基于彩色V-I轨迹特征和边缘机器学习非侵入式负荷识别方法

陆玲霞 ,  孟繁举 ,  于淼 ,  任沁源 ,  包哲静

工程科学与技术 ›› 2025, Vol. 57 ›› Issue (05) : 134 -141.

PDF (2685KB)
工程科学与技术 ›› 2025, Vol. 57 ›› Issue (05) : 134 -141. DOI: 10.12454/j.jsuese.202400042
新型电力系统

基于彩色V-I轨迹特征和边缘机器学习非侵入式负荷识别方法

作者信息 +

Non-intrusive Load Monitoring Based on Colored V-I Trajectory Features and Edge Machine Learning

Author information +
文章历史 +
PDF (2749K)

摘要

非侵入式负荷识别方法作为分析用户用电行为的主要途径,对开展能耗监测、实现用电安全评估具有重要意义。针对传统基于V-I轨迹特征的非侵入式负荷识别方法存在特征重叠和无法识别未知负荷的问题,提出一种基于彩色V-I轨迹特征和轻量级孪生网络的非侵入式负荷识别方法。首先,通过负荷电压电流数据构建具有方向信息的彩色V-I轨迹图像。然后,利用孪生网络计算待识别负荷的V-I轨迹图像和负荷特征库中V-I轨迹图像之间的相似度,以完成初步识别。随后,计算电流谐波特征之间的余弦距离,与阈值对比完成最终负荷识别。在以STM32MP1微处理器为核心的嵌入式Linux系统上,使用实验室电器负荷进行了实物验证。结果表明:彩色V-I轨迹能更详细地反映负荷特征,提高负荷识别准确率,并且由于改进的人工智能模型比较轻量化,对计算量需求大大减小,可以在嵌入式设备端对负荷特征库进行动态实时在线更新,轻松完成模型再训练。与依赖服务器的传统算法相比,无需返回PC或服务器重新训练模型并重新移植模型到嵌入式设备端。该方法仅依赖嵌入式终端便可准确识别未知负荷,避免在出现较多未知负荷后识别准确率下降,保证了负荷识别效果。系统运算一次负荷识别时间为0.2 s左右,可以满足实时性要求,具有重要的研究价值和实用性。

