Research On Non-intrusive Load Monitoring Method For Office Buildings

With the rapid development of the national economy,China’s office building area is increasing,and the energy consumption of office buildings is also growing,so energy saving in office buildings is undoubtedly one of the key points in China’s building energy saving work.With the rapid update of grid intelligence,the massive amount of detailed load electricity consumption information can increasingly reflect its key role in energy saving and emission reduction.Non-intrusive load monitoring can break down the individual electricity consumption data of each appliance based on the aggregated data of the user’s main meter,which can give users a more detailed understanding of the equipment’s electricity consumption in order to make reasonable energy saving plans and reduce electricity consumption costs.Load monitoring data can help the power grid understand the composition of the load,according to the user’s electricity consumption pattern,reasonable and planned composition of the user’s electricity consumption time,to achieve peak and valley reduction.Detailed electricity consumption data can effectively reduce the use of high-energy appliances and effectively implement demand response measures.To this end,this thesis conducts a study on non-intrusive load monitoring for office buildings,The specific work is:（1）The experimental data collection platform is built,and select five kinds of office building electrical appliances such as water dispenser,heater,electric vehicle charger,microwave oven and LED as the experimental object,collect steady-state data and transient data according to different trigger mechanisms,store the data in the MySQL database,extract and process the data to establish the experimental data set.（2）Aiming at the problems of low accuracy,slow decomposition speed and poor timeliness of the load decomposition algorithm of the traditional network model,this paper models the non-intrusive load monitoring method based on steady-state data,a load disaggregation model based on Temporal Convolutional Network（TCN）combined with an improved sequence-to-point architecture is designed to address the problems of traditional network models in terms of low accuracy and slow decomposition speed of load decomposition algorithms.The original sequence-to-point architecture is improved by moving the prediction point from the midpoint to the endpoint of the output window,which improves the real-time performance of the model decomposition and enables the model to learn more historical information.TCN uses dilated causal convolution so that it can capture long-term dependencies and prevent information loss,and can effectively avoid problems such as gradient disappearance and explosion caused by too many deep layers of the network.（3）Aiming at the existing non-intrusive load monitoring methods,the research mainly focuses on improving the accuracy of load identification,resulting in problems such as high model complexity and difficulty in application on embedded devices,this thesis designs a model of non-intrusive load monitoring methods based on improved kNN algorithm with transient steady-state features.The kNN algorithm is selected as the load identification model,and the distance weight statistics method is used to improve the kNN algorithm,and the cosine similarity judgment mechanism is added to check the accuracy of the load identification results of the kNN algorithm,which can realize the recognition of unfamiliar devices.The use of event waveform separation mechanism in the data acquisition process eliminates the disturbance of normal load in the main circuit,reduces the difficulty of load identification.The separated event waveforms are then used to extract transient features and steady-state features as load features to improve load feature recognition.（4）A non-intrusive load monitoring system platform is established based on the non-intrusive load monitoring method designed by transient data,and the B/S architecture is used to realize the development of the non-intrusive load monitoring system and the deployment of the cloud server.A non-intrusive load monitoring system platform with data collection,remote data storage and display in one was established.