Research And Application Of Non-Intrusive Low-Voltage Power Load Composition Identification And Aggregate Modeling

With the continuous advancement of the construction of Jiangsu Smart Grid,a series of changes have taken place in the load of Jiangsu Power Grid,resulting in more complicated load characteristics of Jiangsu Power Grid.This requires the power dispatching department to accurately grasp the load composition and achieve deep perception of the load composition.Non-intrusive load composition identification and modeling technology analyzes and collects the electrical information of the user terminal to decompose and obtain the working status of each user's electrical equipment,which can not only make the user more aware of his own power consumption information,but also provide demand response data for the power grid Support to create conditions for the grid to increase dispatch flexibility and enhance renewable energy consumption.Therefore,the identification and aggregation modeling of low-voltage power load has obvious theoretical significance and engineering application value.In this paper,based on the massive low-voltage power load data accumulated by the AMI system,the basic principles of non-intrusive load monitoring and decomposition are analyzed,the classification and characteristics of the load are summarized,the framework of the non-intrusive load monitoring and decomposition algorithm is studied,and the coarse-grained components of the load are identified.Summarized the low-voltage power load modeling method,relying on the construction of the power load cloud integrated analysis platform,to achieve load cloud awareness and online aggregation modeling.The main tasks are:A non-intrusive residential load composition recognition method for mining data is proposed,and a ZIP model of aggregate residential load based on massive mining data in the cloud is constructed.A power load decomposition model based on a factor hidden Markov model is proposed,and the decomposition model is tested using the short-term data of five households provided by the REDD dataset.The results show the ability of non-intrusive load component identification.A low-to-up low-voltage power load aggregation modeling method based on the mining data is established to realize the online update of the aggregate load model parameters.Based on AMI data,the power load decomposition model training and calculation work based on factor hidden Markov model.Based on the typical staticload model and the proportion of electrical appliance power,a household load model is obtained.The load on the feeder is calculated based on the load of each household,and the online aggregate modeling of low-voltage power load is realized,and the model parameters are updated in real time.The results of the example analysis prove that the proposed load identification and modeling method more closely matches the actual operating conditions of the load,is more in line with the modeling expectations,and has stronger analytical capabilities.And the proposed identification and aggregation modeling method is applied to the problem of the ability to respond to the demand side of the interruptible load.Through non-intrusive load identification and aggregation based on power feature matching,the load component and the interruptible load proportion are obtained,which can compare Good implementation of demand side management requirements.The identification and aggregation modeling method of low-voltage power load composition can effectively tap the potential value of massive cloud-use mining data,efficiently realize the real-time identification and online update of feeder-level load aggregation model,and reveal the new ubiquitous power of "marketing data assisted operation scheduling" The data-driven application paradigm of the Internet of Things has also explored new application scenarios for the promotion and application of new-generation smart energy meters and other sensing collection devices.