摘要
随着风电渗透率的提高,电网对风电场的并网适应性特别是对其孤岛检测的要求越来越高。为了提高风电场并网点状态检测的准确性,首先对并网点的电压信号进行S变换形成复时频矩阵,提出了基于S变换的能量熵、奇异熵及相位变化等特征集合。然后利用深度学习策略中的LSTM模型对更加高级抽象的特征进行自动提取,从而实现对孤岛状态的精准检测。为了对检测效果进行校核,以某地风电场为例,进行现场数据的校验。测试结果表明,策略的检测效果比较满意。
characteristics,so as to realize accurate detection of the island status.In order to check detection effects,the wind farm at a certain place was taken as an example to verify field data.The test results showed that the detection effect of the proposed strategy was quite satisfactory.
作者
朱凌
徐大勇
Zhu Ling;Xu Dayong(Huizhou Power Supply Bureau,Huizhou Guangdong 516003,China)
出处
《电气自动化》
2020年第3期20-23,共4页
Electrical Automation
基金
广东电网有限责任公司科技项目(031300KK52160028)。
关键词
风电场
孤岛检测
深度学习
长短时记忆神经网络
S变换
混淆矩阵
wind farm
island detection
deep learning
long and short time memory(LSTM)neural network
S transformation
confusion matrix