| Based on advanced metering infrastructure(AMI),this thesis designed a practical domestic energy center with functions of electric appliance identification and charge load forecast,it is of great significance for the household users to percept the working status of domestic devices,save electric energy,make reasonable planning of domestic electricity consumption,and improve smart family.The thesis analyzed advanced metering infrastructure,designed a concept model that meets AMI system and the hardware and software architecture of energy center in detail.Also designed the remote communication network in detail and realized GPRS communication mode,in the user’s indoor network,the Zig Bee network was created to realize a series of functions such as the management and control of the other nodes in the room.For the equipment of the non-intrusive online decomposition of recognition was at the entrance of the home energy center collected data.By the decomposition of the total current to monitor indoor each electrical appliances with power and working condition,also can be learned within the family of electrical equipment with working status and use the rule and so on.The electrical steady-state current as decomposition and recognition of feature value,and according to the total current of the time series curve area,the thesis designed a single feature value of two different steady-state and jump variable pattern recognition mode,and multi feature value spatial localization pattern recognition method.By solving the minimum value of the two integer programming,these methods achieved the recognition of the work of each device.Through simulation results,compared with the actual equipment operation,the equipment identification rate was in a reasonable range.For household power consumption forecast,a practical algorithm was designed with reference to time series ARIMA model and this algorithm can be easily applied to embedded devices.And three kinds of algorithm prediction experiments are carried out for the five sets of sample data of daily power consumption.These algorithms include the classic ARIMA model,the improved slip time series algorithm,and 2 different input mode BP network algorithms.Through simulation analysis,the results show that the difference between the time series algorithm and the classical ARIMA simulation was not significant,in the forecast of fifth sample data,the slip time series algorithm was more accurate than the classical ARIMA forecast.In the comparison of forecast results,between the slip time series algorithm and the BP neural network algorithm with 2 different input modes,the slip time series algorithm is more accurate than the BP neural network algorithm to forecast the results. |
