Time series temperature data from an industrial steam turbine are used in the present analysis to develop methodology for anomaly detection. Simple and exponential smoothing techniques are used to study the effectiveness of the technique for prediction considering different periods for analysis. The analysis of the lags between the predicted and observed data is performed using associated parameters like average deviation, root mean square deviation and split error. Exceedance test is also applied to the data set and the results obtained are found to be consistent and satisfactory in identifying sharp anomaly in the observed real time data.