Research On Load Monitoring Method Of Non Intrusive Household Appliances Based On Transient Process

If you can monitor the load of residents,so that electricity bills can be the same as the telephone bill shows the power consumption of each electrical appliances,then the energy-saving emission reduction,environmental pollution and other aspects of significance is extraordinary.Traditional intrusion monitoring methods have the problems of high cost,complicated data processing and difficult operation and maintenance of equipment.In recent years,with the technological innovation,non-invasive load monitoring system came into being.This system only needs to install the information collection device at the general distribution line of the residents,which reduces the cost and maintenance workload.Firstly,this paper uses Simulink to build a model to simulate a variety of electrical equipment in home,selecting the most representative of the four electrical appliances:refrigerators,air conditioners,water heaters and power electronic devices(such as:TV,charging lines,etc.)After the simulation model is completed,the energy ratio algorithm is used to estimate the moment of load switching,that is,the use of active fluctuations to explore whether the electrical input or removal.Then,by using the wavelet analysis technique,the current information is decomposed to extract the characteristic parameters of the load.Finally,based on the algorithm of depth belief network,we study the load characteristic pattern of each electrical equipment to achieve the ultimate goal of identifying the load.The main contents and conclusions of this paper are as follows:(1)Matlab software is used to simulate the power consumption of families.Electric motors are used to simulate refrigerators,resistors are used to simulate electric lights and water heaters,and induction motors,or asynchronous motors,are used to replace air conditioning.So it supports the follow-up experiment to simulate the typical single and combined load operation events,analyze their characteristics,establish the corresponding load characteristics database,verify the practical application effect of the typical algorithm,and analyze the advantages and disadvantages.(2)energy ratio algorithm is used for event detection.Because the energy ratio algorithm has the advantages of easy to understand,superior stability,convenient parameter selection and so on,it is more reasonable and applicable as the research method of this subject.(3)use wavelet decomposition to extract features.This method is now more mature,research papers are also many,for the future application of high value,so it is worth learning.Transient features are expanded on the frequency scale,and some representative frequency waveforms with high identification are excavated or combined into samples for feature selection.(4)pattern recognition is used as a method of load identification module.Compared with mathematical optimization,the method of pattern recognition is more intelligent.It achieves the goal of load identification by learning load characteristic patterns of electrical equipment.This paper mainly adopts the pattern recognition method based on deep belief network DBN,but this method can only analyze the situation that the type of load involved is not many,the state involved is relatively simple,and the pattern recognition method is too complex to study under complex circumstances.In this paper,the non-intrusive load monitoring system is divided into four blocks,which are:Simulink simulation for data acquisition;Energy ratio algorithm for event detection;Wavelet decomposition for feature extraction and depth belief networks for load recognition.For the data acquisition part,the resident load is divided into three types:resistor type,electric type and electric electronic type load.Each household assumes a refrigerator,two air conditioners,a water heater and an electrical electronic device.And three families are a group of load units,A,B,C three phases with three load units.For the event detection part,this paper uses the energy ratio algorithm to judge the timing of the electric appliance by the change of the power energy.This algorithm is very reliable and simple to use,because when a certain point in time,when the electric appliance changes,it will inevitably be accompanied by active changes,which will inevitably lead to changes in the power.On the basis of the original research,the feature parameter extraction is improved.The square sum of wavelet coefficients in 0.5 S or 1 S is defined as wavelet energy,and it is used as the characteristic parameter of load.Finally,using the depth belief network,the third layer of the fourth layer wavelet coefficient after the wavelet transform is used as the input layer,and the custom five electrical function values are used as the output layer to train,carry out the final load recognition,and then set a set of known data to detect.,judge the practicality of the algorithm.