Recognition Of Electrical States For Home Appliances And Power Load Forecasting Method For Buildings

In recent years,because of the popular use of white electrical home appliances such as air conditioners of heating or cooling appliances,the power consumption of the buildings especially of the public buildings is growing rapidly and take too much proportion in the total social energy consumption.New energy micro-grid as the power system for buildings can save energy and reduce emissions efficiently,and the monitoring of the electrical states of the low voltage electric appliance and the accurate load peak prediction for the buildings is one of the basic and key technology for the safe and economic operation of micro-grid.However,the truth is most of the buildings are not smart home equipped with smart meters,and one house or floor of public buildings or one power supply monitoring point only has a total current sensor or meter and with it we cannot directly collect the independent current signal and monitor electrical states and the power consumption of each independent electrical device.To solve the above problems,we focused on the three issues in this thesis,which are the electrical states recognition of the home appliances,and the single-channel blind source separation for the aggregated current signal of the multiple home appliances,and the buildings power load forecasting method.The main research work and conclusions are as follows:(1)The recognition method for the electrical states of single home appliance.After the full research on the characteristics of steady-state signals and transient signals of many different types of home appliances and the relationship between the control mode of home appliances and the state-series,we proposed a novel method to extract the features for steady-state signals and transient signals separately and modeling the state-series graph.According to the degree of stability of the envelope the steady-state signals are divided into two types,one is the long-period steady-state signal with long-time of multi power frequency period with exponential gradient envelope,the other is short-period steady-state signal of single power frequency period with steady envelope.For the long-period steady-state signals the FFT is used to extract the features such as the harmonics and envelope,phase,average power of each frame.For the short-time steady-state signal the Hilbert-Huang Transform(HHT)is adopted to extract the features.For the transient signals the Wavelet Packet Transform(WPT)is done to extract the features.A certain type of microwave-oven,air-conditioning,washing machine were used as the test object.The experimental results showed that the identification algorithm proposed in this thesis has high recognition rate and could be widely used,and with which we can real-time identify the electrical state and acquire the state-series graph,and calculate the state entropy used for fault prediction.(2)The method of the single-channel blind source separation for the aggregated current signal of the multiple home appliances.A novel method of semi-blind source separation based on the inner product of the feature ranking vector is proposed to separate the semi-simulation mixed current signals and the measured aggregated current signals.The semi-simulation mixed current signals means to simulate a mixed signal mathematically based on the measured data of single home appliance,the measured aggregated current signal is captured by the single current sensor for the multiple home appliances which are powered by the same voltage.We chose an air conditioner,a microwave oven,a washing machine,computers and other common electrical equipment as the test object to testify the proposed method.The experimental results showed that with this method the aggregated current signal of the three different kinds of home appliances can be fast and accurately separated.(3)The short-term power load forecasting method for buildings.The traditional coarse-grained forecasting based on neural network model and the fine-grained forecasting based on the working states of the home appliance were discussed.The coarse-grained forecasting combining the Elman neural network and the particle swarm optimization algorithm was tested using the measured data of public buildings in AnHui JianZhu university.The fine-grained load state estimation and peak prediction model was built according to the state-series graphs of the home appliances.The experimental results showed that the method and model of this paper can effectively shorten the convergence time of network and have higher load forecasting accuracy and stability.The fine-grained forecasting method based on the working states of the home appliance of this thesis proposed a exploratory ideas for fine-grained load prediction of the micro-grid for buildings.The work and contributions of this thesis will be widely used in many fields such as in micro-grid and energy management for buildings,automatic detection and health diagnosis for low voltage electrical appliances,network smart home appliances,and so on.This project deserves further research.