Research Of Load Forecasting Based On Occupant Behavior Model

Some renewable energy technologies such as solar energy and wind energy greatly affected the matching of energy supply side and load side due to their instability and discontinuity.As an important technology to solve the matching problem,load forecasting plays an increasingly prominent role in the renewable energy system.In order to ensure the stability,efficiency and safety of ener gy system,to achieve fast and accurate load forecasting has attracted more and more attention and been widely studied.The traditional load forecasting model mainly considered historical load and environmental parameters such as temperature,humidity and solar radiation intensity as the input parameters,lacking of attention to the interaction between the occupant behavior and the built environment.However,occupant behavior is one of the main factors that affect the building load.Therefore,how to lead the complex and random occupant behavior into the process of building load forecasting becomes one of the keys to improve the accuracy of load forecasting.Firstly,based on BP neural network,this paper proposes two load forecasting methods,that is traditional load forecasting method and the load forecasting method based on occupant behavior model.In the traditional load forecasting method,the input parameters selected are: weather data,historical load,and fixed on and off work schedule,while for the load forecasting method based on occupant behavior model,hourly occupancy rate and hourly air conditioning schedule are used to describe occupant behavior in detail.In addition,the mean absolute error and the root mean square error are selected to evaluate the accuracy of load forecasting.Next,this thesis constructs a renewable energy system for an office building,the demand side of which is composed of photovoltaic panels,wind turbine,a battery and city power.Besides,the control system is composed of load forecasting system and dispatching platform.Then the modeling of photovoltaic and wind turbine was completed in TRNSYS,and the building model was constructed in DeST,in the building occupant behavior module of which,the number and beha vior model of occupant were further set.Through simulation,the cumulative annual electricity generation of renewable energy and the annual load of buildings are obtained.Then,the hourly occupancy rate and air-condition schedule based on occupant behavior model of each room are calculated,and they are compared with the traditional fixed on-off work schedule,the results shows that there exists obvious difference between the random schedule by the occupant behavior model and traditional fixed on-off work schedule.Next,in the calculation process of load forecasting,no matter the room or the whole building,the predicted value of the load forecasting method based on occupant behavior model is closer to the actual value,and the error value is smaller.The results show that the mean absolute error(MAE)and root mean square error(RMSE)values obtained by the optimized method are consistently within 0.4 for individual room.In addition,a marked reduction of the whole building is achieved by 59.05% and 55.57% for the MAE and RMSE values respectively,so it can effectively verify the effect of the detailed description of occupant behavior on the improvement of load forecasting.Finally,this thesis makes the energy management and control strategy of both sides of the supply and demand based on the electricity price policy,and evaluates the matching of the two sides of the system by using the two indicators of on-site energy fraction and on-site energy matching.The results show that the load forecasting method based on occupant behavior model not only improves the accuracy of load forecasting,but also promotes the matching of building renewable energy system both under the time scale of ‘day’ and ‘hour’.