基于特征量和卷积神经网络的暂态电压稳定评估

Transient voltage stability evaluation based on feature and convolutional neural network

针对目前传统方法无法准确快速判断系统暂态电压稳定性的问题,提出一种基于特征量和卷积神经网络(convolutional neural networks,CNN)的暂态电压稳定评估方法.首先基于传统电压稳定指标分解及响应数据,从原始电气量中进行36维电压稳定构建特征量,将特征量输入卷积神经网络中进行有监督学习,最后将训练所得到的模型应用于电力系统暂态稳定评估中.利用新英格兰10机39节点标准算例对电网进行了仿真,结果表明该方法具有错误率低、精确率高、测试时间短的特点,能够准确快速判断出系统的暂态电压失稳,提高了电力系统的稳定性.

Aiming at the problem that the traditional method can not accurately and quickly judge system transient voltage stability,this paper proposes a transient voltage stability evaluation method based on feature and convolutional neural network(CNN).Firstly,based on the traditional voltage stability index decomposition and response data,the 36-dimensional voltage stability feature is constructed from the original electrical quantity;and the feature is inputed into the convolutional neural network for supervised learning.Finally, the training model is applied to the power system transient stability assessment.The simulation study in the New England 10-machine 39-node standard example grid shows that this method has a low error rate,high precision and short test time.It can accurately and quickly determine the transient voltage instability of the system and improve the stability of the power system.