Time-Dependent Dielectric Breakdown (TDDB) testing is a crucial method for investigating the reliability and predicting the lifetime of power device gate oxides. In traditional lifetime prediction, the Weibull distribution of failure time often exhibits distortion in the Weibull slope, which impacts the accuracy of lifetime prediction. Firstly, a TDDB experimental platform was developed to enable long-term continuous online measurement of gate leakage current, resulting in more precise failure time data. Secondly, an analysis of the Weibull distribution results for failure time and failure charge was conducted based on percolation model theory and E-model theory for lifetime prediction, revealing that internal accumulated charge in the gate oxide can better reflect device failure. Finally, through theoretical analysis, hypothesis testing, and goodness-of- fit testing for Weibull distribution, we compared and analyzed the representative effects of offline gate leakage current, threshold voltage, and composite flat band voltage on accumulated charge. The findings indicate that composite flat band voltage can more accurately assess the progression of time-dependent dielectric breakdown failure.