Power-side Load Prediction Combined With FHMM Load Decomposition In Scenario Context

The monitoring and analysis of the electricity consumption status of residential users can provide data support for user-side demand response,energy efficiency management,and refined electricity service.Non-Intrusive Load Monitoring(NILM)is an important technical means to realize user load status monitoring.Therefore,this paper aims to realize non-intrusive load monitoring of residential power consumption from the two aspects of load feature extraction and load decomposition method based on the existing load data of the power consumption side,and based on NILM to study the short-term residential load forecast,the specific content is as follows:The detection of switching events of electrical appliances is the premise of load feature extraction.In view of the difficulty of selecting thresholds in existing event detection algorithms,the power ratio algorithm is used for event detection,and the control parameter of the algorithm,the window width is discussed.Aiming at the overlapping of load characteristics of some electrical appliances,the steady-state characteristics of the load are extracted from the time domain and frequency domain,the similarity of the similar load characteristics of different electrical appliances is quantified,and the active power,reactive power,current amplitude and harmonics are quantified.The distortion rate is used as a template feature to establish a load feature database.To solve the problem of low recognition rate when using single feature optimization in load decomposition,based on template features,a multi-load feature objective function optimization model is established,and differential evolution algorithm(Differential Evolution,DE)is used for load decomposition.Considering that DE has premature convergence and late evolution and other problems,the mutation strategy of DE is improved.Aiming at the mathematically optimized load decomposition model without considering the operation of multi-state electrical appliances,a factor hidden Markov model(FHMM)load decomposition model is established,and the Viterbi algorithm is used to solve the model.The problem of high algorithm complexity,combined with the context information of the context,constrains the state space and state transition path to reduce the complexity of the Viterbi algorithm.Based on NILM,the short-term forecast of residential load is studied.First,the daily load sequence of a single appliance is predicted,and then the predicted load of all appliances is added to obtain the total load sequence of the predicted day.Considering the influence of meteorological factors on the forecast,the meteorological factors and the daily load sequence of the load equipment are correlated,and the characteristic quantities of the load forecast are screened according to the analysis results.Finally,a load forecast model of a single load equipment is established based on the BP neural network.Aiming at the problem of slow convergence speed of BPNN,combined with improved DE,the network parameters are optimized,the back propagation error function is used as the objective function,the weights and offsets of the network are optimized,and the optimized parameters are used as the network The initial parameters improve the accuracy of the load forecasting model.