| 引用本文: | 刘凯,符玲,杨金刚,熊思宇,蒿保龙,刘丽娜.基于简化HMM和时间分段的非侵入式负荷分解算法[J].电力自动化设备,2024,44(2):198-203,210. |
| LIU Kai,FU Ling,YANG Jingang,XIONG Siyu,HAO Baolong,LIU Lina.Non-intrusive load decomposition algorithm based on simplified HMM and time segmentation[J].Electric Power Automation Equipment,2024,44(2):198-203,210. |
|
| 摘要: |
| 针对现有非侵入式负荷分解算法需要以过去时刻的分解结果为依据，从而造成误差累积的问题，提出一种基于简化的隐马尔可夫模型和时间分段的非侵入式负荷分解算法，以实现居民家庭的负荷分解。对负荷的低频功率信号进行分层抽样和聚类分析，构建负荷功率模板并利用独热码对超状态进行编码表示。基于简化的隐马尔可夫模型和普遍生活规律对家庭用电时间段进行划分，在每个时间段内单独训练参数。结合总线数据和各时间段参数实现对各时刻负荷功率的独立求解。基于2种国外公开数据集的测试结果验证了所提算法的准确性和实时性。 |
| 关键词: 负荷分解 隐马尔可夫模型 亲和力传播聚类 时间分段 超状态 |
| DOI：10.16081/j.epae.202304010 |
| 分类号: |
| 基金项目:四川省科技计划项目(2021YFG0294) |
|
| Non-intrusive load decomposition algorithm based on simplified HMM and time segmentation |
|
LIU Kai1, FU Ling2, YANG Jingang1, XIONG Siyu2, HAO Baolong2, LIU Lina3
|
|
1.Tangshan Institute, Southwest Jiaotong University, Tangshan 063000, China;2.College of Electrical Engineering, Southwest Jiaotong University, Chengdu 611756, China;3.Metering Center of State Grid Sichuan Electric Power Corporation, Chengdu 610045, China
|
| Abstract: |
| Aiming at the problem that the current non-intrusive load decomposition algorithms need to be founded on the decomposition results of past moments, which causes the error accumulation, a non-intrusive load decomposition algorithm based on a simplified hidden Markov model(HMM) and time segmentation is proposed to realize the load decomposition of residential households. The stratified sampling and cluster analysis are carried out for the low-frequency power data of the load, the power template of the load is constructed and the super-state is coded and represented by the one-hot code. The time segments of household electricity consumption are divided based on the simplified HMM and universal life pattern, and the parameters are trained separately within each time segment. The load power at each moment is independently solved by combining the bus data and the parameters of each time segment. The accuracy and real-time performance of the proposed algorithm are verified by the test results based on two foreign public datasets. |
| Key words: load decomposition hidden Markov model affinity propagation clustering time segmentation super-state |
