Latent structure mapping algorithm-based soft measurement method for DXN discharge concentration in municipal waste solid incineration process

Abstract

The invention discloses a latent structure mapping algorithm-based soft measurement method for DXN discharge concentration in a municipal waste solid incineration process. The method comprises the steps of firstly, collecting and preprocessing process variables of the complete municipal waste solid incineration process and easily detected exhaust gas concentration; secondly, based on variable importance in projection (VIP) of a linear latent structure mapping algorithm and an input feature selection ratio set according to experience, determining input features of a soft measurement model; andfinally, building a selective ensemble kernel latent structure mapping model based on adaptive selection kernel parameters of a control training sample ensemble construction strategy. The method can preset a feature selection parameter and a structure parameter of the soft measurement model and a selection threshold and a weighting strategy of an ensemble sub-model according to actual demands of an industrial process, and is suitable to build the small sample collinear data-based difficultly detected parameter soft measurement model.

Description

technical field [0001] The invention belongs to the technical field of solid waste discharge, and in particular relates to a soft measurement method for dioxin discharge concentration in a solid waste incineration process based on a latent structure mapping algorithm. Background technique [0002] The advantages of solid waste incineration (MWSI) in terms of harmlessness, reduction and recycling of domestic waste are more significant than those of solid waste landfill, but the dioxin (DXN) emitted by this process is extremely chemically and thermally stable. The highly toxic persistent organic pollutants are known as the "poison of the century" [1], especially the significant accumulation and amplification effect of DXN in organisms will cause huge actual and potential harm to the ecological environment and human health[ 2]. MWSI is a typical complex industrial process with the characteristics of multi-variable, strong coupling, large inertia, nonlinear, etc. It is composed...

AI Technical Summary

Problems solved by technology

[0004] Facing the DXN emission concentration modeling problem, using some key process variables and easy-to-detect gas concentrations in the solid waste incineration process, literature [11] constructed a simple linear regression model for different types of incinerators. Obviously, it is difficult to describe the DXN emission concentration model Inherent nonlinear characteristics; literature [12] built a nonlinear model based on genetic programming (GP), its predictive performance is stronger than multiple linear regression and BP neural network modeling methods, but its generalization performance still needs to be improved; literature [13] proposed the DXN emission concentration soft sensor using genetic algorithm to optimize the BP neural network model, but when small sample data is used for modeling, the inherent random characteristics of this method will make it difficult to obtain stable prediction performance; literature [14] proposes to adopt Resampling and noise injection processing of small sample data increases the number of samples, and then constructs a DXN emission concentration model based on the maximum entropy neural network; the above methods all use a single BP neural network to construct a soft sensor model, which is difficult to overcome the inherent characteristics of neural network modeling algorithms. It is easy to fall into the local minimum, poor prediction performance stability when facing small sample data modeling, etc.

[0005] Studies have shown that the latent structure mapping algorithm and its kernel version can effectively extract linear / nonlinear latent variables related to both input and output data to construct a soft sensor model, which can eliminate collinearity between input features and reduce the requirement for the number of modeling samples[ 15]; but it also has the following problems: too many input features will reduce the generalization performance and interpretability of the soft sensor model, and it is difficult to effectively select the kernel parameters suitable for modeling the characteristics of the sample itself[16]

The selective ensemble algorithm using the ensemble construction strategy of "training sample resampling" verifies that integrating partially available candidate sub-models can obtain better generalization performance than integrating all candidate sub-models [19], but the BP neural network algorithm used Not suitable for modeling small samples and complex distribution data

For multi-source small-sample high-dimensional collinear data, literature [20] proposed a two-layer selective integration latent structure mapping modeling method that comprehensively considers multi-source features and multi-working condition samples, but the soft sensor model constructed is complex High, limited universality and other shortcomings, and at the same time, the input features are not effectively reduced, so it is difficult to apply to small-sample low-dimensional data modeling

In summary, the selective integration strategy that integrates functions such as input feature selection, kernel parameter adaptive selection, and model complexity control based on actual needs has not been reported in DXN emission concentration soft sensor modeling.

Images