Abstract

Objective A large amount of data is generated during the process of power consumption, which serves as the basis for informed user decisions and helps realize home energy efficiency monitoring, safety protection, and demand-side management. Therefore, obtaining real-time electricity consumption information through load identification has significant research value. Non-intrusive load monitoring (NILM) refers to a method that does not require the installation of a monitoring device for each individual load. Instead, it analyzes only the voltage and current data on the bus to obtain information about various factors, such as load type, operating status, and power consumption. Studies have shown that non-intrusive load recognition helps users reduce their energy consumption by up to 15%. As a primary tool for analyzing users’consumption behavior, non-intrusive load identification is crucial for both energy-use monitoring and electrical safety assessment. Methods based on the V-I trajectory have indicated promise, but several limitations remain. First, V-I trajectories often overlap: many appliances produce very similar shapes, especially after normalization, so using the trajectory alone makes it challenging to separate look-alike loads. This calls for integrating auxiliary cues, power and power-factor statistics, harmonic features, or other spectral or temporal descriptors. Second, many approaches are closed-set. They frame recognition as multiclass classification and identify only classes seen during training, while unknown loads are mishandled. The associated neural classifiers are also relatively complex. Third, training pipelines typically depend on server-class compute. When ported to embedded edge devices, limited resources prevent timely model updating or retraining. As the number of unseen loads grows, accuracy degrades without on-device or real-time adaptation, making it challenging to guarantee effectiveness and latency in practice and revealing a gap between lightweight AI and deployable systems. Methods This study proposed a NILM method that combined colored voltage-current (V-I) trajectory features with a lightweight Siamese network. The aim is to address the overlapping features caused by different loads. The method has two main objectives: to enhance the information conveyed by V-I images and to maintain a compact inference pipeline suitable for embedded, real-time NILM applications. This study introduced a simple and effective method for constructing colored V-I trajectory images. The approach involved creating colored V-I trajectory images that incorporated directional information using load voltage and current data, while employing the RGB color channels of the images. Compared to traditional V-I trajectories, this new method captured more load characteristics and clearly reflected the distinct features of each load. The improved colored V-I trajectory enabled a more detailed depiction of load characteristics, enhancing the accuracy of load identification. Given that the load V-I trajectory images were not overly complex and that NILM must operate online on embedded devices, a highly complex neural network structure was unnecessary. This study employed a Siamese network to calculate the similarity between the V-I trajectory image of the load to be identified and the V-I trajectory images in the load feature database, which enabled preliminary identification. The Siamese network model consisted of two components: the convolutional neural network (CNN) model and the backpropagation (BP) model. The CNN model extracted feature vectors from the colored V-I trajectory images, while the BP model computed the similarity between these feature vectors. The CNN model referenced was the classical lightweight neural network structure known as LeNet-5, which produced a 32-dimensional feature vector. LeNet-5 has a simple architecture and low complexity, which makes it suitable for real-time operation. The BP model took as input a 64-dimensional vector derived from merging two 32-dimensional feature vectors obtained by the CNN model, and its output was a continuous value between 0 and 1. An evaluation was made on whether the input vectors belonged to the same class by comparing the output results of the Siamese network with a predetermined threshold. The second stage incorporated harmonic features for identification after the initial recognition using the Siamese network to prevent misidentification of different loads with similar V-I trajectories. Specifically, the cosine distance between the harmonic features of the current was calculated and compared against a threshold to complete the load identification process. Initially, loads in the harmonic feature database corresponding to the V-I trajectory image that has matching harmonic features (including the 1st, 3rd, 5th, and 7th harmonics) were identified, and the similarity between these harmonic features was assessed using cosine distance. A greater cosine distance indicated higher similarity between the two loads. If the similarity exceeded the threshold, the load to be identified was classified as a known load in the database. In contrast, if the similarity fell below the threshold, the load was considered a new load, and its feature vector was added to the feature database. Through this dynamic updating of the feature database, the load was recognized as known if it appeared again. The method effectively avoided misidentifications that could have arisen from the similarities in V-I trajectories of different loads by integrating the features of V-I trajectories with harmonic features. Results and Discussions The Siamese network model was trained using V-I trajectories from the WHITED dataset and was deployed on an embedded Linux system powered by an STM32MP1 microprocessor. It was then validated using laboratory electrical loads. The results indicated that the colored V-I trajectory provided a more detailed representation of load characteristics, which enhanced the accuracy of load identification. In addition, the improved artificial intelligence model was lightweight, which significantly reduced computational requirements. The on-device feature database was updated online in real time, and local retraining or incremental updates were supported. Unlike server-dependent pipelines, this deployment eliminated the requirement for a round-trip to a PC or server for retraining and subsequent redeployment to the edge device. Conclusions The method employs the embedded terminal to accurately identify unknown loads, preventing a reduction in recognition accuracy as the number of unknown loads increases and ensuring effective load recognition. The system's runtime for identifying a single load is approximately 0.2 seconds, which satisfies real-time requirements and demonstrates substantial research value and practicality. The colored V-I trajectory combined with a lightweight Siamese network model provides a robust and deployable approach for real-time NILM on resource-constrained embedded hardware.

Graphical abstract

关键词

非侵入式负荷识别 / 边缘机器学习 / 孪生网络 / 嵌入式Linux系统

Key words

non-intrusive load monitoring / edge machine learning / siamese network / embedded Linux system

引用本文

引用格式 ▾
陆玲霞,孟繁举,于淼,任沁源,包哲静. 基于彩色V-I轨迹特征和边缘机器学习非侵入式负荷识别方法[J]. 工程科学与技术, 2025, 57(05): 134-141 DOI:10.12454/j.jsuese.202400042

