**5. Conclusion**

In the chapter, three data processing approaches to improve data quality are demonstrated. For the importance of properly controlling the steam system performance normally, The data obtained from EMS should be accurate and reliable. However, the data may be influenced many outer factors. The approaches proposed in the chapter are to detect the fault data, locate and remove the gross errors and reduce the random errors.

Four main reasons induce the low accuracy of the mass flow rate measurement. Combining the principle of "3σ" and empirical distribution function to determine control limit is proposed for single variable monitoring, and applying PCA to determine the control limits for the multivariate process. With the limits, most of fault data can be identified easily. For the fault data of flow rates, the approach to setup the mathematical model of the steam network and calculate the flow rates is proposed. The simulation and experimental results show the effectiveness of the approaches.

Two approaches, MT and MP, to detect the gross errors are demonstrated. Both are preceded by selecting the statistical variables, which follow standard normal distribution, and applying hypothesis test. Some notations for the two algorithms are stated.

The constrained least-squares problem applied for present application is discussed. The four assumptions are approximately satisfied when the steam network is normally function and the state is nearly static. The pipe network loss can be considered to add to the constraint equations for more accurate results. The weighed parameter matrix has influence on the results of data reconciliation. To estimate the deviations of the instruments online and apply several instruments with high precision will improve the quality of reconciled data.
