether the current number of iterations or the updated new population Pt + 1 meets an end condition or not until the end condition is met, and outputting the updated new population. According to the method, the task load can be ensured to be satisfied when the task is implemented, and the balance of the task load borne by each region is also considered.

The invention provides a heterogeneous cloud multi-target task scheduling method and system based on quantum collaborative optimization, and the method comprises the steps: generating a DAG graph based on a user request, and constructing a chromosome code of charging mechanism collaboration; an initial population is randomly generated, quantum fitness evaluation is performed, non-dominated sorting is accelerated by applying a Grover search algorithm, a quantum-assisted crowding distance is introduced, and a to-be-evolved individual is obtained. Performing crossover and mutation operation on the sample by using a quantum genetic algorithm, and stopping when a convergence condition is reached; and finally, converting a Pareto solution set obtained by the quantum genetic algorithm into pheromones, initializing by using an ant colony algorithm and carrying out routing updating to complete scheduling and distribution. According to the invention, global search is carried out through the quantum genetic algorithm, and local optimization is carried out through the ant colony algorithm, so that the overall efficiency of the task scheduling method is improved; meanwhile, by improving a genetic algorithm chromosome coding mode and redefining a virtual machine cost function, the problem of billing mechanism difference in heterogeneous clouds is solved.

The invention discloses a heterogeneous cluster parallel multi-stage fuzzy job scheduling method, which comprises the following steps of: receiving a processing request which is submitted by a user and contains multi-stage jobs, and modeling task execution time, data transmission time and request deadline into triangular fuzzy numbers, constructing a scheduling model taking the minimization of the total lease cost and the minimization of the request tardiness as targets; carrying out global exploration by adopting an improved second-generation non-dominated sorting genetic algorithm to generate parent and offspring populations; based on population similarity threshold judgment, dynamically triggering local search guided by a multi-agent near-end strategy optimization algorithm, and adaptively selecting a local search operator for each individual to generate an adjacent population; combining various populations, performing non-dominated sorting and crowding distance calculation, and screening out a new generation of populations; and iterating the process until convergence, and outputting a Pareto optimal solution set. According to the method, the problem of multi-target scheduling with fuzzy time variables in a heterogeneous environment is effectively solved, and the quality of a scheduling scheme and the algorithm search efficiency are improved.

The invention discloses a limited transportation resource flexible job shop scheduling system and scheduling method based on a novel evolutionary algorithm, and relates to the technical field of optimal scheduling, and the method comprises the steps: calculating the process completion time of a new scheduling scheme, and generating an optimal scheduling plan; performing non-dominated sorting on the optimal scheduling plan, calculating a congestion distance of a solution, and performing solution set screening to generate an optimal Pareto solution set; and integrating the optimized objective functions, generating comprehensive objective functions, and selecting an optimal scheduling scheme corresponding to the minimum comprehensive objective function based on the optimized Pareto solution set. According to the invention, a self-adaptive crossover and mutation probability adjustment mechanism is introduced through the genetic scheduling module, the global search capability of the evolutionary algorithm is improved, an optimization objective function is constructed through operation data, and the equipment utilization rate and the production efficiency are improved in combination with a self-adaptive evolutionary strategy and multi-objective optimization.

The invention relates to the technical field of target scheduling and warehouse logistics, in particular to a multi-vehicle scheduling method based on a genetic algorithm. Comprising the steps of generating an initial solution by adopting an improved repair mode; taking the total time as a target, performing roulette selection on parents according to the fitness, and performing crossover and mutation operation to generate offspring; repeating improvement and restoration in the initial solution for the filial generation; combining the parent generation and the child generation, and reserving individuals with high fitness to enter the next generation; with the conflict probability as a target, multi-vehicle trajectory planning is optimized through timestamp estimation and an artificial potential field; taking the conflict cost and the total distance as optimization targets, and performing target optimization through a non-dominated sorting genetic algorithm to obtain a non-dominated solution; time and conflict cost are normalized, the Euclidean distance between each solution and an ideal point is calculated, and the solution with the minimum distance is selected as the optimal solution. The method has the advantages that multi-objective optimization is realized through non-dominated sorting and crowding degree distance calculation; conflict risks are considered while global planning is performed, and a reasonable route is planned.

