[en] To improve the computational accuracy of core axial power shapes in COLSS (Core Operating Limit Supervisory System) of ABB-CE reactors, a new method using extra 4 pseudo-detector signals to evaluate axial power shapes was proposed and tested for Younggwang Nuclear Unit (YGN) 3 cycle 3 and YGN 4 cycle 4. To find optimal correlation between each pseudo-detector signal and 5 real detector signals, the Alternating Conditional Expectation (ACE) algorithm was used. And the conventional Fourier fitting method was adopted to calculate 20-node axial power shapes with 9-detector information. To verify the usefulness of new method, a total of 3462 axial power shapes per each cycle produced by ROCS (Reactor Operation and Control Simulation) code were recalculated by different axial power shape reconstruction methods. The results were compared with those of the existing Fourier fitting method and stochastic method using the ACE algorithm. The average Root Means Square (RMS) error and average of axial peaking, DFZ, error of the proposed 9-detector method shows about 50% and 70 reduction, respectively, relative to the existing 5-detecotor method. Because the proposed 9-detector method, compared with the stochastic power prediction method, has no restriction on expanding from 20-node shape to 40- or 50-node axial power shape, it may be an useful method for precise reconstructing of axial power shapes when only 5-detector information are available