基于Levenberg-Marquardt算法的静态地层温度反演方法OA北大核心

Static formation temperature(SFT)is a crucial parameter for assessing the thermal state of geothermal and hydrocarbon reservoirs,playing a vital role in reservoir characterization and development planning.Traditional SFT estimation methods typically rely on long-term field measurements or physical simulations,which are often time-consuming,costly,and sensitive to data incompleteness or external disturbances.To enhance the efficiency and accuracy of SFT estimation,this study proposes a nonlinear fitting method based on the Levenberg-Marquardt(LM)optimization algorithm.The approach integrates the Horner-LIU method(HLM)model to establish the predictive relationship and utilizes the LM algorithm for iterative optimization of the model parameters.Six sets of bottom hole temperature(BHT)data from geothermal and petroleum wells were used for validation.Results indicate that the proposed method achieves high prediction accuracy using only a few early BHT data points(e.g.,the first 3 to 5 measurements).All datasets yielded prediction errors below 3.5%,with coefficients of determination(R2)exceeding 0.987,demonstrating excellent fitting performance and model stability.Furthermore,the Theil inequality coefficient(TIC)is employed to assess prediction accuracy under limited data conditions,achieving all TIC values remaining below 3%,confirming the robustness and practical reliability for applications of the proposed approach.