基于图信号处理的船舶电力系统超分辨率量测生成方法

doi:10.12382/bgxb.2025.0643

摘要/Abstract

摘要：

船舶在运行过程中，易因设备老化、操作失误或其他突发事件影响造成电力系统通信系统损伤，导致电力系统关键数据缺失，影响系统的正常运行和安全态势的准确感知。针对现有电力系统缺失数据恢复算法速度较慢且还原精度较低的问题，提出一种基于少量量测信息的超分辨率量测生成方法。根据电力系统量测配置分布空间分布不均且量测精度不同的特点构建电力系统空间稀疏量测状态方程，结合图傅里叶分解技术构建电压的低维近似等式来减少参数量，进而实现低阶线性的近似利用，通过引入掩码概念，构造基于图信号处理的状态估计优化方程，实现了配电网电压的快速重建并提高了全局电压的质量。通过IEEE-85节点算例验证，所提方法相较半正定规划最小二乘法（Semi-Definite Programming Least Squares，SDP-SE）、矩阵补全方法在不同量测数量下的电压恢复精度平均提升50%，相较SDP-SE在运算时间上从分钟级降低至秒级，支持系统在线运行。在Cloudpss中搭建船舶电力系统模型并进行实验，在2s内实现了缺失电压的数据恢复，且电压的平均绝对误差为0.88%，验证了该方法在船舶电力系统通信系统部分损毁的情况下，能够准确掌握全局信息，提高系统的抗干扰能力和自愈能力。

关键词: 通信系统受损, 船舶电力系统, 图信号处理, 高分辨率量测

Abstract:

The shipboard equipment is constantly exposed to complex and harsh marine environments during the actual operation of ship.Some factors such as seawater corrosion，high humidity and mechanical vibrations accelerate the aging process of electrical components.Moreover，the human-induced operational errors，sudden natural disasters like severe storms，and unexpected equipment malfunctions can all cause significant damage to the shipboard power system and its communication infrastructure.Once damaged，a large amount of key data within the power system is lost，severely influencing the normal operation of the system and making it extremely difficult to accurately perceive the security situation，thus posing potential risks to the overall safety and reliability of the ship.Aiming at the issues of slow computation speed and subpar data restoration accuracy of the existing missing data recovery algorithms for power systems，this paper proposes an ultra-high-resolution measurement generation method based on minimal measurement information.In this method，the uneven spatial distribution of measurement points and their varying accuracies are fully considered to construct a spatial sparse measurement state equation for the power system，and the graph Fourier decomposition is employed to introduce a low-dimensional approximate equation of voltage for reducing the model parameters and enabling efficient low-order linear approximation.The innovative concept of a mask is also incorporated to build a state estimation optimization equation based on graph signal processing，achieving the rapid reconstruction of distribution network voltage and improving the global voltage quality.Through the IEEE-85 bus test case，the proposed method increases the voltage recovery accuracy by an average of 50% in the case of different measurement quantities compared with traditional methods like SDP-SE and MCSE，and reduces the computation time from minute-level to second-level compared with SDP-SE，supporting online system operation.Experiments in CloudPSS with an actual shipboard power system topological diagram show that the method can recover the missing voltage data within 1.2 seconds with an average absolute error of only 0.88%，validating its ability to accurately grasp the global information and enhance the anti-interference and self-healing capabilities of shipboard power system even when the communication system is partially damaged，ensuring the stable operation of shipboard power system.

Key words: communication impairment, shipboard power system, graph signal processing, high-resolution measurement

苏朝玉, 陈亮, 吴本祥, 张扬, 李耕. 基于图信号处理的船舶电力系统超分辨率量测生成方法[J]. 兵工学报, 2025, 46(S1): 250643-.

SU Zhaoyu, CHEN Liang, WU Benxiang, ZHANG Yang, LI Geng. An ultra-high-resolution Measurement Generation Method Based on Graph Signal Processing for Shipboard Power System[J]. Acta Armamentarii, 2025, 46(S1): 250643-.

