This paper proposes an energy consumption-prediction method for metro heating

ventilation and air-conditioning (HVAC) systems based on an auto-regressive moving average (ARMA) model using a time-series data analysis. Firstly

stationarity analysis and white-noise analysis (also known as pure stochastic analysis) were carried out on the collected energy-consumption data from actual metro HVAC systems. Secondly

optimal model parameters were determined using the autocorrelation function (ACF)

and partial autocorrelation function (PACF) and Akaike information criterion (AIC). Finally

an effective energy consumption-prediction model was established. Four different methods were employed to test the effectiveness of the established ARMA model. Meanwhile

two performance indexes

namely

mean absolute error and root mean square error

were adopted to evaluate its performance in terms of fitting the observed energy consumption data. The results demonstrate that the proposed method based on the ARMA model could extract useful information from the energy data and is thus effective for energy consumption prediction of metro HVAC systems.