Research And Application Of Non-invasive Load Recognition Algorithm

With the establishment of the concept of global energy Internet,how to make full and reasonable use of electricity resources and promote the construction of smart power grid has become a hot issue of concern to all countries in the world.Power load monitoring technology is one of the key technologies to realize smart grid.Non-invasive power load monitoring technology is gradually replacing the traditional invasive power load monitoring technology due to its advantages of easy installation,strong maintainability,low cost and wide coverage.However,there are still many problems with this technology that have not been solved yet and it has not been widely used in practice.Based on this,the core part of this technology,load identification method is studied in this paper.The research work of this paper is of great significance and value for solving the problem of electricity consumption monitoring in some complex scenarios.Compared with the traditional intrusive power load monitoring technology,the current non-intrusive power load monitoring technology has the following problems:(1)when new equipment is added,manual intervention is needed to extract the characteristics and establish the characteristic database;(2)when the load characteristics of electrical appliances are similar,the recognition rate is low;(3)most of the existing methods are for single load identification.When multiple electrical appliances are turned on at the same time,the identification effect is not ideal.This paper will focus on these problems,aiming at the low accuracy of load identification in non-intrusive power load monitoring technology,single identification type,and the difficulty in building feature library,etc.,to build a non-intrusive power load system framework based on feature identification.The traditional pattern recognition method is adopted,and the emerging in-depth theory focuses on the research of load feature extraction and recognition in the non-invasive power load monitoring technology,and some research results with practical application value are obtained:(1)Taking the steady-state current of electric load as the research object of this paper,a non-intrusive power load system framework based on feature recognition is proposed;(2)A load information collection device is designed by using a single chip microcomputer.Household appliances are used as the collection objects to build a power load data set,which lays a foundation for the subsequent algorithm research.(3)A non-intrusive power load recognition algorithm based on traditional pattern recognition method is proposed.The load characteristics were extracted by Fourier transform and empirical mode modeling.Load is identified by k-means and type 2 fuzzy clustering method.In comparison,characterized by EMD decomposition component,the two-type FCM clustering method has the highest recognition rate as a classifier.To some extent,this algorithm solves the problem that the existing methods need manual intervention to build the feature library and improves the recognition rate of the load.(4)Based on deep study of non-invasive power load identification algorithm,is proposed based on stacked the encoder single model identification of the structure of the load identification algorithm,based on LSTM network model has been put forward to identify the load on the structure characteristics of the modeling method,and USES the four types of similarity evaluation method to evaluate the model so as to realize the recognition of electrical appliances.The two models are suitable for different electricity consumption scenarios.The proposed algorithm is verified by experiments,and the average recognition rate can reach 98.0%.This algorithm solves the problem of similar load identification effectively.(5)Completed the software design of the entire non-intrusive power load monitoring system in Python environment.This system has the functions of power consumption monitoring and information feedback.By simulating the real power consumption environment,the effectiveness of the proposed algorithm in practical application is verified.The method proposed in this paper is not only useful for the object identification in this paper,but also effective for other load identification.