登录浏览全文

4963

注册一个新账户 忘记密码

本刊网刊
21 网络出版日期:2024 ‒ 网上出版日期:03 ‒ 29
电能消费过程中产生了海量数据,这些用电信息可作为用户用电决策的依据,有助于家庭能效监测、安全防护及需求侧管理的实现[13]。因此,如何通过负荷识别实时获取用电信息具有重要的研究价值[46]。非侵入式负荷识别(NILM)方法由Hart在1992年第一次提出,该方法不需要对每个负荷安装监测装置,只需对总线上的电压、电流数据进行分析就可获得负荷类别、运行状态及耗电情况等多种信息[79]。研究表明,NILM的应用可帮助用户节省高达15%的能耗[10]
近年来,国内外研究者从不同角度、方向提出了许多不同的非侵入式负荷分解方法[11]。荆永震[12]对负荷电流进行快速傅里叶变换得到谐波分量作为负荷特征,结合决策树和Logistic回归等机器学习方法进行非侵入式负荷识别。王爱囡等[13]在文献[12]基础上,通过TOPSIS相似度匹配算法进行负荷监测。但是,电流谐波特征无法表示设备阻抗特征,也难以处理多状态负荷[14],而V-I轨迹是电器负荷的高频特征之一,可以反映负荷稳态运行时的电流波形、阻抗特性等电器特征[1517]。周任飞等[18]通过卷积神经网络(CNN)算法对负荷V-I轨迹图像进行特征提取,可以提高功率特征相似但V-I轨迹不同的电器负荷的识别准确率。Liu[19]和宰州鹏[20]等利用彩色V-I轨迹图像和迁移学习进行负荷监测。王守相等[14]结合负荷功率特征和V-I轨迹特征来获得负荷特征,并利用神经网络实现负荷监测。汪颖等[21]也将负荷功率和V-I轨迹特征结合,先利用K-means算法将负荷功率特征聚类,将功率特征相似的电器设备聚为同组;再利用V-I轨迹特征和CNN针对功率相似的电器负荷进行精细化识别。
虽然基于V-I轨迹的方法在负荷识别方面取得了很好的效果,但仍存在以下问题:第一,V-I轨迹容易产生特征重叠,很多负荷的V-I轨迹形状相似,若只利用V-I轨迹,难以识别这些V-I轨迹相似并重叠的负荷,所以还需融合其他特征,如功率或者谐波特征等。第二,很多方法对未知负荷效果欠佳。因为这些方法将负荷识别作为多分类问题,所以只能对参加训练的负荷进行识别,并且用于分类的神经网络复杂度较高。第三,很多负荷识别方法训练模型依赖服务器或电脑的算力支持,若移植到嵌入式边端使用,会因其算力有限而不能很好地完成模型再训练。随着未知负荷增多,若不实时训练,模型识别准确率下降,难以保证负荷识别效果和实时性,与人工智能轻量化和部署落地还有一定距离[22]
基于上述问题,本文提出了一种基于彩色V-I轨迹特征和边缘机器学习的非侵入式负荷识别方法。首先,通过负荷电压和电流数据构造具有轨迹方向信息的彩色V-I轨迹图像,使其较传统V-I轨迹携带更多负荷特征,将其输入到轻量级孪生网络,计算与特征库中负荷的相似度,进行初步分类。与传统基于V-I轨迹的NILM方法利用神经网络进行分类处理不同,本文所设计的方法利用计算特征相似度进行判别,从而大大降低了模型复杂度要求。然后,为实现V-I轨迹相似负荷的识别,计算负荷谐波特征之间的余弦距离,根据余弦距离的大小完成最终识别。在以STM32MP1为核心的嵌入式Linux平台上结合TensorFlow Lite工具包进行负荷识别实验,对实验室多个电器负荷进行了实物验证。

1 彩色V-I轨迹和边缘机器学习

1.1 彩色V-I轨迹

虽然V-I轨迹能够反映负荷稳态运行时的电流波形、阻抗特性等特征,但是传统的V-I轨迹无法反映负荷的功率特征[14,23],而且很多负荷的V-I轨迹具有特征重叠的现象,如加热用的家用电器基本为纯电阻型,其V-I轨迹均为一条直线。目前,虽有很多改进的V-I轨迹图像构建方法,但大都比较复杂,需要大量的计算时间和资源[24]

针对这一问题,本文提出一种简单有效的彩色V-I轨迹图像构建方法,利用彩色图像的RGB 3个通道,充分反映每个负荷之间的不同特征。构建长×宽为2N×2N的彩色V-I轨迹图像的具体方法如下:

1)设置每个像素点的初始值为(0,0,0)

2)设负荷稳态时的电压、电流数据由s个采样点组成,计算电压、电流绝对值的最大值UmaxImax

3)计算Δu=UmaxNΔi=ImaxN,其中,N为长和宽尺寸的一半,NZ+

4)对于每个采样点数据(UjIj),其中,0< j  s,计算Xj=N+ceilUjuYj=N+ceil(IjΔi),其中,ceil( )为向上取整函数。

5)设置(XjYj)像素点的值:当0<j<s3时,将(Xj,Yj)像素点的值置为(255,0,0);当s3j<2s3时,将(Xj,Yj)像素点的值置为(0,255,0);当2 s3j<s时,将(Xj,Yj)像素点的值置为(0,0,255)

