Research On Non-Intrusive Load Monitoring Algorithms In Residential Building

With the threat of exhaustion of global energy resources and environmental degradation,energy saving and emission reduction has become one of topics of human society.In order to deal with this problem,power grid is progressing towards the intelligent and saving tendency.As a key technology of smart grid,load monitoring in residential houses has a great significance for the development of society.In order to meet the requirements of intelligent measurement and demand-side management in smart grids,efficient and low-cost load monitoring has become a technical problem to be solved urgently.Traditional load monitoring methods are usually invasive.The current and voltage detection equipment is installed on each load branch to monitor the power of each appliance branch.The biggest problem with this approach is cost and installation complexity.Consequently,researchers proposed the Nonintrusive Load Monitoring(NILM)approach.Non-intrusive load monitoring requires only the installation of current and voltage transformers at the power supply entrance,and uses the total current and voltage data to obtain the operating status and power status of a single electrical load to realize the monitoring of each electrical load.At present,there are two research directions for NILM,namely non-intrusive load identification based on event detection and non-intrusive load decomposition based on load model.This article proposes solutions for these two directions respectively:Event detection refers to the detection of changes in operating status of electrical loads.This paper proposes an event detection method based on mathematical morphology.It uses a multi-resolution mathematical morphological gradient to detect current changes.Spike detection of current gradients is used to detect event occurrence and locate the time of event.The example analysis shows that the event detection method proposed in this paper can accurately detect events.This paper presents a load identification algorithm based on event detection considering multiple operating states.The algorithm uses the k-means clustering method to cluster the load signals of different operating states of an individual appliance in an unsupervised manner,and extracts harmonic features or V-I trajectory features from the load signals after the label.The feature vectors are used to train the classification model.The trained classification models are used to achieve load identification.The example analysis shows that the proposed load identification method has a good recognition effect.This paper also proposes a non-intrusive load decomposition algorithm based on Hidden Markov Model(HMM).The algorithm uses a clustering method based on Density-Based Spatial Clustering of Applications with Noise(DBSCAN)to establish an accurate Hidden Markov Model for a single appliance.On this basis,the total load model based on Factorial Hidden Markov Model(FHMM)is used to transform the problem of load decomposition into an optimization problem for finding the maximum probability.The integer programming method is used to solve the optimization problem.