基于Gibbs抽样算法的造纸工艺能源消耗预测模型仿真研究

Simulation Research on Energy Consumption Prediction Model of Papermaking Process Based on Gibbs Sampling Algorithm

针对造纸工艺能源消耗预测模型效果不佳的整体现象,基于Gibbs抽样算法设计了造纸工艺能源消耗预测模型,通过以贝叶斯参数估计为载体的马尔科夫蒙特卡洛算法优化传统预测模型,以能源消耗数据特性合理选择最佳先验分布,生成与目标分布要求相符的马氏链,以Gibbs抽样算法进行参数后验分布均值提取分析,当作模型参数估计值。同时进行模型仿真试验验证,结果表明,优化前模型与优化后模型预测值都可准确拟合真实值,但是就整体拟合效果来讲,贝叶斯参数估计算法能源消耗预测模型的效果更优;且贝叶斯参数估计算法能源消耗预测模型的预测精确度较高,平均绝对百分误差相对更小。

A papermaking process energy consumption prediction model based on Gibbs sampling algorithm was designed,the traditional prediction model by Markov Monte Carlo algorithm based on Bayesian parameter estimation was optimized,the best prior distribution according to the characteristics of energy consumption data was reasonably selected,the Markov chain which meet the requirements of target distribution was generated,and Gibbs sampling algorithm was used to extract and analyze the mean value of parameter posterior distribution as ARIMA model parameter estimation.The results showed that both the pre-optimized model and the pre-optimized model fit the real value accurately,but the Bayesian parameter estimation algorithm ARIMA better for the overall fitting effect,and the Bayesian parameter estimation algorithm had higher prediction accuracy ARIMA the model,and the average absolute percentage error was relatively small.