Application Research Of Resource Optimal Allocation Model Based On LSTM And NSGA2 Algorithm

As the "Made in China 2025" national strategy continues to advance,it is becoming more and more important to use artificial intelligence technology to help traditional industries optimize the allocation of production resources to reduce production costs.This article takes the production of ammonia in traditional industries as an example,uses deep learning and multi-objective optimization algorithms to build a set of combined models,and finds an optimal solution to guide production so that production can reduce production costs while meeting predetermined goals.First,in deep learning modeling,this paper uses Long Short-Term Memory(LSTM)to build deep learning models.LSTM neural networks have become the most popular due to their ability to process data with complex time correlation One of the timing modeling tools.In this paper,the first aspect of LSTM neural network modeling is to select the appropriate excitation function,loss function,optimization method,learning rate and other parameters according to the logical structure of the LSTM neural network and the actual problem background.Second,through the use of grid search to optimize the parameters of the number of hidden layers,the number of neurons and the length of the data sequence of the LSTM network,the effects of the three parameters on the performance of the LSTM network under different combinations are explored.Third,compare the LSTM network with the BP and RNN networks to find that LSTM has the best performance and is more suitable for modeling and analyzing this problem.Secondly,in terms of multi-objective optimization modeling,this paper adopts a non-dominated sorting elite strategy genetic algorithm(NSGA2)with an elite strategy.It has the advantages of optimal individual retention mechanism and parameter sharing.In the construction of the NSGA2 algorithm model,first,based on the established LSTM model,the problem description of multi-objective optimization is combined with the actual problem background,including the determination of the objective function,the choice of optimization direction,and the setting of constraints.Second,determine the coding method of the optimization problem and the selection of crossover and mutation genetic operators in the NSGA2 algorithm.In this paper,the spatial evaluation index and the reverse generation distance are used as the evaluation indexes of model performance,and the relationship between different genetic operators is explored.The effect of the combination on the performance of the NSGA2 model.Third,repeat the simulation experiment on the NSGA2 model and count the experimental errors.It is concluded that the model error fluctuation is small and within the allowable acceptance range,the model has good stability.Finally,this article conducted an example analysis of the established LSTM-NSGA2 combination model,taking the real data results as a comparative reference,the results show that the optimal solution obtained by the LSTM-NSGA2 model established in this article can be based on meeting the production requirements Reduce the cost of raw materials for production.Compared with the previous application research results,this paper mainly improves and expands in the following aspects.First,because the data used in this paper is time series data,the first type of modeling analysis is the traditional time series analysis.Methods such as autoregressive models or differential autoregressive moving average models,but such methods can only handle stationary data or perform autoregressive analysis.The other type uses BP neural network,but the general neural network model modeling defaults that the training samples are independent of each other so as to automatically ignore the correlation of each sample in time sequence,and the LSTM network used in this article can make up The shortcomings of class methods.Second,regarding such multi-objective optimization problems,most of the traditional processing methods are to merge multiple objective functions to be optimized into a single objective function through hierarchical sequence and other methods,and optimize the single objective function on this basis Solve,but the method has problems such as unit inconsistency,which leads to poor performance in practice.The NSGA2 multi-objective optimization algorithm used in this paper can effectively solve the above shortcomings.Third,the traditional industrial production optimization mainly relies on the mechanism model,but the actual production process is complicated and involves many influencing factors,so it is difficult to rely on the mechanism model to accurately describe and analyze the relationship between its input and output.Mold technology can effectively solve the above problems.