This paper addresses the issue of significant fuel measurement fluctuations in aircraft during non-cruising states, analyzing the reasons for these fluctuations and potential solutions. It proposes an optimized method for fuel measurement based on consumption prediction, incorporating actual flight data. Using machine learning algorithms, a fuel consumption prediction model is established, optimizing measurements without the addition of onboard sensors. Simulation analysis with out-of-sample data from the training dataset indicates that the optimized fuel measurement approach is effective, showing potential practical application value in engineering projects.