Research And Application Of Home Load Event Detection And Identification Method

Electric energy is the most widely used energy in the world,and the rational use of electric energy is an important link in creating ecological civilization.In recent years,the rapid growth in electricity consumption has led to peak load saturation on the grid.Residential electricity consumption is an important component of peak load and even peak load of power grid,and concentration is its unique property,which affects the stable operation of power grid.Intelligent power consumption can reduce the consumption of building electric energy,achieve peak clipping and valley filling and load transfer by means of power flow regulation and time-of-use electricity price.Load monitoring and measurement is an important part of realizing intelligent electricity consumption.Through load monitoring,the electricity consumption of each electrical appliance can be timely understood by the user,and the user can actively optimize the way of electricity consumption and reasonably manage the family energy.By monitoring the components of residential load,the power grid can accurately manage the demand side and reasonably allocate resources for residential load.As monitoring technology development trend and future direction,noninvasive power load monitoring is an important way of load monitoring,application of NILM technology will use data for a single device,household energy consumption data classification can be decentralized and centralized metering user power consumption,this method can not only reduce the installation and maintenance cost,also can understand the power law,It plays an important role in data support for power system demand-side management,the choice of the best power consumption scheme for users and the promulgation of relevant policies by relevant government departments.This paper takes non-intrusive load monitoring as the research object,and focuses on the field of event detection and identification.The main work is as follows:(1)To summarize the research status and related technologies in the field of noninvasive monitoring.On the basis of extensive reading related literature both at home and abroad,firstly analyzed the research background and significance of this topic,and then expounds event detection of non-invasive intelligent building household electric equipment and the basic process of load identification,the final two research for event detection and identification of load plate,respectively in this paper,the commonly used several methods and their advantages and disadvantages.(2)A non-invasive event detection method based on the ARIMA-F test is proposed.Load switching events are affected by multiple factors,such as excessive or abnormal power fluctuations in the operation of some devices,which lead to the failure of detection algorithm to correctly identify whether they are switching events or not,and long switching time on or off of some devices,etc.Due to the above problems,this chapter is based on the ARIMA-F test model.A non-intrusive load switch event detection method for residential electrical equipment in intelligent buildings is proposed.Firstly,using the advantage of median filtering to effectively deal with noise,the problem of misjudging events caused by equipment power fluctuation is solved.Then,the typical transient active power waveform of each appliance is analyzed in detail,and the disadvantageous factors of the equipment with long transition time are dealt with by different functions.Then,after preprocessing,the original sequence was decomposed into trend sequence,periodic sequence,and residual sequence by the ARIMA algorithm,and the residual sequence was extracted as the data to detect load switching events.Finally,the validity of the method proposed in this chapter is verified and analyzed by Blued public data set.(3)A non-intrusive load identification method based on NP-MLSTM is proposed.To solve the problems of the excessive workload of identification due to diverse load characteristics,an insufficient number of partial load samples and low accuracy of identification due to too complex operation of multi-category identification,a short and long time memory network LSTM with "memory" function was taken as the basis.A non-intrusive load identification method based on NP-MLSTM for typical household electrical equipment in intelligent buildings is proposed.In this chapter,the KCCA method is used to select the features with high correlation with the sample labels as the alternative features.Then,the Mixup method is used to expand the small number of samples,and the similarity of the samples before and after the expansion is tested by cosine similarity.Then,a multilayer preferred binary identification model is constructed to identify the load data.Finally,the high precision of the proposed load identification method is verified through experimental and comparative analysis.(4)The self-collected data sets are used to verify the methods mentioned in the first two chapters of this paper,and a non-intrusive home load monitoring visual platform is designed.Given the limitations of different data sets used in switch event detection and load identification mentioned above,this chapter uses self-collected load data sets of electrical equipment to conduct event detection and load identification experiments.Firstly,load data is collected by a data acquisition device,and load data are preprocessed accordingly.Then,non-intrusive load switch event detection and load identification experiments were carried out to verify the overall effect of the two task modules on the same data set.Finally,a load monitoring visualization platform is designed,which includes two main function modules,switch event detection and load identification,to highlight the practicability of the method proposed in this paper is a visual display way.