Capacitive voltage transformers (CVTs) are essential in high-voltage systems. An accurate error assessment is crucial for precise energy metering. However, tracking real-time quantitative changes in capacitive voltage...Capacitive voltage transformers (CVTs) are essential in high-voltage systems. An accurate error assessment is crucial for precise energy metering. However, tracking real-time quantitative changes in capacitive voltage transformer errors, particularly minor variations in multi-channel setups, remains challenging. This paper proposes a method for online error tracking of multi-channel capacitive voltage transformers using a Co-Prediction Matrix. The approach leverages the strong correlation between in-phase channels, particularly the invariance of the signal proportions among them. By establishing a co-prediction matrix based on these proportional relationships, The influence of voltage changes on the primary measurements is mitigated. Analyzing the relationships between the co-prediction matrices over time allows for inferring true measurement errors. Experimental validation with real-world data confirms the effectiveness of the method, demonstrating its capability to continuously track capacitive voltage transformer measurement errors online with precision over extended durations.展开更多
【目的】随着智慧城市建设中信息技术的深度应用,GNSS轨迹数据呈爆炸式增长,但其轨迹生成过程易受信号干扰与传感器故障影响而产生噪声。本文旨在设计新型噪声识别与修复算法,以提升原始GNSS轨迹数据的处理精度与质量。【方法】针对轨...【目的】随着智慧城市建设中信息技术的深度应用,GNSS轨迹数据呈爆炸式增长,但其轨迹生成过程易受信号干扰与传感器故障影响而产生噪声。本文旨在设计新型噪声识别与修复算法,以提升原始GNSS轨迹数据的处理精度与质量。【方法】针对轨迹噪声识别问题,本文提出基于密度矩阵的自适应DBSCAN算法,其具有超参数无关特性,可敏感捕获低幅值噪声点,同时避免连续转向点的误判。针对噪声修复问题,提出基于轨迹分段的函数构造式修复算法:首先采用道格拉斯-普克(Douglas-Peucker,DP)算法压缩轨迹数据实现分段;其次定位含噪声轨迹段,基于段内有效点构造拟合函数;最终依据相邻点时空属性修复噪声数据。相较于主流插值算法(如拉格朗日、牛顿、埃尔米特、线性、三次样条及最近邻插值),本方法通过规避全局特征依赖,显著保留了噪声点蕴含的局部信息特征。【结果】基于长春市1500名志愿者2024年8月19日—9月1日的原始GNSS轨迹数据,设计2组对比实验。第1组将新型识别算法与原始DBSCAN及其主流衍生算法(KANN-DBSCAN、BDT-ADBSCAN)进行对比。实验表明:新算法在轮廓系数(SC)、Calinski-Harabasz指数(CHI)、Da‐vies-Bouldin指数(DBI)3项指标均取得最优值,优化幅度分别为40.17%~381.80%、20.03%~235.18%、23.42%~79.53%。第2组实验对比新型修复算法与6类经典插值方法(拉格朗日、牛顿、埃尔米特、线性、三次样条、最近邻),结果显示:新算法在轨迹相似性度量指标(Dynamic Time Warping,DTW)上全面优于对比方法,整体优化幅度达43.18%~80.43%。【结论】本文提出的噪声识别与修复算法显著提升了原始GNSS轨迹的质量精度,可高效支撑大规模轨迹数据预处理任务,为时空轨迹挖掘研究提供高质量数据基础。展开更多
文摘Capacitive voltage transformers (CVTs) are essential in high-voltage systems. An accurate error assessment is crucial for precise energy metering. However, tracking real-time quantitative changes in capacitive voltage transformer errors, particularly minor variations in multi-channel setups, remains challenging. This paper proposes a method for online error tracking of multi-channel capacitive voltage transformers using a Co-Prediction Matrix. The approach leverages the strong correlation between in-phase channels, particularly the invariance of the signal proportions among them. By establishing a co-prediction matrix based on these proportional relationships, The influence of voltage changes on the primary measurements is mitigated. Analyzing the relationships between the co-prediction matrices over time allows for inferring true measurement errors. Experimental validation with real-world data confirms the effectiveness of the method, demonstrating its capability to continuously track capacitive voltage transformer measurement errors online with precision over extended durations.
文摘【目的】随着智慧城市建设中信息技术的深度应用,GNSS轨迹数据呈爆炸式增长,但其轨迹生成过程易受信号干扰与传感器故障影响而产生噪声。本文旨在设计新型噪声识别与修复算法,以提升原始GNSS轨迹数据的处理精度与质量。【方法】针对轨迹噪声识别问题,本文提出基于密度矩阵的自适应DBSCAN算法,其具有超参数无关特性,可敏感捕获低幅值噪声点,同时避免连续转向点的误判。针对噪声修复问题,提出基于轨迹分段的函数构造式修复算法:首先采用道格拉斯-普克(Douglas-Peucker,DP)算法压缩轨迹数据实现分段;其次定位含噪声轨迹段,基于段内有效点构造拟合函数;最终依据相邻点时空属性修复噪声数据。相较于主流插值算法(如拉格朗日、牛顿、埃尔米特、线性、三次样条及最近邻插值),本方法通过规避全局特征依赖,显著保留了噪声点蕴含的局部信息特征。【结果】基于长春市1500名志愿者2024年8月19日—9月1日的原始GNSS轨迹数据,设计2组对比实验。第1组将新型识别算法与原始DBSCAN及其主流衍生算法(KANN-DBSCAN、BDT-ADBSCAN)进行对比。实验表明:新算法在轮廓系数(SC)、Calinski-Harabasz指数(CHI)、Da‐vies-Bouldin指数(DBI)3项指标均取得最优值,优化幅度分别为40.17%~381.80%、20.03%~235.18%、23.42%~79.53%。第2组实验对比新型修复算法与6类经典插值方法(拉格朗日、牛顿、埃尔米特、线性、三次样条、最近邻),结果显示:新算法在轨迹相似性度量指标(Dynamic Time Warping,DTW)上全面优于对比方法,整体优化幅度达43.18%~80.43%。【结论】本文提出的噪声识别与修复算法显著提升了原始GNSS轨迹的质量精度,可高效支撑大规模轨迹数据预处理任务,为时空轨迹挖掘研究提供高质量数据基础。
