Signature Analysis Of Non-intrusive Load Identification

With the rapid development and large-scale grid connection of renewable energy,the stable operation and dispatching of power grids are facing enormous challenges.Improving the level of demand side response is a reliable way to promote the consumption of renewable energy.The non-intrusive appliance load monitoring(NALM)of the user's power consumption behavior in the smart grid is an important part of realizing the demand side response with the home power user as the main body.Non-intrusive load identification is the core of non-intrusive load monitoring.Improving the accuracy of non-intrusive load identification is beneficial to the development and promotion of non-intrusive appliance load monitoring technology,and it has important theoretical and practical significance for the stable and economic operation of power grid and for the economic and safe use of electricity by users.Aiming at the current problems existing in non-intrusive load identification cases,such as the weak basis of the selected signatures,the lack of interval determination research on transient signatures analysis,the lack of close combination of identification algorithm and selected signatures,resulting in low identification accuracy.Focusing on the analysis of load signatures,the interval determination method of transient signature analysis,the selection method of load signatures,and the combination of signatures and identification algorithms based on the high-frequency sampling signal with "high noise and high resolution" signatures,and the signatures with "high-dimensional and small-sample" signatures.The main work of this paper is as follows:Firstly,the signatures that can be used for non-intrusive load identification are discussed in depth,the calculation methods and advantages and disadvantages of each signature are summarized,and the concept of virtual transient process duration is defined for the complex situation of transient process duration.By comparing and analyzing the identification effect of each signature used in non-intrusive load identification with an example,the shortcoming of using only one signature to identify non-intrusive load is pointed out,which provides an optional signature for signature selection and load identification research in the future.Secondly,in order to extract the transient signatures accurately,an interval determination method of transient signature analysis based on the combination of permutation entropy algorithm and Yamamoto algorithm is proposed.The time window of load state change is determined by the jump of multi-scale permutation entropy value,and then the time of state change is located according to the difference of signal-to-noise ratio of each sampling point in the time window,so that the interval of transient signature analysis can be determined.The fusion method can quickly and accurately determine the interval of transient signature analysis,and reduce the influence of errors in transient signature extraction on subsequent non-intrusive load identification results.Then,in order to obtain a high quality and non-redundant subset of load signatures,an improved Relief-F and mutual information hybrid signature selection method is proposed to reduce the dimension of all signature sets composed of optional signatures.Firstly,the signature subset is selected based on Relief-F feature importance sorting algorithm,and then the signature subset is redundantly processed by mutual information between signatures.Finally,the signature subset and the corresponding signature weight value are obtained after the selection.The selection of high-dimensional signature data can effectively save the storage space required for non-intrusive load monitoring,while ensuring the accuracy of subsequent load identification,greatly reducing the amount of calculation.Finally,a non-intrusive load identification algorithm based on fuzzy C-means clustering algorithm combined with signature weights is deduced and constructed,and three non-intrusive load identification indicators are given,i.e.load group identification accuracy,load single-class identification accuracy and load random identification accuracy.The identification algorithm is used to analyze the influence of interval error,signature selection or signature redundancy on load identification results.The results show that it is necessary for non-intrusive load identification to determine the analysis interval accurately,and it is also necessary to carry out the signature selection and de-duplication processing.Besides,the proposed algorithm combines the fuzzy C-means clustering algorithm with feature weights to achieve good identification results.In a word,the analysis and research of load signatures is the basis of non-intrusive load identification.The comprehensive understanding,accurate extraction,reasonable selection of load signatures and effective combination with identification algorithm are helpful to improve the accuracy of non-intrusive load identification.This topic provides a new idea for the research of non-intrusive load identification,which is in line with the development trend of load monitoring technology.