Comprehensive Study On Nonintrusive Load Decomposition Algorithm

With the rapid development of science and technology,the smart grid becomes the future direction of the power system.At the same time,the shortage of energy also restricts the economic and social development of our country.Therefore,it is particularly important to reduce the energy consumption and monitor the operation status of various loads in the power system.At present,most of the means of load decomposition are intrusive decomposition methods,namely,for the system to be decomposed,a measuring device is installed at each load in the system to collect the electrical quantity of each load so as to realize the monitoring of the operating status of each load in the system,and a great deal of manpower,material and financial resources are required in the aspects of purchasing,operation and maintenance,and which poses great difficulties for later load handling and data analysis.In response to these problems,people put forward a non-intrusive means of load decomposition.Non-intrusive load decomposition refers to the need to monitor the electrical power at the terminal of the power system without entering the system,so as to monitor the operating status of each load in the system.However,for now,non-intrusive load decomposition is still in its development stage,and complete system and mature algorithms have been not formed yet.Based on this,this paper respectively studies the steady-state process and transient process of power system comprehensively,improves the decomposition method and decomposition process of each load in the system,and puts forward a more complete algorithm to achieve the load decomposition.The main achievements are as follows:According to the characteristics of the collected data,two objective functions are proposed for the steady-state process of the power system.The objective of the two objective functions is optimized respectively to achieve the purpose of load decomposition.The actual measured cases are used to verify the objective function.The results show that the algorithm achieves the purpose of monitoring the load in the steady state,and then the paper combines the two objective functions to make up for the possible error of a single objective function,and which verified by the examples.Secondly,for the transient process of the power system,the algorithm of transient event monitoring in the system is put forward.After that,the steady-state load current data of the load causing the transient process are extracted by waveform extraction.With the aid of the cross-correlation function,the type of the load causing the switching process is recognized.The goal of monitoring the load status during the transient process is achieved.Finally,the cross-correlation analysis between signals can only monitor the load with large differences in current waveforms,and further exploration is carried out.The main research can uniquely identify the transient information of load transient process,select a series of characteristic quantities to identify each transient process,and use the decision tree algorithm to realize the load decomposition in transient process.Afterwards,in order to solve the possible misjudgment problem caused by a single decision tree,the paper further studies the characteristics of the data and proposes multi-layer decision tree solution.Compared with the single decision tree,the correct rate of load decomposition is improved.The load decomposition results of steady-state process and the transient process complement each other,common decision-making,which is more perfect to achieve the task of power system load decomposition.