由上述步骤可知,构建彩色V-I轨迹图像时根据采样点的时间先后顺序为其对应的像素点设置了不同的RGB颜色值,由此,彩色V-I轨迹相较传统V-I轨迹增加了轨迹方向信息,能反映更多负荷特征。

传统的V-I轨迹和改进的彩色V-I轨迹如图1所示。图1中,加热器和空调两个负荷的功率接近,只有电流信号有微小的区别。

图1可以发现:传统的V-I轨迹无法对加热器和空调两个负荷进行有效区分;本文提出的彩色V-I轨迹利用轨迹的颜色信息,可以准确区分这两个负荷。改进的彩色V-I轨迹能够更详细地反映负荷特征,从而有助于提升负荷识别的准确率。

1.2 嵌入式人工智能

训练和运行深度学习模型需要很高的运算能力,目前深度学习框架大多基于具有GPU运算能力的PC终端或功能强大的服务器[24]。近年来,随着高性能、低能耗的嵌入式设备的出现,边缘机器学习成为可能。

嵌入式人工智能作为能在边缘设备上进行机器学习的一种轻量化人工智能技术,其优点是无需在设备与服务器间来回发送数据,有助于缩短延迟、保护隐私和降低功耗。与依赖服务器的传统算法相比,嵌入式人工智能方法具有较低的成本和良好的便携性。但由于神经网络需要大量训练数据以保证准确率,受限于嵌入式设备算力,现有的非侵入式负荷识别方法仅能将预训练好的模型从服务器部署到终端,还难以在边缘设备上进行模型训练,缺少终端神经网络模型在线学习和更新功能,从而无法识别未参加训练的负荷,甚至因模型太大无法通过轻量化将其移植到嵌入式设备,准确率和实时性均得不到保障。传统的嵌入式人工智能方法通过模型剪枝、特征降维等方式将复杂神经网络进行简化,以适应嵌入式设备算力要求,但不可避免地降低了模型的准确度。

本文通过设计一种轻量化孪生网络模型、利用计算特征相似度的方式来实现负荷识别,有效降低了模型复杂度,并将该模型无损地移植到嵌入式设备。所用嵌入式设备以STM32MP157为核心,主控采用MCU+MPU架构。为能在STM32MP1嵌入式设备上运行神经网络算法,需通过轻量化工具TensorFlow Lite将服务器上预训练好的模型移植到嵌入式设备,实现基于边缘机器学习的负荷识别和模型再训练,其脱离服务器也能保证准确率和实时性。TensorFlow Lite作为边缘机器学习的强大工具,是TensorFlow的轻量级版本,可以在嵌入式Linux、安卓、iOS等多个平台部署,而且具有适用于多种编程语言的应用程序编程接口,包括Java、C++、Python、Swift[2526]等。TensorFlow Lite转换器将训练好的自定义模型转换为“.tflite”格式,再使用TensorFlow Lite解释器在边缘设备端快速运行自定义模型。

3 基于边缘机器学习的非侵入式负荷识别系统设计

3.1 系统的整体设计

本文设计的基于边缘机器学习的非侵入式负荷识别系统整体框图如图2所示。

硬件系统主要包括以STM32MP1为核心的嵌入式Linux系统、数据采集模块、用户界面。STM32MP1核心板包含一个M4内核及双A7内核,其中:M4内核负责实施实时性要求高的数据采集及负荷投切事件检测,根据事件点的起始点信息检测结果提取负荷特征等信息。当检测到有效的负荷投切事件后,才进行后续的非侵入式负荷识别,负荷投切事件检测是实现负荷识别的第1个环节,而且是很重要的一环。A7内核运行嵌入式Linux系统,主要负责算力要求高的算法运算,如神经网络的运行。数据采集模块由电压互感器和电流互感器组成,STM32MP1自带的ADC通道连接数据采集模块,来获取电压和电流信号值。用户界面实时显示当前用电设备的运行情况,遇到未知负荷时可以录入即新增负荷名称。非侵入式负荷识别系统实物如图3所示。

3.2 非侵入式负荷识别算法

3.2.1 基于彩色V-I轨迹和孪生网络的第1阶段识别

在第1阶段识别中,将彩色V-I轨迹图像作为输入,利用孪生网络完成特征提取和相似度计算。孪生网络[2728]作为一种结构特殊的神经网络,相比于传统的网络模型具有两个优点:第一,训练数据由两个输入组成,这种两两组合的形式可以扩展数据集,因此每个负荷不需要太多样本数据;第二,孪生网络的输出是两个输入之间的相似度,所以通过设置阈值的方式可以对未知负荷进行识别。

