非侵入式电器识别算法的研究
发布时间：2018-07-25 19:40
【摘要】：对于电力系统的智能化发展,负荷监测具有非常重要的意义。传统的负荷监测方法一般是在每个负荷配电输出端,安装传感器等监测设备,这种侵入式的负荷监测方法在安装和维护方面需要大量的时间和金钱,且硬件维护成本较高。因此,研究人员提出非侵入式负荷监测(NILM)方式,只需要在电力入口处安装监测设备,通过监测人口处的电压、电流等信号就可以分解得到系统内单个负荷类别和运行情况。对于能源提供者来说,NILM有助于电力提供方了解用户的负荷构成,用电习惯和能源使用情况,加强负荷用电的监测和管理,合理安排负荷的使用时间,调节峰谷差、降低输电损耗等；单从技术本身考虑,有助于改善电力负荷的预测精度,为负荷监测的仿真分析、系统规划提供更准确的数据；对于电力用户来说,通过NILM可以对负荷能耗数据进行有效的分析,减少不必要的能源消耗,达到节能降耗的目的。家用电器用电情况在线监测是在智能电表中加入非侵入式家用电器用电监测模块,为满足在线用电管理提供有效且全面的数据支持。本文从三个方面进行非侵入式负荷识别的简单研究,首先根据空调负荷在夏季是家用负荷用电的主要耗能元件,基于k-means算法的改进应用于空调负荷的分解,使用边缘检测和k-means聚类方法将数据进行分类,利用数据确定空调行为的关键参数,这个参数用于确认空调的启停事件；其次提取负荷电流参数,选用电流最大值、平均值和均方差作为负荷识别特征参数,进行简单的识别。负载启动瞬态电流波形可以被获取到,激励瞬态特性的几个数值提取自获取到的与三个特性参数相关的瞬态电流波形,提取到瞬态特性参数,将其进行训练完善,标识为负荷识别特征参数,进而进行仿真验证识别效果；最后,根据提取到的电流、电压波形,计算负荷的多特征参数,加权赋值法来完成负荷类型的匹配,选择用电负荷仿真,将实验数据代入识别算法,验证算法的准确性与可应用性。具体工作如下：(1)首先检测到负荷的启停,根据电流波形的差分,获得投切负荷的波形图,之后对每个电流周期强度进行差分运算,得到总的瞬态时间,进而提取到该时间段内负荷电流的最大值、平均值和均方差作为负荷识别的参数设定。提取多负荷的这三个瞬态识别参数,进而可仿真验证算法准确性。(2)研究家用电器的稳态和暂态特征,提取家用电器的多特征参数。以16种家用电器作为参照设备进行实验,采样稳态运行的电压、电流波形数据,计算其多特征参数,建立特征参数模型库作为电器类型辨识数据库。(3)提出家用电器类型辨识算法。选取参照电器以外的某种电器进行仿真识别,将电压电流波形数据带入辨识过程进行计算分析,结果证明该辨识算法的正确性。选取两个家用电器做混合类型识别实验,利用上述方式进行分析。结果证明提出的辨识算法可以成功辨识多个设备同时在线运行情况。
[Abstract]:The load monitoring is very important for the intelligent development of the power system. The traditional load monitoring method is usually in the output end of each load, the installation of sensors and other monitoring equipment. This intrusion detection method needs a lot of time and money in installation and maintenance, and the cost of hardware maintenance is high. In this case, the researchers propose a non intrusive load monitoring (NILM) method, which only needs to install monitoring equipment at the power entrance. By monitoring the voltage of the population, the current and other signals can decompose the single load category and operation in the system. For the energy provider, NILM helps the power provider to understand the user's load composition. With the use of electricity and energy, the monitoring and management of load power is strengthened, the use time of load is reasonably arranged, peak and valley difference is adjusted, transmission loss is reduced, and the prediction accuracy of power load is improved by the single technology itself, and more accurate data for the simulation analysis of load monitoring, and more accurate data for the system planning; For the household, the data of the load energy consumption can be analyzed effectively by NILM to reduce the unnecessary energy consumption and achieve the purpose of saving energy and reducing consumption. The on-line monitoring of household electrical appliances is to add non intrusive household electrical monitoring module to the intelligent meter to provide effective and comprehensive data support for line management. In this paper, a simple study of non intrusive load identification is carried out from three aspects. First, according to the air conditioning load in summer is the main energy consumption component of the household load, based on the improvement of the k-means algorithm, the air conditioning load is decomposed. The data are classified by the edge detection and K-means clustering method, and the air conditioning behavior is determined by the data. The key parameter, this parameter is used to confirm the start and stop event of the air conditioning. Secondly, the load current parameters are extracted, the maximum current value, the mean value and the mean square deviation are selected as the characteristic parameters of the load identification. The load starting transient current waveform can be obtained, and several values of the transient characteristics are extracted from the acquisition to three. The transient characteristic parameters related to the characteristic parameters are extracted and the transient characteristic parameters are extracted. They are trained and improved. The characteristic parameters of the load identification are identified, and then the simulation verification and recognition results are carried out. Finally, according to the extracted current, voltage waveform, the multiple characteristic parameters of the load, the weighted assignment method is used to match the load type, and the selection of the load type is completed. Using the electrical load simulation, the experimental data are replaced by the recognition algorithm to verify the accuracy and applicability of the algorithm. The specific work is as follows: (1) first, the load starts and stops are detected, and the waveform diagram of the load is obtained according to the difference of the current waveform, then the difference calculation is carried out for each current cycle strength, and the total transient time is obtained, then the extraction is obtained. The maximum value, average value and mean square of load current in this period are set as parameters of load identification. The three transient identification parameters of multi load are extracted, and then the accuracy of the algorithm can be verified. (2) study the steady and transient characteristics of household electrical appliances and extract the multiple characteristic parameters of household appliances. 16 kinds of household appliances are used as reference equipment. In the experiment, the voltage and current waveform data are sampled in the steady state, and the multiple characteristic parameters are calculated. The characteristic parameter model library is set up as the identification database of the electrical type. (3) the identification algorithm of the household electrical type is proposed. The results prove the correctness of the algorithm. Two household appliances are selected to do the hybrid type identification experiment, and the above method is used to analyze. The results show that the proposed identification algorithm can identify the on-line operation of multiple devices at the same time.
