融合小波阈值降噪和INFO-LSSVM算法的化工过程参数预测

Chemical Process Parameter Prediction Based on Combining Wavelet Threshold Denoising and INFO-LSSVM Algorithm

针对化工过程非线性、时滞性高的实际问题,以某氨合成工段为研究对象,首次提出基于小波阈值降噪的向量加权平均算法(INFO)优化最小二乘支持向量机(LSSVM)的氨合成塔预测方法 INFOLSSVM,将小波降噪理论与INFO-LSSVM算法结合创建了预测精度较高的氨合成塔预测模型,用模型筛选出6个氨合成工段中的样本数据进行小波阈值降噪预处理,然后用INFO-LSSVM训练降噪后的数据,得到小波阈值降噪的INFO-LSSVM氨合成塔预测模型,对比INFO-LSSVM、PSO-LSSVM、LSSVM模型的预测结果,得到三者的均方根误差分别为0.231 8、0.447 7、0.496 4,可为多因素作用下类似预测提供借鉴。

Aiming at the chemical process with nonlinearity and high time lag and through taking the ammonia synthesis section of a chemical enterprise in Yunnan Province as the object of research,a prediction method for ammonia synthesis tower which having the vector weighted average algorithm(INFO)based on wavelet threshold denoising adopted to optimize least-squares support vector machine(LSSVM) was proposed,in which,having the wavelet denoising theory combined with INFO-LSSVM algorithm to create an ammonia synthesis tower prediction model with high prediction accuracy,including having the sample data screened from six ammonia synthesis sections for wavelet threshold denoising and then having the denoising data trained with the INFO-LSSVM algorithm so as to obtain INFO-LSSVM ammonia synthesis tower model with wavelet threshold denoising.Comparing its prediction results with those of PSO-LSSVM and LSSVM shows that,the root mean square errors of the three prediction results are 0.231 8,0.447 7,and 0.496 4,respectively and it provides a reference for similar prediction under the effect of multiple factors.