本文提出的孪生网络模型包括两个部分,如图4所示。

图4可见,CNN模型负责提取彩色V-I轨迹图像的特征向量,BP模型负责计算V-I轨迹特征向量之间的相似度。CNN模型参考经典的轻量级神经网络模型Lenet-5[29],最后输出32维特征向量。虽然有很多比Lenet-5结构上更深度的神经网络模型,如AlexNet、VGG16、GoogleNet等,但是因为负荷V-I轨迹图像并不复杂,而且NILM需要在嵌入式设备上在线运行,所以神经网络模型并不需要非常复杂的结构。Lenet-5结构简单,复杂度低,能够满足实时运行需求。BP模型的输入是两个V-I轨迹图像经过CNN模型获取的32维特征向量合并得到的64维向量,输出是一个0~1之间连续取值的数,将孪生网络的输出结果与给定的阈值比较后得到输入向量是否属于同一类的判定结果。

3.2.2 基于谐波特征和余弦距离的第2阶段识别

利用孪生网络对彩色V-I轨迹图像进行初步识别后,为避免V-I轨迹相似的不同负荷的误识别,通过谐波特征进行第2阶段识别。在V-I轨迹图像对应的谐波特征库里找出谐波特征(包含1次、3次、5次、7次)匹配的负荷,利用余弦距离计算谐波特征之间的相似度,计算公式如下:

cosa, bi=x1yi1+x2yi2+x3yi3++xnyin(x12+x22+x32++xn2)(yi12+yi22+yi32++yin2)

式中:a为通过负荷投切事件和快速傅里叶变换得到的负荷谐波特征向量,a=(x1,x2,x3,,xn)bi为特征库中负荷的谐波特征向量,bi=(yi1,yi2,yi3,,yin)。余弦距离越大,则两个负荷的相似度越大。当相似度大于阈值时,即可将待识别负荷识别为特种库中负荷。当相似度小于阈值时,认为待识别负荷是新的负荷,并将该负荷的特征向量添加到特征库里。通过动态更新特征库,当该负荷下一次出现时,即可识别为已知负荷。

3.2.3 系统整体算法流程

结合两个阶段的识别,负荷识别算法整体流程如图5所示。

图5可知:系统开始运行之后,对负荷投切事件进行检测。当发生负荷投切事件后,根据电压、电流数据构建彩色V-I轨迹图像。通过CNN网络提取彩色V-I轨迹的特征向量,利用快速傅里叶变换(FFT)提取谐波特征。提取特征之后,查看负荷特征库是否为空。当特征库为空时,直接把提取到的特征添加到负荷特征库里。当特征库已有负荷,先计算待识别负荷和特征库中负荷V-I轨迹特征向量之间的相似度。

当V-I轨迹特征向量最大的相似度大于阈值(经多次测试,本文设为0.8)时,即认为该负荷的V-I轨迹图像已在特征库中。继续计算待识别负荷和特征库中负荷谐波特征向量之间的相似度(余弦距离)。当最大的相似度大于阈值时,即可将待识别负荷识别为特征库中的负荷。当最大的相似度小于阈值时,认为待识别负荷为未知负荷,将其谐波特征添加到相应的V-I轨迹组合,并做标签。当V-I轨迹特征向量最大的相似度小于阈值时,认为待识别负荷为新的负荷。将该负荷的V-I轨迹特征向量和谐波特征添加到负荷特征库,并做标签。

4 非侵入式负荷识别系统实现和测试结果

4.1 孪生网络模型预训练

为了构造训练数据集,利用WHITED数据集中的数据构建彩色V-I轨迹图像。WHITED数据集[30]包含44 kHz采样频率下的54种不同类型家用电器的电压、电流数据。本文利用20个不同负荷的彩色V-I轨迹图像构造训练集,每个负荷有15张V-I轨迹图像样本,一些典型负荷的彩色V-I轨迹图像如图6所示。图6中,V-I轨迹图像的分辨率设置为32×32像素点,分辨率过低会损失较多细节特征[20]

