Research On Non-intrusive Load Disaggregation Method Based On Electrical Operation Characteristics And Time Probability Distribution

Non-intrusive load disaggregation is to identify the operation status and energy consumption of each household appliance from the aggregate load readings collected at users' power entrance.The data obtained by non-intrusive load disaggregation technology is helpful to figure out the pattern of users' power consumption behavior,which is of great significance to users and power grid.Most of the research on NILD at home and abroad is based on the data sampled at high-frequency.However,the sampling interval of smart meters commonly used in china is around 10 minute,which is quite different from the sampling frequency required in most NILD research.It is hard to realize load disaggregation with high precise based on data obtained at the intervals of 10 minute since the information of the data is limited.Besides,most of the existing NILD researches regard a set contained all the household appliance in user's home as a candidate set.And each appliance in this set might be active at each time stamp.In light of this,the algorithm would be less efficient when the number of appliance is big.In order to optimize the candidate set,a method can preliminarily analyze the types of electrical appliance,probably be active during a given period,is proposed in this thesis.The method is based on the time characteristics appeared when different appliance be active at the same time.In addition,in order to overcome the limited information of the data sampled at the interval of 10 minute,the operation characteristics such as operating state conversion process and energy consumption are introduced in this paper.And it apparently could provide more information for NILD.The main research contents are shown as follows:1.Based on time probability distribution,a preliminary load disaggregation algorithm that could preliminarily analyze the types of electrical appliance,probably be active during a given period,is proposed in this thesis.The procedure of this method is as follows: the aggregate active power sequence is segmented into a series of load subsequences according to its variation trend;The time probability distribution characteristics of single electrical equipment and electrical combination are analyzed in this paper,and the types of appliance could be active in each subsequence can be preliminarily analyzed based on these time probability distribution characteristics.The simulation results show that the preliminary load disaggregation algorithm can accurately identify the types of appliance each subsequence contained,which could reduce the computational complexity of load disaggregation.2.The state transition sequence and energy consumption sequence of electrical appliances are extracted,and the operation characteristic library of electrical appliances is established in this thesis.Different from the current method extracting the state transition characteristics by using a large amount of training data,this paper proposes a method which can extract electrical state transition sequences based on typical periodic curve of electrical appliances.In addition,the energy consumption information was introduced to enrich the information of the sampling data.In order to improve the accuracy of load disaggregation,the energy consumption sequence corresponding to the state time series is established to match the aggregate household energy consumption.3.Further load disaggregation method based on electric appliance operating characteristics is proposed in this thesis.The further load disaggregation model consists of two optimization models.1)Considering the time correlation between operation states in the state transition sequence and some prior knowledge,an optimization model based on the state transition sequence is established.2)The time correlation between energy consumptions in energy consumption sequence are considered to construct the other optimization model.The specific electrical appliances and their start-stop time can be identified by solving the optimization models above.The simulation results show that comparing with the traditional auxiliary integer linear programming algorithm,the algorithm proposed in this paper has less misjudgment about the operation of electrical appliances.