The invention provides a power network three-phase balance optimization method based on a multi-target particle swarm optimization algorithm, which belongs to the technical field of power grid three-phase balance adjustment, and comprises the following steps: acquiring low-voltage transformer area network topology and user power consumption data, initializing a particle swarm population, updating particle positions through a dynamic inertia weight and a self-adaptive learning factor, and optimizing the power grid three-phase balance. The three-phase unbalance degree, the voltage deviation and the economic cost are evaluated in combination with low-voltage power flow calculation, an external archiving mechanism, a crowding distance strategy and variation maintenance solution set diversity are adopted, an optimal phase sequence adjustment scheme and wire diameter configuration are output, and aiming at the characteristics of high impedance and three-phase load imbalance of a low-voltage transformer area, the voltage quality is remarkably improved, and the power consumption is reduced. The method is suitable for intelligent reconstruction of low-voltage distribution networks in residential areas, small industrial and commercial areas and the like, and has good practicability and economical efficiency.

This invention discloses a multi-objective optimization method for high-pile piers and abutments based on an improved Grey Wolf algorithm. The method includes the following steps: establishing a design variable with pile inclination and torsion angle as the design variables, and simultaneously minimizing the standard deviation of pile driving force, maximum pull-out force, and maximum pile bending moment as the optimization objective; designing an integer index discrete encoding strategy to map discrete inclination and torsion angle parameters in the project to integer indices; using an improved multi-objective Grey Wolf optimization algorithm for solving the problem, efficiently searching for Pareto optimal solutions by introducing a nonlinear convergence factor, external archive maintenance, and a leader selection mechanism based on crowding distance; and finally outputting a series of pile placement schemes that achieve a balance between pile driving force uniformity, pull-out force, and bending moment control. This invention achieves collaborative optimization of multiple engineering objectives, with strong algorithm adaptability and significant optimization effects, providing scientific and efficient design decision support for high-pile piers and abutments and other pile foundation projects.

The invention provides a collaborative learning grouping method and system based on a multi-objective optimization algorithm, and the method comprises the steps: 1, collecting student data which comprises the gender, the number of owned skills, the skill mastering degree, the learning style and the score of a pre-repair course; 2, randomly grouping class members to obtain an initial grouping scheme under the condition of ensuring that the number of students with low score of the first course in each group is controllable and the group leader is fixed; 3, calculating a diversity value Dg of a single group, a diversity value Dc of the whole class, a pressure value # imgabs0 # among members in the group and a pressure value Pc of the whole class; 4, gradually improving the grouping quality through a multi-target genetic algorithm NSGA-II; and 5, taking the grouping scheme with the minimum congestion distance in the Pareto leading edge as a final output scheme. According to the method, grouping can be carried out by considering the student attributes in multiple aspects, so that grouping is more scientific and accurate, the advantages of each student are fully played, and good conditions are created for effective collaborative learning.

The invention relates to the technical field of cloud computing and distributed computing scheduling, in particular to a server-free MapReduce multi-target scheduling optimization method. Comprising the following steps: constructing a directed acyclic graph, coding a resource allocation decision into a solution from a source node to a target node, generating an initial population by adopting a hybrid initialization strategy, evaluating fitness, calculating a congestion distance of a solution in each non-dominated leading edge, and sequentially executing tournament selection, a self-adaptive crossover strategy and a topological repair variation strategy. And periodically executing a solution injection strategy guided by heuristic search, carrying out crowding distance selection, updating the global Pareto optimal solution set, outputting the current global Pareto optimal solution set when the time is up, and otherwise, returning to iteration. The method has the positive effects that the execution cost and the execution time of the MapReduce operation are balanced, the global Pareto optimal solution set quality of the multi-target scheduling scheme is improved, and the convergence speed and the robustness of the algorithm are enhanced.

The embodiment of the invention relates to the field of agricultural chemical production base site selection, in particular to an agricultural chemical production base multi-target site selection method and device, equipment and a medium. A specific embodiment of the method comprises the following steps: determining an agricultural chemical production base multi-target site selection model; performing fitness non-dominated sorting processing on the initial population; determining iteration information; executing the following iteration steps: determining a filial generation site selection population corresponding to the initial site selection population; performing combination processing on the initial site selection population and the offspring site selection population; performing non-dominated crowding distance sorting processing on the combined site selection population to generate a target offspring site selection population; executing the following determination steps: updating the iteration information to generate target iteration information; determining a new population for site selection; and determining the site selection new population as the agricultural chemical production base multi-target site selection information. According to the embodiment, when a large-scale population is processed, the efficiency of non-dominated sorting can be improved.

