双阶段配电网线损率计算方法

Line Loss Rate Calculation Method of Two-Stage Distribution Network

摘要

针对低压配电网线损特征识别与理论计算中存在的特征权重分配主观性强、非线性映射精度不足等问题, 提出了一种融合多特征加权聚类与智能优化神经网络的线损分析方法。首先建立了多特征加权改进型k均值聚类算法(MFW-IKCA: Multi-Feature Weighted Improved k-Means Clustering Algorithm), 通过熵权法与互信息法的融合策略实现特征维度的自适应加权, 并用交替方向乘子法(ADMM: Alternating Direction Method of Multipliers)优化目标函数; 其次采用改进遗传算法优化的莱文贝格-马夸特反向传播神经网络(IGA-LMBP-NNM: Improved Genetic Algorithm Optimized Levenberg-Marquardt Back Propagation Neural Network)参数, 通过模拟二进制交叉、多项式变异及自适应阻尼因子提升网络收敛速度与预测精度; 最后利用决策树分类器提取聚类规则, 并结合核密度估计实现异常值检测。实验结果表明, 与主流算法相比, 所提方法聚类纯度提升14.5%, 均方误差(MSE: Mean Squared Error)降低29.3%, 决定系数R²提高至0.892。该研究不仅为配电网线损精细化分析提供了新思路, 也为提升台区降损决策的科学性提供了有力支持, 具有显著的工程应用价值。

Abstract

A novel method integrating multi-feature weighted clustering and intelligent optimized neural network is proposed to address the issues of subjective feature weighting and insufficient nonlinear mapping accuracy in low-voltage distribution network line loss analysis. First, an improved k-means clustering algorithm (MFW-IKCA: Multi-Feature Weighted Improved k-Means Clustering Algorithm) is established by introducing dynamic weight matrices and reference weights, combining entropy weighting and mutual information methods for adaptive feature weighting, and optimizing the objective function via the ADMM (Alternating Direction Method of Multipliers). Subsequently, an IGA(Improved Genetic Algorithm) is employed to optimize the parameters of the LMBP-NNM (Levenberg-Marquardt Backpropagation Neural Network), enhancing convergence speed and prediction accuracy through simulated binary crossover, polynomial mutation, and adaptive damping factors. Finally, decision tree classifiers are utilized to extract clustering rules, while kernel density estimation enabled anomaly detection. Experimental results on 710 transformer district datasets have demonstrated a 14.5% improvement in clustering purity, a 29.3% reduction in MSE (Mean Squared Error), and a 0.892 coefficient of determination (R ²). This study provides a new perspective for refined line loss analysis and offers significant engineering value for improving scientific decision-making in loss reduction of distribution network.

关键词

台区线损特征识别 / 多特征加权聚类 / 改进遗传算法 / 莱文贝格-马夸特神经网络 / 动态权重分配 / 配电网理论线损计算

Key words

line loss feature recognition / multi-feature weighted clustering / improved genetic algorithm / levenberg-marquardt neural network / dynamic weight allocation / distribution network line loss calculation

参考文献

[1] 崔婷婷, 钟庆, 梁怡雪, 等. 配电网串/并联型动态电压恢复器组合优化配置方法[J]. 电力电容器与无功补偿, 2025, 46(2): 135-142.

[2] CUI T T, ZHONG Q, LIANG Y X, et al. Collaborative Optimal Configuration Method for Series/Parallel Dynamic Voltage Restorers in Distribution Network[J]. Power Capacitors & Reactive Power Compensation, 2025, 46(2): 135-142.

[3] 王宣元, 张玮, 李长宇, 等. 考虑随机性的主动配电网有功无功可行域计算方法[J]. 中国电力, 2025, 58(4): 182-192.

[4] WANG X Y, ZHANG W, LI C Y, et al. The Large-Scale Integration of Renewable Energy Poses Challenges to the Power Systems, Such as the Shortage of Flexibility[J]. Electric Power, 2025, 58(4): 182-192.

[5] 叶臻, 叶鹏, 程绪可, 等. 供电网络线损计算研究综述[J]. 沈阳工程学院学报(自然科学版), 2022, 18(2): 58-64.

[6] YE Z, YE P, CHENG X K, et al. Review of Line Loss Calculation of Supply Network[J]. Journal of Shenyang Institute of Engineering (Natural Science), 2022, 18(2): 58-64.

[7] 贺远, 王耀晨, 李军. 同期线损系统在电网消缺中的智能应用[J]. 电世界, 2023, 64(1): 1-5.

[8] HE Y, WANG Y C, LI J. Intelligent Application of Synchronous Line Loss System in Grid Defect Elimination[J]. Electric World, 2023, 64(1): 1-5.

[9] 佘健铭. 基于大数据配网线路降损系统的研究与设计[J]. 电力设备管理, 2025(3): 133-135.

[10] SHE J M. Research and Design of a Big Data-Based Loss Reduction System for Distribution Lines[J]. Electric Power Equipment Management, 2025(3): 133-135.

[11] 张振刚, 杨勇, 闫高飞. 基于聚类分析的台区阻抗特性及线损模型构建[J]. 电子设计工程, 2025, 33(7): 36-41.

[12] ZHANG Z G, YANG Y, YAN G F. Construction of Line Transmission Impedance Characteristics and Line Loss Model of Substation Based on Cluster Analysis[J]. Electronic Design Engineering, 2025, 33(7): 36-41.

[13] 汪颖, 李有煜, 汪清, 等. 基于台区线损状态识别的通用线损率概率密度分布模型[J]. 供用电, 2025, 42(4): 80-90,98.