【学位授予单位】：中国海洋大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TM715
本文编号：2144869
[Abstract]:The load monitoring is very important for the intelligent development of the power system. The traditional load monitoring method is usually in the output end of each load, the installation of sensors and other monitoring equipment. This intrusion detection method needs a lot of time and money in installation and maintenance, and the cost of hardware maintenance is high. In this case, the researchers propose a non intrusive load monitoring (NILM) method, which only needs to install monitoring equipment at the power entrance. By monitoring the voltage of the population, the current and other signals can decompose the single load category and operation in the system. For the energy provider, NILM helps the power provider to understand the user's load composition. With the use of electricity and energy, the monitoring and management of load power is strengthened, the use time of load is reasonably arranged, peak and valley difference is adjusted, transmission loss is reduced, and the prediction accuracy of power load is improved by the single technology itself, and more accurate data for the simulation analysis of load monitoring, and more accurate data for the system planning; For the household, the data of the load energy consumption can be analyzed effectively by NILM to reduce the unnecessary energy consumption and achieve the purpose of saving energy and reducing consumption. The on-line monitoring of household electrical appliances is to add non intrusive household electrical monitoring module to the intelligent meter to provide effective and comprehensive data support for line management. In this paper, a simple study of non intrusive load identification is carried out from three aspects. First, according to the air conditioning load in summer is the main energy consumption component of the household load, based on the improvement of the k-means algorithm, the air conditioning load is decomposed. The data are classified by the edge detection and K-means clustering method, and the air conditioning behavior is determined by the data. The key parameter, this parameter is used to confirm the start and stop event of the air conditioning. Secondly, the load current parameters are extracted, the maximum current value, the mean value and the mean square deviation are selected as the characteristic parameters of the load identification. The load starting transient current waveform can be obtained, and several values of the transient characteristics are extracted from the acquisition to three. The transient characteristic parameters related to the characteristic parameters are extracted and the transient characteristic parameters are extracted. They are trained and improved. The characteristic parameters of the load identification are identified, and then the simulation verification and recognition results are carried out. Finally, according to the extracted current, voltage waveform, the multiple characteristic parameters of the load, the weighted assignment method is used to match the load type, and the selection of the load type is completed. Using the electrical load simulation, the experimental data are replaced by the recognition algorithm to verify the accuracy and applicability of the algorithm. The specific work is as follows: (1) first, the load starts and stops are detected, and the waveform diagram of the load is obtained according to the difference of the current waveform, then the difference calculation is carried out for each current cycle strength, and the total transient time is obtained, then the extraction is obtained. The maximum value, average value and mean square of load current in this period are set as parameters of load identification. The three transient identification parameters of multi load are extracted, and then the accuracy of the algorithm can be verified. (2) study the steady and transient characteristics of household electrical appliances and extract the multiple characteristic parameters of household appliances. 16 kinds of household appliances are used as reference equipment. In the experiment, the voltage and current waveform data are sampled in the steady state, and the multiple characteristic parameters are calculated. The characteristic parameter model library is set up as the identification database of the electrical type. (3) the identification algorithm of the household electrical type is proposed. The results prove the correctness of the algorithm. Two household appliances are selected to do the hybrid type identification experiment, and the above method is used to analyze. The results show that the proposed identification algorithm can identify the on-line operation of multiple devices at the same time.
【学位授予单位】：中国海洋大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TM715
【参考文献】
相关期刊论文 前3条
1 牟龙华;朱国锋;朱吉然;;基于智能电网的智能用户端设计[J];电力系统保护与控制;2010年21期
2 张钦;王锡凡;付敏;王建学;;需求响应视角下的智能电网[J];电力系统自动化;2009年17期
3 张文亮;刘壮志;王明俊;杨旭升;;智能电网的研究进展及发展趋势[J];电网技术;2009年13期
，本文编号：2144869
本文链接：https://www.wllwen.com/kejilunwen/dianlilw/2144869.html
