Reaserch On Current Feature Extraction Algorithms For Nonivasive Load Monitoring

Today,science and technology are changing with each passing day,the energy has become a hot issue attracting the world-wide attention.In the United States,the power consumption of commercial and residential industry accounts for 70% of the total electricity consumption.The electricity load,especially all kinds of electrical appliances,occupies a large part of building energy consumption,and the proportion is likely to increase in the future.Extensive use of all kinds of load makes the harmonic pollution,the electricity disturbance and safety problems more prominent.And the accelerated aging of equipment and the potential risk of overvoltage and overcurrent makes the control of grid more and more complex,the hidden danger of the fire is also can not be ignored.So it put forward higher requirements for user behavior analysis of the electricity grid and load identification and monitoring to solve these problems.In this paper,based on the insight into the the research status of load feature extraction and recognition at home and abroad,in order to explore the methods should be taken in the process of extracting different load characteristics,we have made some progress in the following several aspects of research.First of al,on the basis of traditional noninvasive load monitoring(NILM),we designed the structure of the load identification system and collected relevant data.By modeling the load identification system,we distinguished the function of each module and ways of realization.Then,we collected the current data from resistive loads as well as capacitive loads in different operation conditions.Secondly,we focuses on the method of load transient features extraction.The technology of change–points detection is introduced into the recognition process of load transient features,by using the method based on dynamic analysis of similarity,we identified the change-points result from transient switching of load and then calculated the energy of the transient current as transient characteristics in load identification.By dealing with the blending load current data in the matlab experiment,we have proved that the recognition accuracy of the proposed algorithm can achieve 98% above.Finally,we have studied the process and methods of wavelet analysis and fractal theory in the extraction of load steady characteristics.By using the wavelet packet decomposition to analyze the load current signal,we selected an appropriate frequency band as a steady characteristics to identify the load.The fractal theory was applied to extract the characteristics of the load current,by using GP algorithm,we calculated the correlation dimension of different load,explored the feasibility of making correlation dimension as a steady current feature to identify of the load.