基于粒子群优化支持向量机的火电厂主汽温预测模型

Main Stream Temperature Forecasting Model Based on QPSO and SVM in Power Plant

针对支持向量机在大规模训练中算法收敛速度慢、复杂程度高等问题,采用量子粒子群算法选取最小二乘支持向量机的模型参数,避免了人为选择参数的盲目性,提高了预测模型的训练速度和泛化能力.实验结果表明,该算法具有容易实现、节省计算成本、提高收敛速度等优点,应用于火电锅炉主汽温预测模型,取得良好的效果.

According to the low speed of constringency and high complexity of training methods in SVM large scales training, quantum-behaved particle swarm algorithm （QPSO） is presented to solve the problem. Parameters selection is an important problem in the research area of support vector machines. Meanwhile, quantum-behaved particle swarm algorithm is used to choose the parameters of least square support vector machines, which can avoid the man-made blindness and enhance the efficiency and capability of forecasting. The experimental results indicate that this QPSO-SVM forecasting model can be trained quickly and good generalization, is easy to be realized, can save the calculating cost and improve the constringency speed.