Research On Electricity Scheduling Technology Of Home Energy Management System Based On Load Forecasting

The continuous development of distributed energy technologies will make renewable energy sources appear in more decentralized forms in individual residences,such as renewable energy generation and storage(e.g.,photovoltaic rooftops),so that home-based electrical energy systems will become an important part of future power systems.The internal home energy system is also being increasingly emphasized,and its research facilitates the rational use of home energy,both to achieve the need for stable operation of the external power system and to reduce the electricity bill of the home user.The key to a home energy management system is the accurate prediction of the household load.In order to more accurately dispatch the household electrical load,it becomes particularly important to forecast the residential load at the individual user level.In previous studies,most of the forecasting has been done at the system level(e.g.,citywide load forecasting).In contrast,individual household load forecasting is more challenging because of its high volatility and randomness.Machine learning techniques are nowadays more and more widely used and play an increasing role in technical research.And load forecasting is an important application scenario of machine learning,and machine learning is currently the preferred method to perform load forecasting.When using machine learning methods for load forecasting,the prediction effect has a great relationship with the features in the dataset,and the effective elimination of some redundant and wrong features in the dataset can improve the prediction accuracy.In this paper,we first use load data of multiple household appliances in home load forecasting to increase the features of the data,and then propose a genetic algorithm-based feature selection method to effectively eliminate features in the data that are useless or unfavorable for model prediction.The results of load forecasting show that using this feature selection method for load forecasting can effectively improve the forecasting accuracy.This paper next designs a residential electric energy scheduling rule based on time-ofday tariffs.The inputs to this scheduling rule are the forecasted load conditions for the day ahead and the known time-of-day tariffs,and the outputs are the battery charging and discharging power for each period of the day ahead.Experimental results show that load scheduling using this rule is effective in reducing the household’s electricity bill.The rest of this paper is described as follows.Chapter 1 of this paper firstly introduces the research background and the status of domestic and international research,Chapter 2introduces the general design of the home energy management system,Chapter 3 introduces the data processing scheme of the home energy management system,Chapter 4 introduces the load forecasting model and load scheduling model of the home energy management system,demonstrates and explains the experimental results,and finally concludes and analyzes this paper.