The X-ray fluorescence well logging technology is a significant method that can make quantitative analysis orsemi-quantitative analysis on the wellface. This method is very important to mineral exploration. The spectrum shifting is often observed in the X-rayfluorescence well logging because the temperature in the well changes acutely. The hardware is used to release the spectrum shifting and the software method is used to rescale the tiny spectrum shifting. There are too manyspectra to be rescaled in a well logging task by manually. In this paper, an auto method to rescale spectrum shifting, via the expert system model which is based on the special process to rescale spectrum shifting in manual, is presented. The symmetric zero-area conversion method, which is not sensitive to the changes of the baseline, is used to research the peaks. And then, the characteristic peaks will be identified by the standard errors, automatically. The prior knowledge (the last energy scale) and the gauss probability density function are used to analyze the peaks qualitatively and confirm the energy of characteristic peaks. Then the least square method is applied energy calibration. The singular deviation point, away from the calibrationline, will be rejected and the energy ratio will be obtained again. This method is applied for rescaling spectrum shifting in 322 spectra and obtains a satisfactory achievement.