基于三层BP神经网络的低压配电网安全性分析方法

林文硕; 刘奇; 田慧丽; 朱怡莹; 许达

林文硕,刘奇,田慧丽,朱怡莹,许达.基于三层BP神经网络的低压配电网安全性分析方法[J].电测与仪表,2026,63(6):121-129.
Lin Wenshuo,Liu Qi,Tian Huili,Zhu Yiying,Xu Da.Security analysis method for low-voltage distribution network based on three-layer BP neural network[J].Electrical Measurement & Instrumentation,2026,63(6):121-129.

基于三层BP神经网络的低压配电网安全性分析方法

Security analysis method for low-voltage distribution network based on three-layer BP neural network

DOI：10.19753/j.issn1001-1390.2026.06.013

中文关键词: 三层BP神经网络低压配电网安全性分析电能质量谐波畸变容量约束

英文关键词:three-layer BP neural network, low voltage distribution network, security analysis, power quality, harmonic distortion, capacity constraints

基金项目:南方电网管理创新项目（编号GDKJXM20220914）

作者中文名	作者英文名	单位
林文硕	Lin Wenshuo	广东电网有限责任公司广州供电局
刘奇	Liu Qi	广东电网有限责任公司广州供电局
田慧丽	Tian Huili	广东电网有限责任公司广州供电局
朱怡莹	Zhu Yiying	广东电网有限责任公司广州供电局
许达	Xu Da	广东电网有限责任公司广州供电局

摘要点击次数: 138

中文摘要:

低压配电网的负荷和运行状态具有显著的动态变化特性，包括负荷波动、设备老化及故障，这些变化导致安全性评估难度增加。为有效应对这一挑战，提出了一种基于三层BP神经网络算法的低压配电网安全性分析方法。所提方法以电能质量、电流电压谐波畸变和馈线容量作为安全性评估指标，将归一化处理的实时运行状态数据输入三层BP(back propagation)神经网络中，基于安全性评估指标修正网络各层级权重值，以最小化模拟训练误差。在训练完成后，该网络能够准确评估处于不同负荷状态下的电网关键变量，实现对配电网安全性的分析。实验结果表明，所提方法能够有效捕捉输入数据中的关键特征，其准确率达到了97.8%，均方误差为0.008，验证了该方法的高效性和准确性。并且该方法能够准确评估不同负荷状态下的电网关键变量，对细微变化具有敏锐的感知能力，有效提升了电网的稳定性和安全性。

英文摘要:

The load and operating status of low-voltage distribution network have significant dynamic variation features, including load fluctuations, equipment aging, and faults, which increase the difficulty of safety assessment. To effectively address this challenge, a low-voltage distribution network security analysis method based on three-layer BP(back propagation) neural network algorithm is proposed. This method uses power quality, current voltage harmonic distortion, and feeder capacity as safety evaluation indicators. The normalized real-time operating status data is input into a three-layer BP neural network, and the weight values of each level of the network are corrected based on the safety evaluation indicators to minimize simulation training errors. After training, the network can accurately evaluate the key variables of the power grid under different load states, achieving analysis of the safety of the distribution network. The experimental results show that the proposed method can effectively capture key features in the input data, with an accuracy of 97.8% and a mean square error of 0.008, verifying the efficiency and accuracy of the proposed method. And this method can accurately evaluate the key variables of the power grid under different load states, with a keen perception ability for subtle changes, effectively improving the stability and safety of the power grid.

查看全文查看/发表评论下载PDF阅读器

关闭