Improvement Strategy for Abnormal Error of Data-driven Power Flow Calculation Based on Deep Neural Network

Jianglong LEI; Juan YU; Mingxu XIANG; Zhifang YANG; Yan YANG; Wenyuan LI

doi:10.7500/AEPS20210516003

您当前的位置：

首页 >

文章列表页 >

Improvement Strategy for Abnormal Error of Data-driven Power Flow Calculation Based on Deep Neural Network

Special Topic on Data-Driven Methods for Modernizing Power Systems | 更新时间：2025-06-16

- Improvement Strategy for Abnormal Error of Data-driven Power Flow Calculation Based on Deep Neural Network
- Automation of Electric Power Systems Vol. 46, Issue 1, Pages: 76-84(2022)
- 作者机构：
  
  输配电装备及系统安全与新技术国家重点实验室（重庆大学），重庆市 400044
- 作者简介：
- 基金信息：
  
  National Natural Science Foundation of China(52077016);Natural Science Foundation of Chongqing, China(cstc2020jcyj-msxmX0315);Fundamental Research Funds for the Central Universities(2020CDJ-LHZZ-079)
- DOI：10.7500/AEPS20210516003
  CLC：
- Received：16 May 2021，
  
  Revised：2021-09-03，
  
  Published Online：19 November 2021，
  
  Published：10 January 2022
- 稿件说明：
移动端阅览
雷江龙,余娟,向明旭等.基于深度神经网络的数据驱动潮流计算异常误差改进策略[J].电力系统自动化,2022,46(1):76-84.

LEI Jianglong,YU Juan,XIANG Mingxu,et al.Improvement Strategy for Abnormal Error of Data-driven Power Flow Calculation Based on Deep Neural Network[J].Automation of Electric Power Systems,2022,46(1):76-84.
雷江龙,余娟,向明旭等.基于深度神经网络的数据驱动潮流计算异常误差改进策略[J].电力系统自动化,2022,46(1):76-84. DOI： 10.7500/AEPS20210516003.

LEI Jianglong,YU Juan,XIANG Mingxu,et al.Improvement Strategy for Abnormal Error of Data-driven Power Flow Calculation Based on Deep Neural Network[J].Automation of Electric Power Systems,2022,46(1):76-84. DOI： 10.7500/AEPS20210516003.

摘要

在考虑不确定性的

http://html.publish.founderss.cn/rc-pub/api/common/picture?pictureId=23510873&type=

http://html.publish.founderss.cn/rc-pub/api/common/picture?pictureId=23510869&type=

9.56733322

2.87866688

安全校核、可靠性计算等需大规模重复潮流计算的场景中，基于深度神经网络（DNN）的数据驱动方法存在部分潮流变量误差异常的问题，影响潮流越限判别的准确率。对此，首先通过理论推导，分析DNN参数更新过程及数据标准化原理，发现该问题的重要成因之一为：DNN仅根据标准化误差迭代训练模型，未计及潮流变量的真实学习误差及工程实际的精度要求，无法及时针对误差异常的潮流变量调整DNN参数。然后，面向潮流计算提出基于动态学习权重的DNN自适应训练方法。该方法通过每轮迭代中验证集的真实学习误差、越限误判率及误差统计指标，确定各潮流变量的学习权重，有效降低数据驱动潮流计算的异常误差。最后，在IEEE标准算例和Polish 2383节点系统上仿真验证了所提方法的有效性。

Abstract

In the scenarios requiring large-scale repetitive power flow calculations

such as

http://html.publish.founderss.cn/rc-pub/api/common/picture?pictureId=23510881&type=

http://html.publish.founderss.cn/rc-pub/api/common/picture?pictureId=23510879&type=

2.53999996

2.37066650

1 safety checks and reliability calculations considering uncertainty

the data-driven method based on deep neural network (DNN) has the abnormal error problem of some power flow variables

which affects the accuracy of the over-limit judgment of the power flow. To this end

through the theoretical derivation

the update process of DNN parameters and the principle of data standardization are analyzed. It is found that one of the important causes of this problem is that the DNN only trains the model iteratively based on standardized errors

and does not take into account the true learning error of the power flow variables and the actual accuracy requirements of the project. And it is impossible to adjust the DNN parameters in time for the power flow variables with abnormal errors. Then

a DNN adaptive training method based on dynamic learning weights is proposed for power flow calculation. This method determines the learning weight of each power flow variable through the true learning error

over-limit misjudgment rate and error statistical indicators of the verification set in each iteration

which effectively reduces the abnormal error of the data-driven power flow calculation. Finally

simulations on IEEE standard examples and Polish 2383-node system verify the effectiveness of the proposed method.

关键词

Keywords

references

杨耿杰，郭谋发 . 电力系统分析［M］. 2版 . 北京：中国电力出版社， 2013 .

