ss5:A Neural Network-based Energy Consumption Prediction Model for Feature Selection and Paremeter Optimization of Winders

Abstract

Textile industry has become the third largest energy consuming industry after engineering and chemical sectors. In order to reduce the energy consumption in the textile industry, a neural network is used to establish the energy consumption prediction model of the winder. In this research, the model is specially designed as the objective function to optimize the energy consumption of the winders. Firstly, the neural network error back propagation is analyzed and the absolute values of the weight coefficient matrix product are used to approximate the influence of input parameters on the model output. The values are also used to select the core parameters to optimize the model. Secondly, the single-dimensional search method is applied for a set of parameter values within a reasonable interval of the whole input parameters to reduce the energy consumption. Experimental results indicate that a set of core parameters can be determined to remodel after the training of the neural network model. In addition, a set of parameter values obtained by single-dimensional search can also be used to effectively reduce the energy consumption of the winders. The proposed method effectively solves the problem and is efficient and straightforward. The feasibility of the proposed approach is validated through the comparative analysis.