基于灰色-BP网络以及马尔科夫算法的天然气年消费量预测

Prediction of Annual Consumption of Natural Gas Based on Grey-BP Network and Markov Algorithm

采用灰色关联度分析法对我国天然气年消费量的影响因素进行筛选,将灰色关联度数值较大的影响因素作为BP(Back-propagation)神经网络预测的输入向量,进而建立灰色-BP神经网络模型对我国年天然气年消费量进行预测,利用天然气消费量的预测值与实际值的误差得出适用于本次预测的马尔科夫矩阵,再采用马尔科夫矩阵对进一步的天然气消费量预测数值进行误差修正,进而得到更为精准的预测结果。以往数据表明,我国天然气年消费量的主要影响因素为天然气产量及进口量、煤炭、石油以及其他能源消费量。通过灰色-BP模型预测得出的天然气年消费量数值与天然气的实际年消费量数值进行比较,预测结果显示灰色-BP神经网络模型适用于对我国天然气年消费量的预测,通过得到的马尔科夫矩阵对进一步的预测结果进行修正,经修正后误差有减小趋势,说明马尔科夫法可有效减小预测误差。

The factors affecting the annual consumption of natural gas in China were screened by the grey correlation analysis method.The influence factors of the large grey relational value were used as the input vector of the BP(Back-propagation)neural network prediction,and the grey-BP neural network model was established to predict China’s annual natural gas consumption.The error of the predicted value of gas consumption and the actual value was used to obtain the Markoff matrix which was suitable for this prediction,and then the Markoff matrix was used to further correct the prediction value of natural gas consumption,and then a more accurate prediction result was obtained.Previous data showed that the main factors affecting the annual consumption of natural gas were natural gas output and import quantum,the consumption of coal,oil and other energy.The annual consumption value of natural gas predicted by the grey-BP model was compared with the actual annual consumption of natural gas.The prediction results showed that the grey-BP neural network model was suitable for the forecast of annual consumption of natural gas in China.The results of the further prediction were amended by the Markoff matrix,and the post error had a decreasing trend,showing that Marco’s method can effectively reduce the prediction error.