Abstract: Municipal solid-waste incineration (MSWI) is a waste resource treatment technology widely used in China and around the world. The emission of dioxins (DXN), a highly toxic and persistent pollutant emitted during the MSWI process, is one of the main reasons for the "not in my backyard" response to incineration plants. The long-period, high-cost offline detection of DXN and the complex mechanism of the incineration process make it difficult to achieve real-time monitoring and optimal control of the DXN emission concentration. To solve these problems, we propose a soft-sensing method based on multi-layer feature selection to monitor the DXN emission concentration in the MSWI process. First, using the correlation between a single feature and DXN combined with their correlation coefficient and mutual information, a comprehensive evaluation index is constructed to realize the first-level selection of process variables in multiple subsystems of the MSWI process. Then, to prevent multi-feature redundancy and obtain robust small-sample-data feature selection, a feature-selection algorithm based on a genetic algorithm-partial least squares (GA-PLS) is run many times to realize second-level feature selection. Finally, third-level feature selection is realized by combining the statistical frequency of the upper-level selection feature and the prediction performance of the model. The effectiveness of the proposed method is verified by combining the DXN test data of an incineration plant for many years.