Building Predictive Control Oriented Model Practicalization Simulative Study

Model Predictive Control(MPC)is capable of the integrated optimization of the multi-input multi-output problem in building control.It helps building to meet the indoor comfort requirement while minimizing its energy consumption.In the meanwhile,MPC possesses the intrinsic ability of occupancy prediction.To use such ability to control the flow rate of fresh air might unearth noticeable energy saving potential,compared to regular fixed-volume fresh air control.However,when average engineers are trying to use MPC for the control of automated building,they may encounter the following three obstacles: a)describing the actual physical interactions of the building within itself and with the environment as a control model,b)truncating a high-order control model into a low-order to make it more solvable,and c)estimating the unobservable variables generated in the process of model truncation.The MATLAB based toolbox BRCM(Building Resistance Capacitance Model)has already removed the first obstacle for engineers,which makes removing the second and third obstacles even more important in the path of building MPC practicalization.To address these two issues,this research designed a lumped-model approach for the practicalization of building MPC that reduced a 19-zone model to a lumped 1-zone model.Applied the lumped model to a real building for simulative control and analyzed the fresh air control performances and energy saving potential based on occupancy prediction.The main contributions are as follow:a)A stage-wise zoning simplification method was proposed.The simulative deviation of thermal load was located and compensated through stage-wise simplification,and a lumped model for the construction of MPC was obtained.The relative error of the lumped model from the conventional model was calculated to be 2.62%,showing a qualified model balancing the trade-off between model accuracy and paracticalness.b)Applied the lumped model to a real building's MPC strategy,and the controlling performances were analyzed.The temperature was well controlled in its setpoints and the energy consumption level was showed to be comparable to the ideal minimum.Additionally,the energy consumption curve performed by MPC was flatter.However,by a further investigation,noticeable energy saving potential still exists if variable fresh air volume control is integrated.c)Three different occupancy prediction methods were compared by the metric of accuracy,and ARIMA(0,1,1)was proved to be the most accurate and could best track the oscillation of the occupancy curve.The root mean square error(RMSE)of ARIMA(0,1,1)was 18.01,and it was chosen as the predictive method of occupancy disturbance in MPC.d)The variable fresh air volume control was performed based on MPC.The results showed 59.8% of unnecessary fresh air supply was reduced and 20% of energy savings was achieved compared to fixed volume control.Meanwhile,variable fresh air volume control based on occupancy prediction did not require the installation of any explicit sensors and can response more promptly to indoor fresh air request compared to feedback control.