Data Reconciliation is a technique that improves the accuracy of process data by adjusting the measured values so that they satisfy the process equations describing the physical phenomena. Along with Gross Error Detection it can be used to identify possible systematic errors in measurements and validate data for a further diagnostic phase. These techniques have been applied for a long time only to chemical plants with balance equations (mass and composition) and then they have been extended to industrial gas turbine, including characteristic equations.Usually all Data Reconciliation techniques are based on constrained minimization of an objective function which is the total weighted sum square of adjustments made to measurements to fulfill the constraints. The strategy proposed in this paper exploits least squares optimization to solve a system of non-linear equations, which represents the same Data Reconciliation problem in a more efficient way. The system of equations involves both the measurement adjustments weighted with measurement uncertainties, and the process equations characterized by an appropriate weight to maintain the error under expected limit, which allows to consider them fulfilled. In this paper this novel approach to Data Reconciliation applied to industrial power plants has been studied and compared with classic Non Linear Data Reconciliation techniques, based on Sequential Quadratic Programming and Lagrange multiplier. Both linear and non-linear case studies have been developed to test the algorithms performance. Moreover a case study based on real field data has been carried out on the base of an Ansaldo Energia industrial gas turbine. To filter real data a Gross Error Detection technique based on serial elimination has been applied along with Data Reconciliation in order to identify the presence of some possible systematic errors in measurements.The comparison between the several approaches has highlighted better performance in terms of implementation and calculation speed of the least square algorithm which provides the same accuracy both in reconciling measurements and in respecting the constraints.
