基于广域多时间尺度的CVT误差状态预测方法

Error State Prediction Method of CVT Based on Wide-area Multi-timescale

针对现有电容式电压互感器(CVT)误差状态预测只考虑单一粒度构建时序模型的问题,本文提出了一种基于广域多时间尺度的CVT误差状态预测方法,首先,对CVT运行状态下二次输出数据进行STL时序分解,构建多元时序特征;其次,利用不同时间尺度的分解数据构造多元时间序列预测模型;最后,利用多元时间序列预测模型,实现不停电条件下CVT误差状态的精准预测。仿真数据表明,该方法能准确预测CVT超误差状态,与现有方法相比,验证了该方法的准确性与适用性。

Addressing the issue of existing single-granularity sequential models in predicting error states of Capacitive Voltage Transformers(CVT),this paper proposes the error state prediction method of CVT based on wide-area multi-timescale.Firstly,the second-order output data of CVT during operation is subjected to STL time series decomposition.Secondly,multi-dimensional time series prediction models are constructed using the decomposed data from different timescales.Finally,accurate prediction of CVT error states under uninterrupted conditions is achieved using the multi-dimensional time series prediction models.Simulation data shows that this method can accurately predict the over error state of CVT,and compared with existing methods,the accuracy and applicability of this method have been verified.