Modeling multi-pollutant emission concentrations in municipal solid waste incineration processes using virtual-real data-driven approach

Abstract

The concentration of pollutant emissions during the municipal solid waste incineration (MSWI) process has a significant global impact on the atmospheric environment. Developing effective pollutant emission models to support optimization for emission reduction is a critical challenge that must be addressed. To address the challenges of high uncertainty and poor interpretability in pollutant emission concentration models for the MSWI process, this article proposes a novel method for modeling multi-pollutant emission concentrations using a virtual-real data-driven method. First, a whole-process numerical simulation model for the MSWI process is developed using a multi-software coupling strategy. Virtual simulation mechanism dataset under diverse operating conditions is generated through a combination of orthogonal experimental design and implementation. Subsequently, to tackle the challenge of limited sample size resulting from the high cost of simulation, virtual sample generation (VSG) is utilized to enhance the dataset. Finally, a virtual-real data-driven multi-pollutant emission concentration model is developed, leveraging the Interval Type-2 Fuzzy Broad Learning System (IT2FBLS) and the Linear Regression Decision Tree (LRDT) algorithm with a main-compensation mechanism. The proposed methodology is validated using data from an MSWI power plant in Beijing.