基于支持向量机的水电机组状态趋势预测研究

趋势预测是水电机组状态监测与故障诊断系统中的重要内容之一，对保障机组安全稳定运行具有重要意义。针对水电机组振动的非线性、非平稳特性，提出将最小二乘支持向量机应用于水电机组振动状态趋势预测；并将该算法应用于某水电机组振动序列峰峰值的预测，与BP神经网络的预测结果的对比表明：最小二乘支持向量机算法更适合于水电机组状态趋势预测分析。

Research on Condition Trend Prediction of Hydro-turbine Generating Unit based on Support Vector Machines

The trend prediction is one of the most important tasks in the condition monitoring and fault diagnosis system of hydroturbine generating units(HGU),which is great helpful for the stability and security of the running of HGU.According to the nonlinear and nonstationary characteristic in vibration of HGU,the theory of least squares support vector machines(LS-SVM) is introduced to the vibration trend prediction of HGU.The algorithm of LS-SVM is applied in the peak-peak value of a vibration series of a certain Kaplan HGU.The comparison between BP neural network and LS-SVM shows that the prediction results of LS-SVM are more suitable for the trend prediction of HGU than that of BP neural network.