Household Electricity User Behavior Analysis Based On Smart Meter Data

Non intrusive household power load monitoring is to identify the total power change at the entrance of the home power,and analyze the energy consumed by a single device.In this paper,a new algorithm based on convolutional neural network is proposed to identify the power consumption of users.The theoretical part of this paper firstly introduces the background and significance of the non intrusive load monitoring,and introduces the four main steps of the non-invasive load monitoring algorithm flow.Through the brief introduction of the theory and structure of convolutional neural networks and hidden Markov models of these two kinds of classification models,identification and classification of two kinds of classification model and non intrusive load monitoring combined with the experimental analysis and comparison of two kinds of algorithm of intrusive behavior recognition ability type load monitoring in non experimental analysis.The experimental part,the original signal preprocessing the data,through the analysis of equipment selection,filtering,state labels and short-time Fourier transform to obtain frequency domain signal,and combining the signal in time domain and frequency domain signal,finally get the convolutional neural network need to time-frequency image.By debugging experiment parameters on the final classification model,the selection of parameters at the same time,to meet the needs of low error rate and small computation time,by controlling the variables of the iteration parameters,the learning rate,the size and number of convolution convolution kernel parameter adjustment,select the optimal parameters of the convolutional neural network experiment for unit family electricity behavior recognition performance analysis.Finally,the optimal convolutional neural network model is compared with the common hidden Markov model identification results of non intrusive algorithm,and its advantages are summarized.