造纸企业工艺过程能源消耗预测仿真

Prediction and Simulation of Energy Consumption in the process of Paper Making Enterprises

能耗预测是企业能源管理系统中的重要任务之一,如何准确、高效的进行能源消耗预测十分具有挑战性。针对现有造纸企业能耗预测结果精度较低的问题,提出采用贝叶斯估计的马尔科夫蒙特卡洛算法对传统的自回归积分滑动平均模型进行改进。根据能耗数据的特点选取恰当的先验分布,并生成一条满足目标分布的马氏链,利用Gibbs抽样算法提取参数的后验分布均值作为模型参数估计值。实验表明,上述算法在造纸企业能源消耗预测上有较高精度。

One of the important missions in the process of industrial production is the prediction of energy consumption. Energy saving and consumption reduction have attracted more and more attention of enterprises,how to accurately and efficiently predict the energy consumption is very challenging. Aiming at the low prediction accuracy of existing method,we proposed one method which optimizes the traditional ARIMA model using Bayesian estimation and Markov Monte Carlo algorithm. Selecting the appropriate prior distribution of the characteristics of energy consumption data and generating a Markov chain to meet the target distribution,we took use of Gibbs sampling algorithm to extract the mean value of parameters posterior distribution as the model estimation parameters. The algorithm is proved more effective by conducting the comparative experiment. The results show that the algorithm has high accuracy in the prediction of the energy consumption of paper making enterprises.