This invention discloses a green scheduling method for unrelated parallel machines using dual factors of electricity and carbon. The method includes the following steps: S1, collecting parameters related to production conditions, and constructing an unrelated parallel machine scheduling model with the goal of maximizing completion time and minimizing electricity and carbon costs; S2, constructing a distributed evolutionary algorithm with adaptive and neighborhood search mechanisms, inputting the parameters related to production conditions, solving the unrelated parallel machine scheduling model, and obtaining an optimal scheduling method. When solving the unrelated parallel machine scheduling model, the algorithm randomly initializes the population, employs a multi-objective non-dominated sorting method and crowding distance calculation to improve individual diversity, and dynamically adjusts the learning rate and mutation rate through an information entropy adaptive mechanism. Simulation results show that the ANEDA algorithm performs significantly better in optimizing maximum completion time and total electricity and carbon costs, verifying the algorithm's effectiveness and practical value.

The invention provides a feature selection method, device and equipment based on self-adaptive multiple tasks and a medium, and relates to the field of feature selection. The method comprises the following steps: obtaining a population scale and a decision space dimension of feature selection, and adaptively initializing to generate an initial population and task number set; then carrying out mixed breeding to obtain a progeny sub-population and a corresponding task number; searching solutions related to the current task number from the sub-population and the descendant population corresponding to each number in the task number set as a candidate solution set; after deduplication, elite solutions are selected from the candidate solution set through non-dominated sorting and crowding distance, and the elite solutions are reserved and updated to maintain the size of the sub-populations unchanged until each number in the task number set is traversed; then, dynamically combining different tasks and sub-populations related to the tasks to obtain a population after multi-task combination; and when a set standard is met, obtaining a final population, otherwise, continuing to breed and iteratively updating. According to the method, the diversity and convergence dynamic balance of the multi-task sub-population is realized.

A method and system for optimizing the configuration of hierarchical protection point selection for a distribution network includes setting multiple objective functions and multiple constraints for protection point selection optimization, wherein the protection point selection optimization is to optimize the number and location of protection points; establishing a constraint violation degree calculation formula based on all constraints; obtaining distribution network parameters, and iteratively optimizing the multiple objective functions based on these parameters using a particle swarm algorithm that introduces a three-dimensional vector rotation matrix and a dynamic perturbation vector. During the optimization process, particles are screened based on the constraint violation degree, and particle priorities are calculated based on the constraint violation degree and congestion distance; a set number of protection points are randomly selected from the multiple protection points optimized, and the final protection points are obtained through clustering and screening. The present invention realizes three-dimensional exploration of the multidimensional target space of distribution network protection point selection, significantly improving the distribution uniformity and global search capability of the solution set.

The invention discloses a multi-modal multi-target travel salesman problem optimization method, which is used for searching more equivalent optimal solutions. An edge crossing strategy adopting non-dominated sorting is introduced to generate excellent individuals; the Spearman distance is combined with a diversity measurement standard based on the same edge in a decision space, and the similarity between different loop solutions is allowed to be evaluated more accurately; in addition, a concept of a special crowding distance is introduced in an environment selection process so as to ensure that a population keeps diversity in a balanced manner in a decision space and a target space; and after the evolution is finished, selecting the best individual to evaluate on the test set. According to the method, the multi-target traveling salesman problem is solved through MMTSPDS, and the algorithm provides various different solutions for the traveling salesman problem by storing a plurality of high-quality multi-modal solutions, so that the scientificity and flexibility of decision making are improved.

The invention discloses a LID facility space optimization layout method coupling SWMM model and GDE3 algorithm, the method steps include: constructing SWMM model to simulate and evaluate hydrological effects of different LID facility space layouts, and obtaining the runoff reduction rate of the study area; constructing a cost calculator to obtain the total construction cost under different LID facility space layouts; establishing a multi-objective decision model for LID facility space layout, determining the decision variables of LID layout, taking the minimum total construction cost of LID facilities and the maximum runoff reduction rate of the study area as the objective function, and determining the boundary conditions; GDE3 algorithm generates decision variables through crossover mutation, inputs them into SWMM model and cost calculator, obtains the runoff reduction rate and total construction cost of the study area, and uses non-dominated sorting and congestion distance calculation to obtain the optimal solution set of decision variables in the multi-objective decision model for LID facility space layout. The invention can improve the search capability and obtain a higher runoff reduction rate.

