Research On Knowledge And Data-driven Building Electricity Demand Forecasting Method

With the continuous improvement of people’s living standards and quality of life,people’s requirements for the indoor environmental quality of buildings are being upgraded,the increase in building energy consumption brings tremendous pressure to the power grid supply,energy and environment.The optimal dispatch of building energy supply and energy consumption is an important way to balance building energy consumption and indoor environmental quality.Building a faster and more accurate prediction model of building energy consumption is the basis and prerequisite for understanding the characteristics of building energy consumption and realizing the optimal control of buildings.It is very important for balanced distribution of energy and energy saving.The knowledge and data driven modeling methods of building electrical power demand forecasting model are researched.The main contents are as follows:1.The research background and significance of building electricity demand forecasting are analized,summarizes the main research methods currently used in building load forecasting,analyses its main characteristics,gives the classification of large public buildings and the composition of their energy consumption,and puts forward a design method of building electricity demand forecasting model based on knowledge and data.2.Data acquisition system of building energy consumption is constructed to collect building energy consumption data and outdoor air temperature and humidity.At the same time,data normalization and sliding average filtering are used to pre-process the collected sample data to improve the data accuracy.3.Based on the energy consumption characteristics of public buildings,the influencing factors of energy consumption of public buildings are analyzed,and the most obvious influencing factors of energy consumption of office buildings are preliminarily judged byorthogonal test.Two main influencing factors are selected from the influencing factors according to the main research object of this paper.4.Based on the existing modeling methods based on HCMAC(Hyperball Cerebellar Model Articulation Controller)neural network,a prediction model of building power consumption based on GIHCMAC(Genetic Algorithm Ant Colony Clustering Algorithm based on HCMAC)neural network is proposed by using genetic algorithm and ant colony clustering algorithm.The model is simulated by MATLAB software.The simulation results show that the accuracy of the model is significantly improved compared with the existing prediction model based on CMAC.