Abstract: To improve the robustness and accuracy of a data-driven prediction model for furnace temperature in municipal solid waste incineration(MSWI), we propose a modeling method leveraging weighted supervision and double mapping to enhance the stochastic configuration network (WS-DM-SCN) and construct the furnace temperature prediction model. First, we design a weighting matrix based on the kernel risk-sensitive mean p-power error function, which is embedded into the supervision mechanism of SCN to constrain the connection weight configuration and bias parameters in the presence of noise. Second, we linearly integrate the Softplus and Sigmoid functions as the activation function of the SCN hidden layer to realize the double mapping transformation of the hidden layer features during the training. Thus, we confirm the convergence of the SCN model with the two improved strategies. The results of the comparative experiments reveal that the error obtained with WS-DM-SCN decreased under the effects of different outlier proportions, thereby verifying the effectiveness of the proposed method.