The application belongs to the technical field of semi-active suspension control method, and relates to a semi-active suspension control method based on an improved multi-objective particle swarm optimization algorithm, which comprises self-adaptive grid division of an external archive of the MOPSO algorithm to determine a global optimum with the most diversity and a global optimum with the most convergence; evolution state of a population is judged according to a distribution entropy change quantity of all optimal solutions in the external archive, and a global optimum gBest for guiding evolution is selected; flight parameters of population evolution are adjusted according to a change quantity of local crowding distance of current population evolution to balance evolution trend of the population; the above two types of core mechanisms are fused, a multi-information fusion multi-objective particle swarm optimization algorithm is proposed, and the algorithm is used for determination of weight coefficients of a semi-active suspension LQR control strategy. The novel multi-objective optimization algorithm effectively realizes balance between convergence precision and population diversity by synergistically integrating external archive information and current population dynamic information.

The invention belongs to the technical field of unmanned underwater vehicle path planning, and discloses an unmanned underwater vehicle multi-target path planning method and system. According to the multi-target path planning method for the unmanned underwater vehicle, the path length and the path safety serve as optimization targets, firstly, environment modeling is carried out by adopting a grid method combined with a topographic information gray value, and a path smoothness strategy is proposed according to a Bessel function; secondly, introducing an improved particle swarm optimization algorithm, and realizing the balance of diversity and convergence through crowding distance sorting and a Pareto frontier updating strategy; and finally, through non-dominated solution screening, outputting an optimal path solution set. According to the path planning method, the path searching capability and the path distribution diversity can be effectively improved during multi-target path planning, the method has high engineering applicability, and especially under the double consideration of the path length and the path safety index, the good solving and planning capability can be kept.

The invention relates to the technical field of snack product research and development, and discloses a snack product prototype promotion method and device based on deep learning and a cloud platform. The method comprises the following steps: constructing a time sequence multi-modal training set containing formula, process and business metadata and sensory evaluation values, firstly establishing a first model by using a deep neural network, and implementing optimization training for multiple stages such as initialization, data sampling, buffer area capacity, data disruption and hyper-parameter search; the second model takes the physical and chemical simulation value as the input and outputs the corrected predicted value to replace the original theoretical value, the first model is retrained, and then a Pareto frontier-crowding degree distance multi-target algorithm is utilized to efficiently screen out a prototype with a high sensory score from the candidate library for trial production; and if the actually measured deviation exceeds a threshold value, actually measured data is returned to trigger incremental learning, only the last two layers of the model are finely adjusted, and an exponential attenuation weight and an elastic weight are adopted for consolidation to prevent catastrophic forgetting, so that data-model closed-loop iteration is realized, and snack prototypes meeting market requirements are continuously output.

The invention discloses a pump station multi-objective optimization scheduling method based on a multi-strategy enhanced African vulture algorithm, and relates to the technical field of water conservancy projects. The method comprises the following steps: acquiring basic data of pump station operation, and establishing a multi-target optimization scheduling model which takes minimization of electric charge and minimization of the quadratic sum of water shortage of a water receiving area as target functions during the pump station operation period and takes water pump power constraint, water pump pumping capacity constraint and starting-up water pump number constraint as constraint conditions; according to the multi-strategy enhanced multi-target African vulture algorithm, a crowding degree distance sorting mechanism and a hypercube mechanism are adopted to select double leaders to guide a population, an archive-based crossover and mutation strategy is adopted to reduce the possibility that the multi-target algorithm falls into local optimum, and a progressive repair strategy is adopted to repair an out-of-border solution after a main loop is finished. And the multi-strategy enhanced multi-target African vulture algorithm is adopted to carry out pump station multi-target optimization scheduling simulation, so that the operation efficiency of the pump station can be improved.

The present application relates to cloud computing and distributed computing scheduling technical field, specifically belong to a kind of serverless MapReduce multi-objective scheduling optimization method, including: constructing directed acyclic graph, resource allocation decision is encoded as the solution from source node to target node, initial population is generated using hybrid initialization strategy, fitness evaluation, the crowded distance of solution in each non-dominated front is calculated, in turn execute tournament selection, adaptive crossover strategy, topological repair mutation strategy, periodically execute heuristic search guided solution injection strategy, crowded distance selection is carried out, the global Pareto optimal solution set is updated, the global Pareto optimal solution set of current time output is achieved, otherwise return iteration. The present application has the positive effect of balancing the execution cost and execution time of MapReduce job, improving the global Pareto optimal solution set quality of multi-objective scheduling scheme, enhancing the convergence speed and robustness of algorithm.