Algorithm-based Non-invasive HSMM Home Power Load Decomposition

Non-intrusive house load disaggregation algorithm is the process ofdisaggregating total electricity consumption into its contributing appliances. Inthe paper, we proposed a disaggregation algorithm basing on hidden-semiMarkov model and indentified the load operation state using power amplitudechanges. In the end, we verified the effectiveness of proposed algorithmcombining two assessment metrics and REDD dataset.Firstly, in order to eliminate the abnormal data in observation sequence, thedata pre-processing is necessary. It contains removing the always-on devicecontent, filtering analysis, calculating the differential waveform. Featureobservation sequence is obtained. The paper sets REDD dataset as experimentdata. We can obtain the power differential waveform by preprocessing the sixhouse power data of REDD. Finally, in order to the comparative of load power,we used K-means clustering algorithm to classify house electric load accordingto power amplitude range.Secondly, in order to solve the limitation of hidden Markov model (HMM),hidden semi-Markov model is proposed for modeling house electric load.HSMM improve the accuracy of state identification, the state transition ofappliance not only relates to the current state but also the state duration. Thenwe can obtain the best state transition sequence using improved Viterbialgorithm. The power of single house electric load can be computed accordingto the means of each appliance.Finally, we proposed two assessment metrics, total disaggregation error andpercent error of single device from the overall and local. The experiment resultsshow that HSMM provides better modeling and analysis ability. And theproposed improved method has better identification effect and better precision.