针对传统非侵入式负荷分解算法准确率低、计算较耗时等问题,在隐马尔科夫模型（HMM）的基础上提出基于行为影响因子的负荷实时分解算法。使用自适应的迭代K-means聚类方法提取负荷状态,并将负荷状态组合成超状态。针对传统HMM没有考虑用电场景时间特性的缺陷,对参数进行时间分段学习。在分解阶段引入用户用电行为模式的影响因子,改进隐马尔科夫齐次假设,并利用维特比算法分解出用户的各个负荷的实时状态。通过公开数据集验证了所提算法的准确性和实时性。

Aiming at the problems of low accuracy, time-consuming, etc., of traditional nonintrusive load disaggregation algorithms, a usage influencing factor based real-time load disaggregation algorithm is proposed based on HMM(Hidden Markov Model). The self-adaptive iterative K-means clustering method is used to extract load states, and the load states are combined into super-states. Aiming at the shortage that traditional HMM does not consider the time characteristic of electricity consumption scenes, the parameters are learned in time segments. At the disaggregation stage, the influencing factor of users' electricity consumption pattern is introduced to improve implicit Markov homogeneous hypothesis, and the real-time states of each load of users are decomposed by Viterbi algorithm. The accuracy and real-time performance of the proposed algorithm are verified through public data sets.