Fault prediction for power system based on multidimensional time series correlation analysis

Fault prediction for power equipment allows the maintenance personnel to know the operation conditions and the fault to be occurred in advance so as to reduce the risk of fault and the economic loss. The current fault prediction methods are generally based on physical model or stochastic model, which are used to evaluate the remaining life of equipment. However, in fact, there are many interference factors between grid equipment; therefore, it is very difficult to illustrate the fault characteristics by using an accurate mathematical model. Besides, there exist many problems such as uncertain mathematical model parameters and short prediction period. As such, this paper proposes a fault prediction method based on correlation analysis of multidimensional time series, which normalizes the historical time-series data of nodes in the power equipment network topology, decomposes the time series by using the time series decomposition algorithm, extracts the typical events occurred on the nodes before the key equipment fails by using one time series pattern representation method, and explores the implicit relationship between the indicator trend and the operating conditions of equipment by correlation method, with a purpose of effectively predicting the fault or impact. The test shows that this method can make the best of time series data and take advantage of the capability to analyze and express the uncertainty relation of data mining so as to accurately and effectively predict equipment faults.