图/表 13

图1 中压直流综合电力系统典型网络结构

Fig.1 Typical network structure of medium voltage DC integrated power system

图2 系统等效电路

Fig.2 Equivalent circuit diagram of the system

图3 IEEE-85节点配网拓扑

Fig.3 IEEE-85 node distribution net topology

表1 量测设备配置方案

Table 1 Configuration scheme of measurement device

量测设备	数量	观测误差/%
PMU	23	1
AMI	25	3
SCADA	37	5

图4 不同状态估计方式下的量测误差电压幅值

Fig.4 The amplitudes of measured error voltages under different state estimation methods

表2 不同量测设备数量下的3种状态估计方法性能对比

Table 2 Comparison of the performances of three state estimation methods under different numbers of measuring devices

	GSPv-SE		MCSE		SDP-SE
量测设备数量	MAE	F范数	MAE	F范数	MAE	F范数
23	0.0041	0.0437	0.0247	0.2564	0.0821	1.9021
43	0.0038	0.0416	0.0166	0.1673	0.0749	1.5514
64	0.0037	0.0414	0.0096	0.1055	0.0544	1.1662
85	0.0031	0.0210	0.0064	0.0669	0.0057	0.0609

表3 2种状态估计方法的模型参数与性能对比

Table 3 Comparison of model parameters and performances of two state estimation methods

模型	优化参数数量	已知参数数量	MAE	F-范数	运行时间/s
GSPv-SE	21	192	0.0037	0.0468	1.8562
SDP-SE	85	384	0.0512	1.1662	61.2625

图5 船舶电力系统拓扑

Fig.5 Topology of shipboard power system

图6 负荷-D4QF1受损

Fig.6 Damaged load-D4QF1

表4 负荷-D4QF1受损下部分节点状态

Table 4 Status of some nodes under the condition of load-D4QF1 being damaged

节点名称	电压值	注入电流	视在功率	是否可量测
母线-负载-D4QF1				不可量测
母线-负载-D3QF1	0.9998+ 0.0077j	-0.001+ 0.0004j	-0.001- 0.0004j	可量测
船舷配电与跨接板#7				不可量测
船舷配电与跨接板#6	0.9861+ 0.004j	0.0164+ 0.0121j	-0.0161- 0.012j	可量测
船舷配电与跨接板#8	0.9917+ 0.0024j	0.0079+ 0.0059j	-0.0078- 0.0059j	可量测

