Identification and prediction of nonlinear multi-parameter based on least squares support vector machine

Abstract

The circulating fluidized bed boiler is key equipment in safety coal gangue power generation, after analyzing of the typical fault and hidden fault of main components of boiler system, and the treating methods, directed towards the nonlinear characteristics of the oxygen content of flue gas, which has many factors influence, a method based on least squares support vector machine (LS-SVM) used in flue gas oxygen content model recognition is proposed. the measured crucial parameter of influence the stable operation of the boiler are used to identification and prediction, including the oxygen content of the flue gas, coal gangue flow and material return pressure imitation, as the results show that this method has higher precision (the error is less than 7‰); Compared with general SVM and improved BP, it has higher precision and reduces the complexity of the calculation.