基于改进小波神经网络的中国天然气消费量多情景预测研究

Multi-scenario Prediction of Natural Gas Consumption in China based on Improved Wavelet Neural Network

天然气是人民生活和工业生产的一种主要能源,随着天然气使用的逐步推广和天然气消费量的快速增长,正确合理地对天然气消费量进行预测有着重要而深远的意义。利用灰色关联度分析法筛选出影响天然气消费量的六个主要因素——人均GDP、天然气生产总量、居民消费水平、用气人口、城镇化率和管道运输长度,在此基础上提出了基于遗传算法优化的小波神经网络预测模型,以1995-2013年中国天然气消费量的统计数据为例对预测模型进行训练和检验,并将该预测方法与传统BP网络、小波神经网络预测方法作对比,预测结果表明,该方法的预测精度更高。遗传算法和小波神经网络相结合的预测方法,可以科学预测我国未来年份的天然气消费量,有效地提高了预测精度,为天然气行业的发展提供理论依据和实证支持。

Natural gas is one of the main energy resources for people’s life and industrial production,with the gradual promotion of natural gas and the rapid growth in natural gas consumption,precisely simulating and forecasting the consumption of the natural gas in the future have great significance for pipeline con-struction,capacity increase decision-making,network design of compressed natural gas filling stations.Us-ing the correlation analysis has selected five main factors influencing consumption of natural gas-the gross domestic product,the total gas production,the level of consumption,population with gas and pipeline length.On this basis,we provide a genetic algorithm based on wavelet artificial neural networks for simulat-ing and forecasting the natural gas consumption.We take the natural gas consumption data of China in 1995-2013 as the sample data for training and testing the forecasting model,and compare the provided forecasting model with the traditional artificial neural networks.The results indicate that the provided forecasting model can effectively improve the forecasting accuracy.More importantly,it can scientifically predict con-sumption of natural gas for the coming years and provide theoretical basis and empirical support for the de-velopment of the gas industry.