REGRESSION METHODS FOR THE DETERMINATION OF COLLECTOR COEFICIENTS

The Euronorm [ 1 ] proposes the Multilinear Regression (MLR) for the QDT. The efficiency of the collector is determined by a model containing 6 parameters (1). The six variables of the model have to be determined from the measured data. With the regression, the six collector coefficients can be determined. By minimizing the square errors of the differences between the calculated and the modeled efficiencies of all data pairs the coefficients are determined. As the uncertainties of the primary measurement data (radiation, flow rate, temperatures) join the calculation of the efficiency, a procedure would be useful, which considers appropriate weighting of a data pair in respect to the inherent uncertainty of that data pair.

For example, pyranometers with an offset of approximately 10 W/m2 show a higher relative measuring uncertainty in their lower measurement range than in their upper range. The weighted least square method (WLS) weights “measuring points by lower radiation (<300 W/mF)” less than for higher radiation for the regression process. Articles [ 5 ], [ 6 ] and [ 8 ] show that the MLR regression based on the least square method (LS) is not sufficient enough to determine the collector coefficients with their uncertainties for the SST. This paper shows the advantage of WLS for the QDT. In the following we discuss the different results obtained by the LS and the WLS for the QDT.

01