基于多特征序列融合的负荷辨识方法

发布时间：2018-03-22 11:29

本文选题：负荷辨识　切入点：整数规划　出处：《电力系统自动化》2017年22期 　论文类型：期刊论文

【摘要】：针对当前利用低频采样实现非侵入式负荷辨识存在的准确率低的问题,提出了基于多特征序列融合的负荷辨识方法。该方法首先建立负荷存在可能性的整数规划模型,进行初辨识以降低负荷辨识的维度。然后,根据滑动窗口算法获得组合功率序列和原始功率序列,提取其统计特征和奇异值特征,进而利用概率神经网络获得隐马尔可夫模型的观测值序列。同时,利用隐马尔可夫模型对负荷序列信息进行融合,计算观测序列和组合功率序列之间的相似度,从而完成在低频采样下的负荷辨识,并获得各个家用负荷的耗电量。最后,通过单负荷辨识、多负荷辨识、不同采样率辨识和各居民用户负荷辨识的仿真实验,得到负荷准确率和辨识精度的平均值均在85%以上,证明了所提算法的合理性和即时性能够达到在低频采样下负荷的辨识要求。
[Abstract]:In order to solve the problem of low accuracy of non-invasive load identification using low-frequency sampling, a load identification method based on multi-feature sequence fusion is proposed. Firstly, an integer programming model of the possibility of load existence is established. In order to reduce the dimension of load identification, the combined power sequence and the original power sequence are obtained according to the sliding window algorithm, and the statistical features and singular value characteristics are extracted. At the same time, the hidden Markov model is used to fuse the information of the load sequence, and the similarity between the observation sequence and the combined power sequence is calculated. In order to complete the load identification under the low-frequency sampling, and obtain the power consumption of each household load. Finally, through the single load identification, the multi-load identification, the different sampling rate identification and the residents user load identification simulation experiment, The average value of load accuracy and identification accuracy is more than 85%, which proves that the proposed algorithm is reasonable and real-time to meet the requirements of load identification under low-frequency sampling.
【作者单位】：东北大学信息科学与工程学院;国网葫芦岛供电公司;国网盘锦供电公司;
【基金】：国家自然科学基金资助项目(61273029) 中央高校基本科研业务费专项资金资助项目(N160402003) 新能源电力系统国家重点实验室立项资助项目(LAPS17013)~~
【分类号】：TM714

【相似文献】