Font Size: a A A

Topology Analysis Of Load Network And Load Decomposition For Non-intrusive Load Monitoring

Posted on:2020-03-05Degree:MasterType:Thesis
Country:ChinaCandidate:B F HuangFull Text:PDF
GTID:2392330596994967Subject:Electrical engineering
Abstract/Summary:PDF Full Text Request
The load decomposition method for non-intrusive load monitoring is a method for collecting and decomposing the power consumption of a specific load from the total power consumption information at the power input port and simultaneously determining the working state,and the obtained decomposition result can be used not only for the method of obtaining the working state.Let users directly understand the power consumption behavior of each device,and then provide more useful reference information and help the power users to rationally arrange electricity consumption,save electricity,thereby saving energy and reducing consumption,and improve efficiency.The existing power load decomposition scheme for non-intrusive load monitoring mainly establishes the equipment load model according to the electrical characteristics of the collected power load and then uses different optimization algorithms or pattern recognition methods to realize the decomposition of the load of the power equipment.The problem is at the current stage,and there is a good decomposition method to identify the equipment with high precision.At the same time,most of the research attempts to solve the load decomposition in multi-class load mixing operation such as high-noise,high-power identification of low-power equipment.However,in real life,a single detection method is temporarily unable to identify all devices,so this article will proceed from the actual situation and propose a new solution from another angle.In view of the above problems,this paper proposes a non-intrusive power load decomposition method based on network topology.First,it is necessary to collect the electrical information of the electrical equipment and use the data generation technology to simulate a variety of complex power usage conditions and generate sufficient training data.Secondly,the current relatively good machine learning algorithms are tested to avoid the inconsistency of algorithmic results due to the differences in the number of devices tested in other people’s research.Furthermore,based on the actual use environment,this paper simulates its electrical network topology and then generates a topology map for testing the practicability of the algorithm.Based on this,the optimization method is used.Under certain conditions,the global load decomposition is realized from the user’s point of view.Finally,this paper uses the common REDD public data and the real electrical data and network topology of a factory to test,and from two different criteria,from the user’s point of view,we from the accuracy rate(ACC),F1-measure,In the root mean square error,the accuracy rate is chosen as the main indicator of this paper.The experimental results show that the proposed non-intrusive power load decomposition method based on network topology can effectively realize the decomposition of global power load and can be better applied in real conditions,so it has higher generalization performance.
Keywords/Search Tags:Non-intrusive load monitoring, intelligent algorithm, machine learning, Genetic Algorithm, Random Forest
PDF Full Text Request
Related items