YANG Gengjie ， GUO Moufa . Power system analysis ［M］. 2nd ed . Beijing ： China Electric Power Press ， 2013 .

汪洋，夏清，康重庆 . 考虑电网 <math id="M95"><mi>N</mi><mo>-</mo><mn mathvariant="normal">1</mn></math> http://html.publish.founderss.cn/rc-pub/api/common/picture?pictureId=23511776&type= http://html.publish.founderss.cn/rc-pub/api/common/picture?pictureId=23511774&type= 7.28133297 2.20133328 闭环安全校核的最优安全发电计划［J］. 中国电机工程学报， 2011 ， 31 （ 10 ）： 39 - 45 .

WANG Yang ， XIA Qing ， KANG Chongqing . Optimal security constrained generation scheduling considering closed-loop <math id="M96"><mi>N</mi><mo>-</mo><mn mathvariant="normal">1</mn></math> http://html.publish.founderss.cn/rc-pub/api/common/picture?pictureId=23511781&type= http://html.publish.founderss.cn/rc-pub/api/common/picture?pictureId=23511779&type= 7.28133297 2.20133328 security correction ［J］. Proceedings of the CSEE ， 2011 ， 31 （ 10 ）： 39 - 45 .

苏晨博，刘崇茹，李至峪，等 . 基于贝叶斯理论的考虑多维风速之间相关性的概率潮流计算［J］. 电力系统自动化， 2021 ， 45 （ 3 ）： 157 - 165 .

SU Chenbo ， LIU Chongru ， LI Zhiyu ， et al . Bayesian theory based calculation of probabilistic power flow considering correlation between multi-dimensional wind speed ［J］. Automation of Electric Power Systems ， 2021 ， 45 （ 3 ）： 157 - 165 .

孙树明，谢昶，吕颖，等 . 电力系统在线安全稳定分析应用模式［J］. 电网技术， 2015 ， 39 （ 10 ）： 2875 - 2881 .

SUN Shuming ， XIE Chang ， LÜ Ying ， et al . Power system online security and stability analysis application modes ［J］. Power System Technology ， 2015 ， 39 （ 10 ）： 2875 - 2881 .

郭健，周京阳，李强，等 . 高性能在线分析计算现状与协同计算关键技术［J］. 电力系统自动化， 2018 ， 42 （ 3 ）： 149 - 159 .

GUO Jian ， ZHOU Jingyang ， LI Qiang ， et al . Current status of high-performance on-line analysis computation and key technologies for cooperating computation ［J］. Automation of Electric Power Systems ， 2018 ， 42 （ 3 ）： 149 - 159 .

唐灿，董树锋，任雪桂，等 . 用于迭代法潮流计算的改进Jacobi预处理方法［J］. 电力系统自动化， 2018 ， 42 （ 12 ）： 81 - 86 .

TANG Can ， DONG Shufeng ， REN Xuegui ， et al . Improved Jacobi pre-treatment method for solving iterative power flow calculation ［J］. Automation of Electric Power Systems ， 2018 ， 42 （ 12 ）： 81 - 86 .

STOTT B ， JARDIM J ， ALSAC O . DC power flow revisited ［J］. IEEE Transactions on Power Systems ， 2009 ， 24 （ 3 ）： 1290 - 1300 .

卫志农，张清松，赵静波，等 . 电力系统线性化模型研究综述与改进［J］. 电网技术， 2017 ， 41 （ 9 ）： 2919 - 2927 .

WEI Zhinong ， ZHANG Qingsong ， ZHAO Jingbo ， et al . Review and improvement of power system linearization models ［J］. Power System Technology ， 2017 ， 41 （ 9 ）： 2919 - 2927 .

樊哲新，杨知方，冯楠，等 . 潮流方程线性化误差分析及最优状态空间选取［J］. 电力系统自动化， 2020 ， 44 （ 1 ）： 109 - 117 .

FAN Zhexin ， YANG Zhifang ， FENG Nan ， et al . Error analysis and optimal state space selection of linear power flow equation ［J］. Automation of Electric Power Systems ， 2020 ， 44 （ 1 ）： 109 - 117 .

IMEN L ， DJAMEL L . Power flow variation based on extreme learning machine algorithm in power system ［J］. International Journal of Power Electronics and Drive Systems （IJPEDS）， 2019 ， 10 （ 3 ）： 1244 .

MA L X ， ZHANG X J . Study on high capability intelligent algorithm in load flow of power system ［C］// 2008 International Conference on Neural Networks and Signal Processing ， June 7-11 ， 2008 ， Nanjing， China .

