This paper proposed the history-independent algorithm based on CSM (cylindrical source model) in the TRT (thermal response test) with unstable heat rate. It achieved a faster computing speed and more accurate results than the usual superposition algorithms in processing variable heat rate problem. In order to alleviate ill-posed problem and improve the reliability of parameter estimation, a Matlab program was compiled to perform inversion calculation to obtain the ground thermal properties based on the SAA (simulated annealing algorithm) with preset lower and upper bounds for estimated parameters. The estimated ground thermal conductivity and volumetric heat capacity from in-situ TRT with unstable heat rate (case A) showed closest agreement with those (case B) evaluated by the LSM estimation method, and calculated water temperatures corresponding to estimated results agreed well with the measured water temperature. Based on SAA coupled with the history-independent mathematical algorithm in in-situ TRT with unstable heat rate, TRT can be performed continuously, independent of power fluctuation with both time and cost saved. (C) 2015 Elsevier Ltd. All rights reserved.