孪生网络训练完之后,将CNN模型和BP模型分开保存,以便更新特征库时只需保存CNN模型输出的32维向量即可。最后,利用TensorFlow Lite转换器将模型转换为“.tflite”格式,并移植到以STM32MP1为核心的嵌入式Linux系统上。

4.2 实际负荷测试

测试所用到的负荷包括笔记本电脑、小冰箱、吹风器、微波炉、加热器、熨斗等常见负荷,如图7所示。

系统测试界面如图8所示。由图8可见,系统刚开始运行时,将所有负荷都识别为未知负荷,因此,第1次运行时需要人为设置标签。加热器和吹风机有两个工作模式,并且,每个工作模式对应的V-I轨迹相同,所以,这两个工作模式需要通过第2阶段的电流谐波特征进行区分。当每个负荷的所有不同运行状态都被添加到负荷特征库后,可以正常进行识别。图7所示的8个被测试的实际电器负荷识别结果如表1所示。

表1可知:投切20次时,除了电脑以外所有负荷的识别率都达到100%。由于电脑相比吹风机和加热器等负荷而言功率不够稳定,电流波动较大,所以在第2阶段对比电流谐波特征时,有两次出现余弦距离小于阈值,从而被识别成未知负荷。但系统总的负荷识别准确率达到了98.75%。

最后,测试本文提出的负荷识别方法的实时性能,发现在STM32MP1处理器上完成1次识别只需要0.2 s,因此可以满足整个系统的实时性要求。

5 结 论

本文提出了一种基于彩色V-I轨迹特征和轻量级孪生网络的非侵入式负荷识别方法。利用负荷的彩色V-I轨迹图像,先通过孪生网络进行第1阶段识别,再通过计算负荷谐波特征之间的余弦距离进行第2阶段识别。通过V-I轨迹与谐波特征的特征融合,有效避免了不同负荷V-I轨迹相似可能导致的误识别。利用WHITED数据集中的V-I轨迹对孪生网络模型进行训练,并将模型移植到嵌入式Linux系统。在实验室实际电器设备上的测试表明,所提方法能够在负荷不参与预先训练的情况下,对其进行实时、准确识别,从而提升了NILM系统在不同环境下的适应性。结果还表明,彩色V-I轨迹能够更详细地反映负荷特征,提高负荷识别准确率。所提出的方法较轻量化,对计算量需求大大减小,可以在嵌入式设备端对负荷特征库进行动态更新,轻松完成模型再训练,能准确识别未知负荷;且运算一次负荷识别时间为0.2 s左右,满足实时性要求。

下一步,将针对空调、电脑等功率不稳定负荷的非侵入式识别开展进一步研究,以期提升NILM在不同应用场景中的适用性。

参考文献

[1]

Asres M W, Ardito L, Patti E.Computational cost analysis and data-driven predictive modeling of cloud-based online-NILM algorithm[J].IEEE Transactions on Cloud Computing,2022,10(4):2409‒2423. doi:10.1109/tcc.2021.3051766

[2]

Dong Kun, Zhao Jianfeng, Sun Ruichen,et al.Research status and prospects of modeling methods for power loads with new load elements[J].Automation of Electric Power Systems,2023,47(23):70‒83.

[3]

董坤,赵剑锋,孙睿晨,.含新型负荷元件的电力负荷建模方法研究现状与展望[J].电力系统自动化,2023,47(23):70‒83.

[4]

Zhang Li, Liu Qinglei, Zhang Hongwei.Home load optimization scheduling strategy based on improved binary particle swarm optimization algorithm[J].Electric Power,2023,56(5):118‒128.

[5]

张丽,刘青雷,张宏伟.基于改进二进制粒子群算法的家庭负荷优化调度策略[J].中国电力,2023,56(5):118‒128.

[6]

Qu Leitao, Kong Yaguang, Li Meng,et al.A residual convolutional neural network with multi-block for appliance recognition in non-intrusive load identification[J].Energy and Buildings,2023,281:112749. doi:10.1016/j.enbuild.2022.112749

[7]

D'Incecco M, Squartini S, Zhong Mingjun.Transfer learning for non-intrusive load monitoring[J].IEEE Transactions on Smart Grid,2020,11(2):1419‒1429. doi:10.1109/tsg.2019.2938068

[8]

Wang Fanrong, Xiang Kun, Liu Hui.Research on non-intrusive load decomposition based on improved AP clustering and optimized GRNN[J].Advanced Engineering Sciences,2020,52(4):56‒65.