[14] WANG Y, LI Y Y, WANG Q, et al. General Probability Density Distribution Model for Line Loss Based on State Identification in Power Distribution Area[J]. Distribution & Utilization, 2025, 42(4): 80-90,98.

[15] 边舒芳, 张伟. 基于改进LSTM的低压配电网日线损率预测方法[J]. 粘接, 2025, 52(1): 188-192.

[16] BIAN S F, ZHANG W. A Daily Line Loss Prediction Method for Medium and Low Voltage Distribution Networks Based on Improved LSTM[J]. Adhesion, 2025, 52(1): 188-192.

[17] 原俊龙, 张军. 基于深度学习的台区线损异常检测系统研究[J]. 光源与照明, 2024(12): 71-73.

[18] YUAN J L, ZHANG J. Research on Anomaly Detection System for Transformer District Line Loss Based on Deep Learning[J]. Lamps & Lighting, 2024(12): 71-73.

[19] 刘全, 刘晓松, 吴光军, 等. 基于渐近式k-Means聚类的多行动者确定性策略梯度算法[J]. 吉林大学学报(理学版), 2025, 63(3): 885-894.

[20] LIU Q, LIU X S, WU G J, et al. Multi-Actor Deterministic Policy Gradient Algorithm Based on Progressive k-Means Clustering[J]. Journal of Jilin University (Science Edition), 2025, 63(3): 885-894.

[21] 胡雅婷, 陈营华, 宝音巴特, 等. 一种增量式MinMax k-Means聚类算法[J]. 吉林大学学报(理学版), 2021, 59(5): 1205-1211.

[22] HU Y T, CHEN Y H, BAOYIN B T, et al. An Incremental Minmax k-Means Clustering Algorithm[J]. Journal of Jilin University (Science Edition), 2021, 59(5): 1205-1211.

[23] 王海燕, 崔文超, 许佩迪, 等. 一种局部概率引导的优化K-means++算法[J]. 吉林大学学报(理学版), 2019, 57(6): 1431-1436.

[24] WANG H Y, CUI W C, XU P D, et al. An Optimized K-Means++ Algorithm Guided by Local Probability[J]. Journal of Jilin University (Science Edition), 2019, 57(6): 1431-1436.

[25] 冯子玹, 王彩玲, 冀书关. 修正的增广拉格朗日函数内点拟牛顿法[J]. 吉林大学学报(理学版), 2009, 47(2): 245-247.

[26] FENG Z X, WANG C L, JI S G. Interior-Point Quasi-Newton Method of Modified Augmented Lagrangian Function[J]. Journal of Jilin University (Science Edition), 2009, 47(2): 245-247.

[27] 赵伟卫, 李艳颖, 赵风芹, 等. 基于互信息和随机森林的混合变量选择算法[J]. 吉林大学学报(理学版), 2017, 55(4): 933-939.

[28] ZHAO W W, LI Y Y, ZHAO F Q, et al. Hybrid Variable Selection Algorithm Based on Mutual Information and Random Forest[J]. Journal of Jilin University (Science Edition), 2017, 55(4): 933-939.

[29] 王文霞. 数据挖掘中改进的C4.5决策树分类算法[J]. 吉林大学学报(理学版), 2017, 55(5): 1274-1277.

[30] WANG W X. Improved C4.5 Decision Tree Classification Algorithm in Data Mining[J]. Journal of Jilin University (Science Edition), 2017, 55(5): 1274-1277.

[31] 王建刚. 基于小生境遗传算法的网络入侵节点智能检测方法[J]. 吉林大学学报(理学版), 2025, 63(4): 1099-1104.

[32] WANG J G. Intelligent Detection Method for Network Intrusion Nodes Based on Niche Genetic Algorithm[J]. Journal of Jilin University (Science Edition), 2025, 63(4): 1099-1104.

[33] 景雯, 张杰, 傅文博, 等. 基于改进遗传算法的物联网链路负载均衡控制方法[J]. 吉林大学学报(理学版), 2023, 61(4): 922-928.

[34] JING W, ZHANG J, FU W B, et al. Load Balancing Control Method of IoT Link Based on Improved Genetic Algorithm[J]. Journal of Jilin University (Science Edition), 2023, 61(4): 922-928.

[35] 隋振, 张天星, 吴涛, 等. 基于多种群空间映射遗传算法的立体仓库储位优化[J]. 吉林大学学报(理学版), 2022, 60(1): 127-134.

[36] SUI Z, ZHANG T X, WU T, et al. Storage Optimization of Three-Dimensional Warehouse Based on Multi Population Space Mapping Genetic Algorithm[J]. Journal of Jilin University (Science Edition), 2022, 60(1): 127-134.

[37] 尹伟石, 刘晓奇, 徐轩. 热源参数识别问题的Bayes遗传算法[J]. 吉林大学学报(理学版), 2020, 58(4): 841-846.

[38] YIN W S, LIU X Q, XU X. Bayesian Genetic Algorithm for Heat Source Parameter Identification Problem[J]. Journal of Jilin University (Science Edition), 2020, 58(4): 841-846.

[39] 曹名圆, 李蓉, 闫雪丽, 等. 求解非线性方程组的非单调自适应加速Levenberg-Marquardt算法[J]. 吉林大学学报(理学版), 2024, 62(3): 538-546.

[40] CAO M Y, LI R, YAN X L, et al. Nonmonotonic Adaptive Accelerated Levenberg-Marquardt Algorithm for Solving Nonlinear Equations[J]. Journal of Jilin University (Science Edition), 2024, 62(3): 538-546.

基金资助

国家电网科技基金资助项目(1400-202312635a-3-2-zn)