An online prediction method of dioxin emissions based on generative path clustering and box-cox transformation

Abstract

The invention discloses an online prediction method of dioxin discharge based on generation path clustering and Box-Cox transformation, relates to the field of dioxin prediction, and includes the following steps: S1, establishing a data set; S2, cluster analysis, according to different Different dioxin generation pathways play a leading role in the dioxin emission concentration area, and the initial data sets of different clusters are divided into different clusters; S3, the optimal transformation coefficient λ is obtained; S4, the final model is established, and the optimal transformation obtained in step S3 The coefficient λ is used as the conversion coefficient of the dioxin emission data, and the converted dioxin toxicity equivalent is calculated according to the least square method and the online measurement of the chlorobenzene concentration x i Between the linear model coefficients, a predictive model is built. The method has few parameters and stable extrapolation performance, making it an ideal predictive model, moreover, the developed method is beneficial to realize on-line measurement of dioxin emissions in similar incineration processes with dioxin emissions, and feedback control to limit emissions .

Description

technical field [0001] The invention relates to the field of dioxin prediction, in particular to an online dioxin emission prediction method based on generation path clustering and Box-Cox transformation. Background technique [0002] The conventional method detects dioxin emissions from the incineration process by sampling flue gas, followed by more than a week of laboratory pretreatment and high-resolution gas chromatography / high-resolution mass spectrometry analysis. Because the conventional method includes experimental processes such as Soxhlet extraction, purification and concentration, the detection cost is high and the cycle is long. Therefore, the frequency of dioxin monitoring for the incineration process should be at least once a year. Traditional measurement methods are difficult to meet the public's requirements for real-time access to dioxin emissions. A long detection period cannot control the operating conditions of the incineration plant and the fuel compos...

AI Technical Summary

Problems solved by technology

Machine learning based on neural networks predicts dioxin emissions through a large number of observed variables and control variables other than indicators. However, the large number of parameters of machine learning with small sample and high-dimensional data sets limits the effectiveness of this method. characteristics, resulting in the limitation that it can only be applied in a specific situation of a specific incineration plant

Images