Research On Fault Diagnosis And Prediction Of Hydroelectric Generating Set Based On SVM Intelligent Mixing Method

Modern equipment gradually toward large-scale and complication, integration and automation of direction.This improves the working efficiency, but the safe and efficiently operation of the equipment put forward higher requirements.Once a fault occurs, the loss will be very large, and even cause some loss irreparable.Therefore, the equipment fault diagnosis and condition prediction is an important measure to ensure the safe and reliable operation of the equipment.In this paper, based on support vector machine (SVM) Intelligent Hybrid Method for hydro power generating units troubleshooting, historical data of hydro power units is premised on the future status of operation of hydro power units forecast.The main work is as follows:1. In view of the empirical mode decomposition (EMD) method is widely used in rotating machinery problems of (aliasing mode), We first to differential and integral the EMD, namely DEMD method.The method can change the proportion of different frequency components in the signal and the frequency of these similar or weak high-frequency components extracted.Then using support vector regression (SVR) and window function to DEMD of end effect problem in the decomposition process of optimization, simulation mode signal found this aliasing and edge effect problems have a better effect.Will get optimization method is applied to fault diagnosis of hydro power generating units, verify the effectiveness of the proposed method.2. Sample signal incomplete is contribute to the diagnosis of fault, and its contribution is not entirely the same.Aiming at this problem, we use fuzzy SVM is applied in the fault diagnosis of hydroelectric units, and puts forward a kind of optimization method based on class mean distance to determine the membership function.And then will be more classes fuzzy SVM class promotion and carry on the numerical simulation and application on hydro power, has achieved the better diagnostic results.3. Many fault signal of hydro power unit has a certain feature in the vibration signal,Based on the vibration can be seen as a sequence of time, can use regression neural network to the running state of hydro power unit according to the historical data to make predictions.Due to the distribution parameters of the regression neural network has great influence to the prediction accuracy, fruit flies optimization algorithm is adopted to carry out optimization.Finally use the optimization method can be applied to hydro power unit on the state prediction and compared with the BP neural network is analyzed, based on the analysis comparison optimization method has better prediction rate.