BAGHAEE H R ， MIRSALIM M ， B GHAREHPETIAN G . Power calculation using RBF neural networks to improve power sharing of hierarchical control scheme in multi-DER microgrids ［J］. IEEE Journal of Emerging and Selected Topics in Power Electronics ， 2016 ， 4 （ 4 ）： 1217 - 1225 .

YANG Y ， YANG Z F ， YU J ， et al . Fast calculation of probabilistic power flow： a model-based deep learning approach ［J］. IEEE Transactions on Smart Grid ， 2020 ， 11 （ 3 ）： 2235 - 2244 .

WANG L ， ZHOU Q ， JIN S S . Physics-guided deep learning for power system state estimation ［J］. Journal of Modern Power Systems and Clean Energy ， 2020 ， 8 （ 4 ）： 607 - 615 .

XIANG M X ， YU J ， YANG Z F ， et al . Probabilistic power flow with topology changes based on deep neural network ［J］. International Journal of Electrical Power & Energy Systems ， 2020 ， 117 ： 105650 .

PAN S J ， YANG Q . A survey on transfer learning ［J］. IEEE Transactions on Knowledge and Data Engineering ， 2010 ， 22 （ 10 ）： 1345 - 1359 .

YANN L ， LEON B ， GENEVIEVE B O ， et al . Neural networks： tricks of the trade ［M］. Berlin， Germany ： Heidelberg ： Springer ， 1998 .

AL-FAIZ M Z ， IBRAHIM A A ， HADI S M . The effect of z -score standardization （normalization） on binary input due the speed of learning in back-propagation neural network ［J］. Iraqi Journal of Information & Communications Technology ， 2019 ， 1 （ 3 ）： 42 - 48 .

RUDER S . An overview of gradient descent optimization algorithms ［EB/OL］.［ 2021-05-10 ］. http：//sebastianruder.com/optimizing-gradient-descent http://sebastianruder.com/optimizing-gradient-descent .

RASKUTTI G ， WAINWRIGHT M J ， YU B . Early stopping for non-parametric regression： an optimal data-dependent stopping rule ［C］// 2011 49th Annual Allerton Conference on Communication， Control， and Computing （Allerton）， September 28-30 ， 2011 ， Monticello， USA .

LI Y M ， LI W Y ， YAN W ， et al . Probabilistic optimal power flow considering correlations of wind speeds following different distributions ［J］. IEEE Transactions on Power Systems ， 2014 ， 29 （ 4 ）： 1847 - 1854 .

SARKAR S ， AJJARAPU V . MW resource assessment model for a hybrid energy conversion system with wind and solar resources ［J］. IEEE Transactions on Sustainable Energy ， 2011 ， 2 （ 4 ）： 383 - 391 .

SNOEK J ， LAROCHELLE H ， ADAMS R P . Practical Bayesian optimization of machine learning algorithms ［EB/OL］. ［ 2021-05-01 ］. https：//arxiv.org/pdf/1206.2944.pdf https://arxiv.org/pdf/1206.2944.pdf .

DU Y ， LI F X ， LI J ， et al . Achieving 100x acceleration for <math id="M97"><mi>N</mi><mo>-</mo><mn mathvariant="normal">1</mn></math> http://html.publish.founderss.cn/rc-pub/api/common/picture?pictureId=23511787&type= http://html.publish.founderss.cn/rc-pub/api/common/picture?pictureId=23511784&type= 7.28133297 2.20133328 contingency screening with uncertain scenarios using deep convolutional neural network ［J］. IEEE Transactions on Power Systems ， 2019 ， 34 （ 4 ）： 3303 - 3305 .

李峰，李虎成，於益军，等 . 基于并行计算和数据复用的快速静态安全校核技术［J］. 电力系统自动化， 2013 ， 37 （ 14 ）： 75 - 80 .

LI Feng ， LI Hucheng ， YU Yijun ， et al . Fast computing technologies for static security checking based on parallel computation and data reuse ［J］. Automation of Electric Power Systems ， 2013 ， 37 （ 14 ）： 75 - 80 .

Views

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

Online Reactive Power Optimization Method Based on Data-driven Predictive Control for Distribution Network with Low Situation Awareness

Data-driven Oscillation Mode Prediction Scheme and Resonance Mitigation Analysis

Related Author

CHENG Lilin

LUO Zijie

LI Qun

ZHANG Ningyu

LI Yaran

ZANG Haixiang

HUANG Yu

ZHANG Xiaoxiao

Related Institution

School of Electrical and Power Engineering, Hohai University

Electric Power Research Institute of State Grid Jiangsu Electric Power Co., Ltd.

Institute of Advanced Technology for Carbon Neutrality, Nanjing University of Posts and Telecommunications

NARI Group Corporation (State Grid Electric Power Research Institute)

School of Electrical Engineering, Zhejiang University

⁰