Non-intrusive Load Identification Of Resident Users Based On Unsupervised Learning

As the construction of a strong smart grid and the ubiquitous power Io T advances,the demand side management of the grid is becoming more and more important.At the same time,the global energy crisis and environmental degradation are becoming more and more serious,and mankind is in urgent need of taking necessary measures.Resident users are an important part of power users,and their energy consumption characteristics are more flexible than industrial and commercial users.Enhancing the interaction between power supply and power consumption is conducive to cutting peaks and filling valleys of the power grid and ensuring safe and stable operation of the power grid.Non-intrusive load monitoring can obtain fine-grained power user data.It is an advanced sensor measurement technology and has important application value in many aspects such as intelligent power consumption,energy saving and emission reduction,and power grid regulation.Based on the extensive search of relevant domestic and foreign literatures,this paper expounds the background of the subject,related concepts and research significance,and briefly introduces the research status at home and abroad.Focus on analyzing the basic principles of non-intrusive load monitoring,and explain its main processes and key technologies.At the same time,it introduces the current evaluation indicators of non-intrusive load monitoring methods and the technical difficulties in this research field.Summarize the important methods of the s ubject and analyze the use of supervised learning and unsupervised learning methods on the subject.Existing methods mainly use supervised learning models,which require a large amount of targeted training data and cannot effectively identify loads that do not appear in the training data.In view of the unique advantages of unsupervised learning methods,this paper proposes a non-intrusive type of non-invasive load monitoring method.Take the transient power waveform during the state change of household appliances,as well as the power variable and transient time before and after the state change as load characteristics.After detecting the occurrence of the load operation event,the dynamic time warping algorithm is used to calculate the matching degree with the transient power waveform in the historical load event.Based on the waveform matching degree,dynamic clustering is performed,and then the association rule analysis is used to find out the load feature set belonging to the same electrical appliance,so as to achieve the purpose of identifying the load.In addition,the sequence pattern mining algorithm is used to denoise and sort the load features to improve the load recognition accuracy.At the end of the thesis,two examples are given to illustrate the feasibility and practicability of the method proposed in this paper.The first calculation example is based on the experimental data of the ideal situation,adjusts the algorithm parameters,verifies the feasibility of the method,and measures the rec ognition accuracy in the ideal environment.The second calculation example is based on the data set published by foreign countries,and is compared with some algorithms horizontally,and the advantages and disadvantages of this method are explained by relevant indicators.The analysis results of the calculation examples show that the method proposed in this paper is easy to implement and has a significant improvement in accuracy and reliability.