基于变点探测的功率振荡数据挖掘

Data mining of power oscillation based on change-point detection

针对当前功率小幅振荡数据挖掘的不足,引入了变点探测方法判断系统是否发生振荡、主要参与机组以及振荡何时进入平稳阶段,从而提出了一种新的大电网功率振荡特征挖掘方法。该方法通过在海量广域测量系统(WAMS)数据中挖掘电网振荡信息,根据变点探测方法获取的极值特性区分弱阻尼的低频振荡以及强阻尼快速衰减过程,并在弱阻尼振荡情况下确定Prony分析时间窗的起点,从而获取更为准确的振荡模式和强相关机组信息。通过新英格兰10机39节点系统仿真和河南电网WAMS实测振荡数据挖掘验证了所提方法的有效性,结果表明该方法能够从海量数据中有效挖掘大电网振荡特征,并准确识别系统模式信息。

Aiming at the defects of data mining for slight power oscillation features, the change-point detection method is introduced to determine whether the low-frequency oscillation happens or not ,which units are the main par- ticipate units and when the oscillation turns into stable phase. On this basis, a new data mining method of power os- cillation features in large grid is proposed. This method identifies the oscillation information of power grid from mass data of WAMS （ Wide Area Measurement System） , distinguishes the low frequency oscillation with weak damping and dynamic process with faster decay according to the extreme point features obtained by the change-point detection method, identifies the beginning of time window of Prony method in the weak damping oscillation condition, and then obtains more reliable mode information and strong correlation generators. The effectiveness of the proposed method is verified by the simulation of New England 10-generator 39-bus system and the data mining on WAMS practical measured data in Henan Power Grid. The results show that the oscillation features and mode information of power system can be accurately identified from the mass data.