基于小电流电器的滑动窗双边CUSUM事件检测改进算法

事件检测是非侵入式负荷监测中的关键部分,然而事件检测方法对于一些小电流电器存在漏检问题。为此,提出一种基于小电流电器的滑动窗双边CUSUM事件检测改进算法,即在均值计算窗和暂态检测窗的基础上,引入方差计算窗区分运行时电流波动小的电器,通过权重参数δ提高检测过程中投入、切出事件的累计和,解决了滑动窗双边CUSUM事件检测算法的小电流电器漏检问题。采用方差阈值判断电器是否进入稳态,提高了电器进入稳态时检测的准确性,有效记录事件投入点和事件切出点。实测验证表明,所提算法不仅能够准确检测到传统算法易忽略的小电流电器的暂态事件,还能准确记录电器完整的事件投切过程,有利于其暂态过程的分析与处理,保证了特征提取的有效性,为事件检测方法的优化方向提供了借鉴。

Improved Algorithm of Sliding Window Double CUSUM Event Detection Based on Small Current Electric Appliance

Event detection is a key part of non-invasive load monitoring. However, event detection method has the problem of missing detection for some small current appliances. Therefore, an improved algorithm of sliding window double cusum event detection based on small current electric appliance is proposed, that is, on the basis of mean calculation window and transient detection window, variance calculation window is introduced to distinguish the appliances with small current fluctuation during operation. Weight parameter δ is used to improve the accumulative sum of on and off events during detection, which solves the problem of small current electrical leakage. The variance threshold is used to judge whether the electrical apparatus enters the steady state, which improves the accuracy of the detection when the electrical apparatus enters the steady state, and effectively records the event input point and event cutting point. Measurement verification show that the proposed algorithm not only can accurately detect the transient event of small current appliances that the traditional algorithm is easy to ignore, can also accurately record electrical appliances complete event which is helpful for the transient process of the analysis and processing, ensuring the effectiveness of the feature extraction. Thus, it provides reference for optimization direction of event detection method.