Forecasting Occupancy for Demand Driven HVAC Operations Using Time Series Analysis

Abstract

Building heating, ventilation, and air conditioning (HVAC) systems contribute substantially to the energy consumption of buildings. Today, traditional HVAC systems mostly operate according to the maximum occupancy assumption, which in turn increases energy consumption during periods of low occupancy. Although, recently, implementing demand-driven HVAC operations are accepted as an innovate approach for reducing HVAC-related energy consumption, occupancy forecast is important to realize demand-driven HVAC operations in buildings. This study aims at using time series models in order to forecast the daily number of bank customers in a financial center of a bank, which is located in Izmir, Turkey. Data were collected from the computerized tracking system for a period of 60 weeks and two forecasting methods were used: 1) Decomposition Method, 2) Box-Jenkins Method. To determine the final model identified via the Box-Jenkins Method, goodness-of-fit, residual analysis and Akaike information criterion were taken into consideration. The results show that the SARIMA model with a MAPE of 11% yields a good occupancy forecast for supporting demand-driven HVAC operations.