An Application of Support Vector Machines in Cooling Load Prediction

The methodology to predict building energy consumption is increasingly important for building energy baseline model development and measurement and verification protocol (MVP). Improving the energy efficiency of buildings by examining their heating, ventilating, and air-conditioning (HVAC) systems represents an opportunity. To improve energy efficiency, to increase occupant comfort, and to provide better system operation and control algorithms for these systems, online estimation of cooling load is desirable. A difficulty in HVAC system parameter estimation is that most HVAC systems are nonlinear, have multiple and time varying parameters, and require an estimate of the cooling loads for a building zone. This paper presents support vector machines (SVM), a new neural network algorithm, to forecast cooling load for HVAC system. The objective of this paper is to examine the feasibility and applicability of SVM in building load forecasting area. An actual HVAC system in Nanzhou is selected as case studies. In addition, the performance of SVM with respect to two parameters, C and epsiv , was explored using stepwise searching method based on radial-basis function (RBF) kernel. Finally, actual prediction results show that SVM forecasting model, whose relative error is turned out to be about 4%, may be better than autoregressive integrated moving average (ARIMA) ones.