摘要
针对架空线-电缆混合配电线路故障选线更为复杂,而且容易受到过渡电阻和故障位置的影响的问题,提出了一种排列熵和优化后的BP神经网络相结合的混合线路故障检测新方法。排列熵对于识别故障后的信号复杂程度有很好的效果,并将故障发生后一个周期内的三相电压信号的排列熵作为特征向量,输入到经过模拟退火算法优化后的BP神经网络中,并仿真验证了该方法的有效性。结果表明该方法不受故障位置和过渡电阻的影响,而且可以有效识别高阻接地故障,并与现有常见的故障诊断方法进行比较,比没有优化的BP神经网络准确率高出14%,比决策树和随机森林算法准确率高出更多;与以能量分量系数作为特征向量相比,以排列熵作为输入向量,准确率要高出4%。
Aiming at the problem that the fault line selection of overhead line cable hybrid distribution lines is more complex,and is easily affected by transition resistance and fault location,a new hybrid line fault detection method based on permutation entropy and optimized BP neural network is proposed. The permutation entropy has a good effect on identifying the complexity of the signal after the fault. The permutation entropy of the three-phase voltage signal in one cycle after the fault occurs is input into the BP neural network optimized by the simulated annealing algorithm as the feature vector,and the simulation verifies the effectiveness of this method. The results show that this method is not affected by fault location and transition resistance,and can effectively identify high resistance grounding fault. Compared with the existing common fault diagnosis methods,the accuracy of this method is 14%higher than that of BP neural network without optimization,and more than that of decision tree and random forest algorithm. Compared with using the energy component coefficient as the feature vector,the permutation entropy is used as the input vector,the accuracy is 4% higher.
作者
齐沛锋
韩笑
丁煜飞
夏寅宇
汪缪凡
QI Peifeng;HAN Xiao;DING Yufei;XIA Yinyu;WANG Miufan(Nanjing Institute of Engineering,Nanjing 211167)
出处
《计算机与数字工程》
2022年第10期2307-2311,共5页
Computer & Digital Engineering
基金
国家自然科学基金项目“基于换流器与电网控制保护协同的直流配电网保护及故障恢复策略研究”(编号:52107098)资助。
关键词
模拟退火算法
排列熵
BP神经网络
混合配电网
simulated annealing algorithm
permutation entropy
BP neural network
hybrid distribution network
作者简介
齐沛锋,男,硕士研究生,研究方向:电力系统继电保护;韩笑,男,教授,研究方向:电力系统继电保护;丁煜飞,男,硕士研究生,研究方向:电力系统继电保护;夏寅宇,男,硕士研究生,研究方向:电力系统继电保护;汪缪凡,男,硕士研究生,研究方向:电力系统继电保护。
