Abstract: P92 steel is a critical structural material for high-temperature components in advanced nuclear reactors and power plants, where its long-term performance under creep conditions is a primary concern for operational safety and structural integrity. The Kachanov-Rabotnov creep constitutive model (K-R model) is a widely accepted framework for describing the creep damage evolution, particularly the tertiary creep stage leading to fracture. However, the predictive accuracy of K-R model is critically dependent on the precise identification of its five material constants, a task that is often challenged by limited experimental data and the inherent limitations of conventional fitting methods. This study aims to develop a comprehensive and high-fidelity framework for establishing the K-R model for P92 steel at 600 ℃, focusing on overcoming the shortcomings of both parameter identification techniques and the model’s intrinsic structural deficiencies. To achieve this objective, a systematic investigation was conducted. First, a series of uniaxial constant-load creep tests were performed on P92 steel specimens at 600 ℃ under three distinct stress levels (180, 200, and 220 MPa) to obtain the complete creep strain-time curves until rupture. The conventional analytical method, which involves a decoupled, step-by-step fitting procedure, was initially employed to identify the K-R model constants. Subsequently, to address the error propagation issues inherent in the analytical approach, a multi-objective genetic algorithm (MOGA) was implemented. This global optimization technique simultaneously identified all five material constants by minimizing the discrepancies between model predictions and experimental data across three objectives: the overall creep curve, the creep rupture time, and the creep rupture strain. Upon discovering a consistent systematic deviation between the model predictions and experimental results even with optimized parameters, a data-driven time-scaling correction methodology was proposed. This method introduced a stress-dependent correction function, derived from the ratio of experimental time to predicted rupture time, to adjust the model’s temporal axis without altering its underlying physical formulation. The results demonstrate a clear hierarchy in the performance of the methods. The analytical method provides a poor fit to the experimental data, yielding R² values as low as 0.29 and failing to capture the creep behavior accurately due to accumulated errors. In contrast, the MOGA method significantly enhances the parameter identification accuracy and robustness, with R² for the 180 MPa and 220 MPa stress levels improving to above 0.93. Crucially, both methods reveal a shared systematic deviation at the 200 MPa stress level, suggesting an inherent limitation of the classic K-R model in capturing the specific creep mechanisms of P92 steel under these conditions. The application of the time-scaling correction function proves highly effective. The final, MOGA-based corrected K-R model achieves exceptional predictive accuracy across all tested stress levels. It successfully covers the entire dataset, with R² consistently exceeding 0.97 and a substantial reduction in the residual sum of squares (RSS), indicating an excellent agreement with the experimental observations. And the model’s extrapolation capability was validated using an independent 160 MPa dataset. For a rupture life of 2 974 h, which exceeds the training set duration by more than six times, the corrected model reduces the life prediction relative error from 24.48% to 2.45%, and improved the R² from 0.039 to 0.806. In conclusion, this research successfully establishes a robust hybrid framework that integrates an advanced optimization algorithm (MOGA) for reliable parameter identification and a pragmatic correction function to compensate for the constitutive model’s intrinsic limitations. This framework not only delivers a highly accurate and reliable K-R model for P92 steel but also presents a valuable method applicable to other materials and constitutive models facing similar challenges. The developed tool offers a precise basis for the creep life assessment and structural integrity analysis of P92 steel components, thereby contributing to the enhanced design reliability and operational safety of high-temperature engineering systems.