Abstract-In oil tank farm, the fire hazard causes a great loss to both people and property. Once there is a fire, to avoid more loss, a reliable, secure and practical method for putting out a fire is needed. And under this circumstance, the accurate forecasting of the changes of the environment, such as temperature or humidity, is of great value. We employed time series analysis method to forecast the temperature and humidity changes when there was a fire accident in the oil tank farm. According to the record, we modelled a suitable ARMA model. And the results show that ARMA is an effective method to predict time varying series in the oil tank farm. Our results in this paper also provide an important way of studying the fire spreading in the oil tank farm.