图7 场景1高分辨率量测恢复效果

Fig.7 Scenario 1：Super-high-resolution measurement and recovery renderings

图8 配电与跨接板#6受损

Fig.8 Damaged distribution and cross-connect panel #6

图9 场景2高分辨率量测恢复效果

Fig.9 Scenario 2：Super-high-resolution measurement and recovery renderings

参考文献 23

[1]	WANG T, ZHANG S, GU X P. A trend-based approach for situation awareness in power systems[J]. International Transactions on Electrical Energy Systems, 2017, 27(12).
[2]	REN L J, WANG T, SEKHARI SEKLOULI A, et al. A review on missing values for main challenges and methods[J]. Information Systems, 2023, 119:102268.
[3]	郑文迪, 聂建雄, 邵振国, 等. 智能配电网状态估计研究现状和展望[J]. 电力系统及其自动化学报, 2021, 33(4):8-16.
	ZHENG W D, NIE J X, SHAO Z G, et al. Research status and prospect of state estimation for smart distribution network[J]. Proceedings of the CSU-EPSA, 2021, 33(4):8-16. (in Chinese)
[4]	DE MELO V H P, MONTEIRO A L, ASCARI L B, et al. Review of power system state estimation and maturity level of market solutions:preceding steps[J]. Journal of Control,Automation and Electrical Systems, 2024, 35(4):702-719.
[5]	LIU Y, LIN Y Z, YUE K. Modern power system state estimation methods[J]. Encyclopedia of Electrical and Electronic Power Engineering, 2023, 2:259-277.
[6]	YANG T, LIU G L, WANG Y, et al. A tensor completion algorithm for missing user data in spot trading of electricity market[J]. Computers and Electrical Engineering, 2025, 122:109988.
[7]	ESPOSITO R, CERRATO M, LOCATELLI M. Partitioned least squares[J]. Machine Learning, 2024, 113(9):6839-6869.
[8]	晁婉萌, 刘灏, 毕天姝, 等. 考虑 N-1 故障重构下的配电网 PMU 优化配置方法[J/OL]. 中国电机工程学报, 2025(2025-03-06)[2025-06-16]. https://link.cnki.net/urlid/11.2107.tm.20250305.1328.006.
	CHAO W M, LIU H, BI T S, et al. Optimal configuration method of pmus in distribution network considering N-1 fault reconfiguration[J/OL]. Proceedings of the CSEE, 2025(2025-03-06)[2025-06-16]. https://link.cnki.net/urlid/11.2107.tm.20250305.1328.006. in Chinese)
[9]	吉兴全, 姜海洋, 张玉敏, 等. 考虑网络重构的三相不平衡配电网μPMU组合优化配置[J]. 电力系统自动化, 2025, 49(3):145-155.
	JI X Q, JIANG H Y, ZHANG Y M, et al. Combined optimal configuration of μpmus in three-phase unbalanced distribution networks considering network reconfiguration[J]. Automation of Electric Power Systems, 2025, 49(3):145-155. (in Chinese)
[10]	郑骁麟, 王天昊, 王凯奇, 等. 配电系统伪量测数据增强方法[J/OL]. 电力系统及其自动化报, 2025(2025-04-10)[2025-06-16].https://doi.org/10.19635/j.cnki.csu-epsa.001607
	ZHENG X L, WANG T H, WANG K Q, et al ZHANG X J., Pseudo-measurement data enhancement method for distribution systems[J/OL]. Proceedings of the CSU-EPSA, 2025(2025-04-10)[2025-06-16].https://doi.org/10.19635/j.cnki.csu-epsa.001607 in Chinese)
[11]	王玥, 于越, 郭嘉辉, 等. 基于改进 Crossformer 伪量测构建的主动配电网预测辅助状态估计方法[J]. 高电压技术, 2025, 51(6):3029-3042.
	WANG Y, YU Y, GUO J H, et al. Predictive auxiliary state estimation method for active distribution network based on improved crossformer pseudo-measurement construction[J]. High Voltage Engineering, 2025, 51(6):3029-3042. (in Chinese)
[12]	SUN B, XU Y J, GU W, et al. PMU data compression in power systems using adaptive rank-based tensor ring[J]. IEEE Transactions on Industrial Informatics, 2025, 21(7):5264-5275.
[13]	蒋睿珈, 余晓丹, 靳小龙, 等. 基于交替最小化矩阵补全及滑动平均的配电网时空量测数据补齐方法[J]. 电力自动化设备, 2025, 45(8):20-27.
	JIANG R J, YU X D, JIN X L, et al. A method for filling in spatio-temporal measurement data of distribution network based on alternating minimization matrix completion and moving average[J]. Electric Power Automation Equipment, 2025, 45(8):20-27. (in Chinese)
[14]	王子馨, 胡俊杰, 刘宝柱. 基于长短期记忆网络的电力系统量测缺失数据恢复方法[J]. 电力建设, 2021, 42(5):1-8. doi: 10.12204/j.issn.1000-7229.2021.05.001
	WANG Z X, HU J J, LIU B Z. A Method for recovering missing measurement data in power systems based on long short-term memory networks[J]. Electric Power Construction, 2021, 42(5):1-8. (in Chinese)
[15]	马彬喻, 杨军, 彭晓涛, 等. 基于改进 CVAE-GAN的电力系统暂态稳定评估样本增强方法[J/OL]. 电力自动化设备, 2025(2025-05-08)[2025-06-16].https://doi.org/10.16081/j.epae.202504023
	MA B Y, YANG J, PENG X T, et al. Sample enhancement method for power system transient stability assessment based on improved CVAE-GAN[J/OL]. Electric Power Automation Equipment, 2025(2025-05-08)[2025-06-16].https://doi.org/10.16081/j.epae.202504023 in Chinese)
[16]	吕奇峰, 陈颖, 肖谭南, 等. 基于图注意力网络的配电网超分辨率量测生成方法[J]. 电力系统自动化, 2023, 47(5):26-34.
	LU Q F, CHEN Y, XIAO T N, et al. A method for generating super-resolution measurements in distribution networks based on graph attention networks[J]. Automation of Electric Power Systems, 2023, 47(5):26-34. (in Chinese)
[17]	邱凯乐. 图信号处理综述[J]. 物联网技术, 2025, 15(5):111-113.
	QIU K L. A survey of graph signal processing[J]. Internet of Things Technology, 2025, 15(5):111-113. (in Chinese)
[18]	CAPUTO A. Vector-valued Graph Signal Processing[Z].arXiv:2505.22325,2025.
[19]	YOU C, ZHAO Y Q, HOU S Q, et al. Research on centre of inertia based frequency sampling and system frequency distribution characteristics[J]. IEEE Access, 2024, 12:71371-71378.
[20]	SONG W C, LU C, LIN J J, et al. A low-quality PMU dataidentification method with dynamic criteria based on spatial-temporal correlations and random matrices[J]. Applied Energy, 2023, 343:121213.
[21]	HAMMER F, BARBER S. Data imputation for SCADa data using graph neural networks[J]. Journal of Physics:Conference Series, 2025, 3025(1):12014.
[22]	LIU J G, TAN H J, ZHAO R F, et al. Applications of graph computing and graph neural networks in power systems:a survey[C]//Proceedings of the 2024 China International Conference on Electricity Distribution (CICED). Hangzhou,China:IEEE,2024:543-548.
[23]	SAXENA A, ASHA V, LALITHA G, et al. Expanding horizons:graph theory’s multifaceted applications[J]. E3SWeb of Conferences, 2024, 507:1015.