Abstract: Aiming at the corrosion problem of gas field gathering pipelines, a prediction algorithm of internal corrosion rate based on the GRA-IFA-LSSVM combined model was proposed. The theories of GRA (Gray Relational Analysis) model, IFA (Improved Firefly) model and LSSVM (Least Squares Support Vector Machine) model were introduced, and the combined process and evaluation indexes of the combined model were proposed. The prediction accuracy of the GRA-IFA-LSSVM combined model was verified, taking a domestic pipeline as an example, and was compared with those of the other common prediction models. The results show that temperature, H₂S content, CO₂ content, pH value and flow rate were important factors affecting the corrosion of gas field gathering pipelines. When the GRA-IFA-LSSVM combined model was used to predict the internal corrosion rate of gas field gathering pipelines, the average absolute error was 1.946%, the root mean square error was 1.496%, and the absolute coefficient was 97.53%. The three evaluation indexes of the combined model were all smaller than those of the other common prediction models. The GRA-IFA-LSSVM combined model had strong accuracy, robustness and advancement in the prediction of internal corrosion rate of gas field gathering pipelines, and could provide data support for the protection of gas field gathering pipelines.