引用本文: | 张安安,庄景泰,郭红鼎,等.结合图半监督与广义回归神经网络的非侵入式海洋平台负荷监测[J].电力系统保护与控制,2020,48(7):85-91.[点击复制] |
ZHANG An’an,ZHUANG Jingtai,GUO Hongding,et al.Non-intrusive offshore platform load monitoring based on graph-based semi-supervised learning and generalized regression neural networks[J].Power System Protection and Control,2020,48(7):85-91[点击复制] |
|
摘要: |
海洋平台微电网所处环境复杂,对其自动化和智能化要求较高,目前缺少对其负荷实时智能监测和管理的方法。从非侵入式负荷监测的角度,考虑海洋平台的经济性要求和特殊的工业环境,提出结合图半监督与广义回归神经网络的非侵入式海洋平台负荷监测方法。采用图半监督学习算法自动标记训练数据集,减少了人工标记数据的工作量,使系统能自动完成数据标记。并与半监督聚类算法对比分析,表明图半监督学习算法对数据标记具有更高的正确率。再利用广义回归神经网络较强的非线性分类能力,提升负荷识别的识别精度和减少计算复杂度。Matlab/Simulink仿真结果表明,所提出的负荷识别算法不仅减少了人工干预而且具有高精度的识别率。 |
关键词: 海洋平台 数据标签 非侵入式负荷监测 图半监督学习 广义回归神经网络 |
DOI:10.19783/j.cnki.pspc.190561 |
投稿时间:2019-05-19修订日期:2019-08-22 |
基金项目:国家重点研发计划项目资助(2017YFE0112600);中国博士后基金项目资助(2014M562335);四川省科研创新团队研究项目资助(15TD0005) |
|
Non-intrusive offshore platform load monitoring based on graph-based semi-supervised learning and generalized regression neural networks |
ZHANG An’an,ZHUANG Jingtai,GUO Hongding,QU Guanglong,ZHOU Zhitong |
(Southwest Petroleum University, Chengdu 610500, China;XJ Times Technology Company, Xuchang 461000, China) |
Abstract: |
The offshore platform microgrid has a complex environment, it requires high automation and intelligence. At present, there is a lack of real-time intelligent monitoring and management methods for its load. From the perspective of Non-Intrusive Load Monitoring (NILM), considering the economic requirements and the special industrial environment of the offshore platform, this paper proposes a non-intrusive offshore platform load monitoring method based on Graph-Based Semi-Supervised Learning (GBSSL) algorithm and Generalized Regression Neural Networks (GRNN). GBSSL algorithm is used to automatically mark the training data set, which reduces the workload of manual labeling data and enables the system to automatically complete the data labeling. Compared with semi-supervised clustering algorithm, it shows that GBSSL has higher accuracy rate for labeling data. The strong nonlinear classification ability of generalized regressive neural networks is used to improve the recognition accuracy and reduce the computational complexity of load recognition. The simulation results of Matlab/Simulink show that the proposed load identification algorithm not only reduces manual intervention but also has high precision recognition rate. This work is supported by National Key Research and Development Program of China (No. 2017YFE0112600), China Postdoctoral Fund (No. 2014M562335), and Research Program of Sichuan Scientific Research Team (No. 15TD0005). |
Key words: offshore platform data label non-intrusive load monitoring graph-based semi-supervised generalized regression neural network |