一种变结构的时间序列预测算法及其在电力系统暂态稳定预测中的应用

针对电力系统机电暂态过程存在首摆失稳和多摆失稳的特点 ,提出了一种变结构的时间序列预测算法 ,即针对不同的失稳模式采用不同结构的预测算法。利用多项式与智能动态修正相结合实时快速预测发电机的功角 ,若预测功角的多项式不存在极值点 ,即为首摆失稳 ,可终止预测过程 ;多摆失稳是利用同步多参量测量装置实测的数据和多项式预测功角的结果 ,提出的一种基于相邻实际输出之间差别驱动的TD算法与具有遗忘因子的改进BP算法相结合的混合算法 ,并用该算法对电力系统机电暂态过程的不平衡功率和功角的时间序列进行了多步预测。预测结果表明 ,该算法是有效和可行的 ,达到了满意的精度 ,克服了采用多项式或自回归方法预测时存在较大误差的弊病

A Variable Structure Predictive Algorithm For Times Series and Its Application to Real-time Transient Stability Prediction

A variable structure predictive algorithm is proposed based on first swing instability and multi swing instability during the transient stability of power system Variable structure predictive method includes two parts First phasor change is obtained by polynomial approaching which utilizes synchronization multi parameter measurement At the same time,phasor is modified by intelligent control If polynomial have not extremum,it is unstable corresponding to first swing stability Then this paper emphasizes research multi swing instabilty It presented a hybrid learning algorithm based on neural network model to predict generator angles and generator electrical power output with several steps in advances The proposed method combines the temporal differences methods with BP algorithm with forgetting It helps to solve the computing problem incrementally in traditional BP algorithm on multi step predicting and has the ability of structural learning The proposed method is feasible by application to real time transient stability prediction of power systems Numerical simulation results have shown that the proposed method is effective and practically feasible