[9]

汪繁荣,向堃,刘辉.基于改进AP聚类与优化GRNN的非侵入式负荷分解研究[J].工程科学与技术,2020,52(4):56‒65.

[10]

Buddhahai B, Wongseree W, Rakkwamsuk P.An energy pre-diction approach for a nonintrusive load monitoring in home appliances[J].IEEE Transactions on Consumer Electronics,2020,66(1):96‒105. doi:10.1109/tce.2019.2956638

[11]

Guo Hongxia, Lu Jinwei, Yang Ping,et al.Review on key techniques of non-intrusive load monitoring[J].Electric Power Automation Equipment,2021,41(1):135‒146.

[12]

郭红霞,陆进威,杨苹,.非侵入式负荷监测关键技术问题研究综述[J].电力自动化设备,2021,41(1):135‒146.

[13]

Li Ruyi, Zhang Peng, Liu Yongguang,et al.Non-intrusive household load identification method based on random forest[J].Electrical Measurement & Instrumentation,2021,58(4):9‒16. doi:10.19753/j.issn1001-1390.2021.04.002

[14]

李如意,张鹏,刘永光,.基于随机森林的非侵入式家庭负荷辨识方法[J].电测与仪表,2021,58(4):9‒16. doi:10.19753/j.issn1001-1390.2021.04.002

[15]

Parvin K, Hannan M A, Al-Shetwi A Q,et al.Fuzzy based particle swarm optimization for modeling home appliances towards energy saving and cost reduction under demand response consideration[J].IEEE Access,2020,8:210784‒210799. doi:10.1109/access.2020.3039965

[16]

Lin Pingchuan, Lu Lei, Gu Chao,et al.Non-intrusive load monitoring method based on CFSFDP graph Laplace algorithm[J].Advanced Engineering Sciences,2023,55(4):216‒223.

[17]

林平川,路磊,谷超,.基于CFSFDP图拉普拉斯算法的非侵入式负荷监测方法[J].工程科学与技术,2023,55(4):216‒223.

[18]

Jing Yongzhen.Development of non-intrusive load monitoring method[D].Guangzhou:Guangdong University of Technology,2019.

[19]

荆永震.非侵入式负荷识别系统研发[D].广州:广东工业大学,2019.

[20]

Wang Ainan, Yuan Lu, Ding Changbo,et al.Design and implementation of a non-intrusive power load monitoring system[J].Journal of Xi'an University of Technology,2019,35(3):343‒351.

[21]

王爱囡,袁麓,丁长波,.一种非侵入式电力负荷监测系统的设计与实现[J].西安理工大学学报,2019,35(3):343‒351.

[22]

Wang Shouxiang, Guo Luyang, Chen Haiwen,et al.Non-intrusive load identification algorithm based on feature fusion and deep learning[J].Automation of Electric Power Systems,2020,44(9):103‒110. doi:10.7500/AEPS20190625010

[23]

王守相,郭陆阳,陈海文,.基于特征融合与深度学习的非侵入式负荷辨识算法[J].电力系统自动化,2020,44(9):103‒110. doi:10.7500/AEPS20190625010

[24]

Yu Miao, Wang Bingnan, Lu Lingxia,et al.Non-intrusive adaptive load identification based on Siamese network[J].IEEE Access,2022,10:11564‒11573. doi:10.1109/access.2022.3145982

[25]

Kang Jusong, Yu Miao, Lu Lingxia,et al.Adaptive non-intrusive load monitoring based on feature fusion[J].IEEE Sensors Journal,2022,22(7):6985‒6994. doi:10.1109/jsen.2022.3155883

[26]

Liu Yu, Liu Congxiao, Zhao Xin,et al.An improved steady-and transient-state mixed non-intrusive load monitoring using viterbi algorithm[J].Transactions of China Electrotechnical Society,2023,38(19):5241‒5255.

[27]

刘宇,刘丛笑,赵欣,.基于维特比算法改进的稳暂态混合非侵入式负荷识别方法[J].电工技术学报,2023,38(19):5241‒5255.

[28]

Zhou Renfei, Tang Pengfei, Liu Sanfeng,et al.Research of non-intrusive load identification based on convolutional neural network[J].Information Technology and Network Security,2019,38(8):64‒68.

[29]

周任飞,汤鹏飞,刘三丰,.基于卷积神经网络的非侵入式负荷识别研究[J].信息技术与网络安全,2019,38(8):64‒68.

