Research On ELM Based On Improved Sparrow Search Algorithm And Its Application In Ethylene Plant

Petrochemical industry provides a strong guarantee for economic construction and people’s life,while ethylene,as a typical petrochemical product,has complex operation process and high energy consumption.It is an effective way to save energy and reduce emission to find the direction of production allocation and energy efficiency optimization by establishing prediction model.Aiming at the deficiency of the sparrow search algorithm that tends to be caught in the local optimum,this paper presents an improved sparrow search algorithm.It is applied to solve the problem of poor generalization performance caused by random generation of input weights and hidden layer thresholds of extreme learning machines,and an extreme learning machine based on it is proposed.The production capacity prediction model of ethylene plant is established,which can realize effective prediction and optimization guidance of production capacity.The main work of this paper is as follows:(1)To solve the problem that the sparrow search algorithm is difficult to jump out when it converges to the local optimum,an improved sparrow search algorithm based on multi-strategy is proposed.By dynamically controlling the proportion of finder sparrows with the number of iterations,a large proportion of finders in the early stage can ensure global exploration,and a large number of entrants in the later stage can strengthen local development.At the same time,at the end of iteration,the firefly strategy is used to disturb the population position,enlarge the search range and enhance the optimization ability.The results reveal that the improved sparrow lgorithm is superior to other intelligent optimization algorithms by optimizing 8 benchmark functions.(2)Focus on the problem of poor generalization performance caused by random generation of input weights and hidden layer thresholds of extreme learning machine,an optimized extreme learning machine based on improved sparrow search algorithm is proposed.The weights and thresholds are optimized by improved sparrow search algorithm,and the generalization ability of modeling is improved.Compared with back propagation network,radial basis network,traditional extreme learning machine and other optimization algorithm improved extreme learning machine,the generalization performance and prediction ability of UCI data modeling are verified.The proposed method is applied to the ethylene input-output data,and the production capacity prediction model of ethylene plant is built.Compared with other prediction modeling methods,the effectiveness of the presented model is verified,which offered guidance for the optimization of material use plan and the reduction of carbon emission.(3)Based on the research of improved sparrow search algorithm and optimized extreme learning machine,the prototype system of ethylene plant production capacity prediction is designed and developed.After the overall design and functional design,modules of data display,production capacity prediction and optimization direction guidance are realized based on the production data of ethylene plant,providing managers with a convenient management and control platform integrating visualization,prediction modeling and optimization guidance.