一种基于ANN技术的预想事故自动选择方法

A NEW CONTINGENCY RANKING APPROACH USING ANN TECHNIQUES

本文提出了一种基于人工神经元网络（ＡＮＮ）技术的预想事故自动选择方法，利用快速解耦潮流计算的迭代一次法（ＦＤＬＦ１）构造了分别对应于支路有功潮流越限的有功功率性能指标ＰＩｐ 和对应于节点电压模值和发电机无功出力越限的电压－ 无功性能指标ＰＩＶＱ。用ＢＰ算法训练多层感知型神经元网络以求得对应于不同运行状态和不同网络拓扑结构的性能指标。通过算例分析，并与其它经典方法的比较，本方法在计算精度和速度方面非常令人满意，而且具有较高的事故捕获率。

This paper presents an approach for fast contingency screening based on ANN techniques.Firstly,two kinds of performance indices PI p and PI VQ are formed corresponding to line active power violations,bus voltage magnitude and bus reactive power injection violations,respectively,according to the first iteration results of Fast Decoupled Load Flow calculation.The multi-layered ANN is trained by the error back propagation algorithm to get the performance indices within a large range of operating conditions and changes in network topology.Comparing with other classical contingency screening method, the proposed approach can have high calculating accuracy and speed, and ensure high contingency capturing rate at the same time.