[30]

Liu Yanchi, Wang Xue, You Wei.Non-intrusive load monitoring by voltage‒current trajectory enabled transfer learning[J].IEEE Transactions on Smart Grid,2019,10(5):5609‒5619. doi:10.1109/tsg.2018.2888581

[31]

Zai Zhoupeng, Zhao Sheng, Zhu Xiang'ou,et al.Non-intru-sive load monitoring based on color coding and harmonic feature fusion[J].Electrical Engineering,2022,23(12):9‒16.

[32]

宰州鹏,赵升,朱翔鸥,.基于颜色编码与谐波特征融合的非侵入式负荷识别方法[J].电气技术,2022,23(12):9‒16.

[33]

Wang Ying, Yang Wei, Xiao Xianyong,et al.Non-intrusive residential load monitoring method based on refined identification of U‒I trajectory curve[J].Power System Technology,2021,45(10):4104‒4113.

[34]

汪颖,杨维,肖先勇,.基于U‒I轨迹曲线精细化识别的非侵入式居民负荷监测方法[J].电网技术,2021,45(10):4104‒4113.

[35]

Tabanelli E, Brunelli D, Acquaviva A,et al.Trimming feature extraction and inference for MCU-based edge NILM:A systematic approach[J].IEEE Transactions on Industrial Informatics,2022,18(2):943‒952. doi:10.1109/tii.2021.3078186

[36]

Qiu Xing, Yin Shihong, Zhang Zhihan,et al.Non-intrusive load identification method based on V‒I trajectory and high-order harmonic feature[J].Electric Power Engineering Te-chnology,2021,40(6):34‒42.

[37]

裘星,尹仕红,张之涵,.基于V-I轨迹与高次谐波特征的非侵入式负荷识别方法[J].电力工程技术,2021,40(6):34‒42.

[38]

Cui Haoyang, Cai Jie, Chen Lei,et al.Non-intrusive load fine-grained identification based on color encoding[J].Power System Technology,2022,46(4):1557‒1567.

[39]

崔昊杨,蔡杰,陈磊,.基于颜色编码的非侵入式负荷细粒度识别方法[J].电网技术,2022,46(4):1557‒1567.

[40]

Parra D, Escobar Sanabria D, Camargo C.A methodology and open-source tools to implement convolutional neural networks quantized with TensorFlow lite on FPGAs[J].Electronics,2023,12(20):4367. doi:10.3390/electronics12204367

[41]

Li Shuangfeng.TensorFlow lite:On-device machine learning framework[J].Journal of Computer Research and Development,2020,57(9):1839‒1853.

[42]

李双峰.TensorFlow Lite:端侧机器学习框架[J].计算机研究与发展,2020,57(9):1839‒1853.

[43]

Zhou Xiaokang, Liang Wei, Shimizu S,et al.Siamese neural network based few-shot learning for anomaly detection in industrial cyber-physical systems[J].IEEE Transactions on Industrial Informatics,2021,17(8):5790‒5798. doi:10.1109/tii.2020.3047675

[44]

de Baets L, Develder C, Dhaene T,et al.Detection of unidentified appliances in non-intrusive load monitoring using Siamese neural networks[J].International Journal of Electrical Power & Energy Systems,2019,104:645‒653. doi:10.1016/j.ijepes.2018.07.026

[45]

Lu Lingxia, Qiang Zhucheng, Yu Miao,et al.Design of contactless sign-in system based on embedded AI[J].Research and Exploration in Laboratory,2023,42(3):130‒134.

[46]

陆玲霞,强柱成,于淼,.基于嵌入式人工智能的无接触式签到系统设计[J].实验室研究与探索,2023,42(3):130‒134.

[47]

Zhang Yusen, Wu Hao, Ma Qing,et al.A learnable image-based load signature construction approach in NILM for appliances identification[J].IEEE Transactions on Smart Grid,2023,14(5):3841‒3849. doi:10.1109/tsg.2023.3239598

基金资助

国家自然科学基金面上项目(52077194)

浙江省“十四五”第二批本科省级教学改革备案项目(JGBA2024014)

教育部产学合作协同育人项目(2501270945)

浙江大学本科“AI赋能”示范课程建设项目(2024-24)

浙江大学AI For Education系列实证教学研究项目(202402)

AI Summary AI Mindmap
PDF (2685KB)

330

访问

0

被引

详细

导航
相关文章

AI思维导图

/