Research On Non-intrusive Load Disaggregation Considering Signals Of Energy Consumption Based On Low Frequency Sampling

Due to the rapid development of smart grids,especially the popularity of smart meters,it is convenient to obtain the operation status and energy consumption of the electrical appliances in user's home from the data collected by smart meters.It is commonly known as non-intrusive load disaggregation.At present,most researchers realize NILD based on the load data obtained under high-frequency sampling,while the sampling frequency of the smart meter appled in daily life is low.It is difficult to analyze the power signals collected by the smart meters directly,for the information of the power signals is limited.In addition,the signals of energy consumption collected by smart meters has not been fully exploited and utilized.The electricity consumption signals are introduced into the NILD problem.This thesis is mainly focusing on how to use the electricity consumption signal to realize the electrical appliance operation identification.The main research contents are as follows:1.In this thesis,the active power value of load equipment in stable operation is taken as load characteristics.The load characteristics of several typical household load equipment are extracted by affinity propagation algorithm based on the relationship between active power of each appliance and time.And then the load feature library(i.e.the load state table referred in this paper)is established.2.In this thesis,the trend of total active power curve in the sampling interval is estimated based on the data of energy consumption.Taking the simplest case as an example,the method of reconstructing a curve to connect the sampling points is given.The simulation results show that the reconstructed curve is closer to the actual curve than the sampling curve with zero-order holder.3.Compared with the traditional load disaggregation optimization model,the algorithm proposed in this paper considering the energy consumption signal,the operation characteristics of electrical appliances and people's behavior characteristics.By relaxing some constraints and linearizing the objective function,the whole model is transformed into mixed integer linear programing model.4.In this thesis,recall rate,precision rate and F1 value are introduced to compare the performance of the three algorithms in identifying operation status of various devices in relatively complex scenarios.The simulation results show that the proposed algorithm has high recognition accuracy.Even in more complex scenarios,the performance of this algorithm is generally stable.Normally,the detection of on-off action is timelier,and the deviation between the identified start-stop time and the actual time is small.Finally,taking the simulation experiment as an example,the computational efficiency and the applicable scenarios of the algorithm proposed in this paper is briefly analyzed.The above analysis has practical significance in some degree for exploring how to study the non-intrusive load disaggregation algorithms using the data obtained by smart meters in China.