摘要
为了实现对电力系统中所有发电机组暂态稳定性的监视,提出一种可在线识别发电机组实时运行状态的改进卷积神经网络框架。提出的框架使用向量测量单元所获取的测量值作为数据源,利用带有预测池化层的卷积神经网络来实现多标签分类功能,以达到识别多种发电机组运行状态的目的。在模拟多种故障状态的IEEE 118测试系统上开展仿真测试。结果表明,所提出的框架能够快速而准确地识别发电机组的运行状态,是实现发电机组运行状态在线监视的可行方法。
In order to monitor the transient stability of all generating units in power system,an improved convolutional neural network framework is proposed which can identify the real time operation status of generating sets online.The proposed framework uses the measured values obtained by the phasor measurement unit as the data source,and uses the convolution neural network with prediction pooling layer to realize the multi label classification function,so as to achieve the purpose of identifying a variety of generator operating states.The simulation test is carried out on IEEE 118 test system simulating various fault states.The results show that the proposed framework can quickly and accurately identify the operating state of the generator set,which is a feasible method to realize the on-line monitoring of the operating state of the generator set.
作者
潘远
陈章国
蔡新雷
杨民京
Pan Yuan;Chen Zhangguo;Cai Xinlei;Yang Minjing(Power Dispatching Control Center of Guangdong Power Grid Co.,Ltd.,Guangzhou 510600,China;NARI Information & Communication Technology Co.,Ltd.,Nanjing 210003,China)
出处
《能源与环保》
2021年第6期186-191,196,共7页
CHINA ENERGY AND ENVIRONMENTAL PROTECTION
关键词
卷积神经网络
预测池化层
相量测量单元
深度学习
暂态稳定性
convolutional neural network
predictive pooling layer
phasor measurement unit
deep learning
transient stability