Furnace temperature prediction model in municipal solid waste incineration process based on slow feature analysis and sparse stochastic configuration network

Abstract

The municipal solid waste incineration process inherently exhibits a significant time lag, making traditional furnace temperature monitoring methods ineffective in promptly capturing temperature fluctuations. To address this challenge, this paper proposes a furnace temperature prediction model integrating slow feature analysis with a sparse stochastic configuration network. Specifically, mutual information combined with slow feature analysis is employed to extract latent variables from high-dimensional features. A sparse prior is introduced to constrain the effective number of output weights, thereby constructing the sparse stochastic configuration network for furnace temperature prediction. Furthermore, hyperparameters of the proposed model and sparse solutions for output weights are derived via Bayes' theorem and maximum a posteriori estimation. Validation using a dataset from a waste-to-energy power plant demonstrates that the model achieves accurate prediction of furnace temperature changes, highlighting its substantial potential for optimizing control throughout the incineration process.