Calorific Value Prediction of Coal Based on Least Squares Support Vector Regression

The calorific value of coal is important in both the direct use and conversion into other fuel forms of coals. Accurate calorific value predicting is essential in ensuring the economic, efficient, and safe operation of thermal power plants. Least squares support vector machine (LSSVM) is a variation of the classical SVM, which has minimal computational complexity and fast calculation. This paper presents Least squares support vector regression (LSSVR) to predict the calorific value of coal in Shanxi Coal Mining Region. The LSSVR model takes full advantage of the calorific value information. It derives excellent prediction accuracy, including the relative errors of less than 3.4% and relatively high determination coefficients. Experimental results conform the engineering application, and show LSSVR as a promising method for accurate prediction of coal quality.