Multi-Temperature Synchronous Prediction of Municipal Solid Waste Incineration Based on a Non-Stationary Crossformer

In this study, we designed a multi-temperature synchronous prediction method for municipal solid waste incineration (MSWI) based on a non-stationary Crossformer model. It is mainly divided into the following steps: First, use a normalization module to normalize the preprocessed temperature sample data. Secondly, add a non-stationary attention module to calculate the non-stationary attention of a single time series. Next, construct an attention convolutional network to extract non-stationary attention features, and based on this, calculate the attention between multiple temperature time series. Finally, perform data restoration based on a de-normalization module. By using the above methods, not only can the data characteristics lost during the model’s over-processing of the original temperature data during training be restored, but also the deeper level of data features and the correlation between various temperatures can be captured. The non-stationary recovery and correlation calculation of multi-temperature time series were achieved simultaneously. The experimental results based on actual data from Chinese MSWI factories show that this method has high accuracy in the multi-temperature synchronous prediction of MSWI.