Dual Prediction Model and Improvement of Coal Ash Melting Point Based on SMA and Optimal Input Parameters

In order to improve the accuracy of coal ash melting point prediction, the original model was improved in terms of modeling method, input parameters, and generalization ability. Firstly, to address the problem of partial distortion of the ash melting point data, a dual prediction model consisting of Model I, which classifies the ash melting point, and Model II, which accurately predicts the ash melting point, is established by using a support vector machine, and the distorted data are modeled separately from the real data. Secondly, the optimal input parameters of the model are obtained by comparing the prediction results of 256 different input parameters, and the hyperparameters of the model are searched for by the slime mould algorithm (SMA) search. Finally, in order to enhance the generalization ability of the model, an evaluation index considering overfitting is proposed to improve the model. The results show that compared with the existing single prediction model, the mean square error (MSE) of the dual prediction model is reduced by 6.84 %; based on this, the optimal selection of input parameters reduces the MSE by 24.57%; only relying on the SMA to search for hyperparameters leads to overfitting of the model and deterioration of prediction effect; improvement of the model using evaluation index that consider overfitting results in a 4.16% reduction in mean squared error compared to the pre-improvement period. In the end, the improved model reduced the mean square error by 32.31 % compared with the original model, which greatly improved the prediction accuracy.