Research On Non-intrusive Resident Load Monitoring Methods

Smart grid is the ultimate form of future power grid.Compared with traditional power grid,it is based on the highly developed communication network and can realize the Two-way circulation of power grid information through developed sensing technology,measurement technology,decision-making technology and control technology.The traditional invasive load monitoring technology needs to install a large number of sensor acquisition equipment which needs huge installation and maintenance costs.Besides,it is difficult to deal afterwards.Therefore,researchers have proposed the non-intrusive load monitoring(NILM)method.NILM is generally divided into four steps: data acquisition and processing,detection of switching events,load signature extraction and load identification.In this paper,a NILM method is proposed based on wavelet transform and energy ratio algorithm.The specific work is as follows:(1)Some home appliances models are built in MATLAB/SIMULINK to simulate the operation of residential quarters,such as fridge,air-conditioner,calorifier and power electronic equipment,and then collect the voltage,current and active power in each household appliance and the Integrated operation scene.(2)Energy ratio algorithm(ERA)is selected for event detection.According to the energy ratio algorithm and the variation of active power,the current transient process is separated.The results show that it is successful to determine the accuracy switching time and separate the transient process based on ERA.(3)Multi-scale decomposition of wavelet transform is used to extract the transient process.Wavelet energy spectrum and wavelet entropy are selected as load characteristics which used the principal component analysis to reduce dimension.The results show that different kinds of loads have variable feature vector.(4)According to load characteristics,the deep belief network(DBN)is used to identify the comprehensive operation of multi-family households,including two RBM and BP.The results show that the identification accuracy is above 95%,and it can identify multiple household appliances at the same time.