关键词: 故障识别, 计量一体化, 自组织神经网络, 最小二乘, 支持向量机, 故障指数

Abstract: Based on metering automation and integration platform, the online failure-identification method for power equipment of some special variable users is designed in order to solve the problem of quick failure-identification in the value-added service mode provided for those users. In the beginning, the time-series auto regression model for the device state parameters was set up, then the time-series including the device state values was quantized as the inputting ones of the system by self-organizing maps. The learning samples of the least square support vector machines model were built by using process input values in the sliding time window. The differences between the calculating regression results and measuring values of the feature vector were set as the observed ones. The background model of the system configured Background model of multidimensional observation value distribution system was fit by the Gaussian mixture model, in which the failure index was calculated by the matching degree between the individual observation and the background model to achieve the real-time identification of the equipment failure. The experimental results show that this method is able to predict failures online quickly and accurately.

Key words: Online failure identification, metering automation and integration platform, self-organizing maps, least square, support vector machines, failure index