The invention belongs to the field of magnetic abnormality detection, and specifically discloses a dynamic magnetic compensation method for a distributed magnetic abnormality detection system. The method comprises the steps: building a following optimized carrier magnetic field compensation model, determining a final parameter estimated value (shown in the description) of x and a non-parameter estimated value (shown in the description) of an independent identically distributed random subsample g(t) through a double-stage estimation method, carrying out the dynamic compensation for the carriermagnetic interference, and solving the magnetic field intensity after carrier magnetic compensation. On the one hand, the method enables a carrier interference magnetic field model to accord with thereality in a better way, and improves the accuracy of the model; on the other hand, the method also improves the parameter estimation precision and reduces the measurement error. Meanwhile, the methodimproves the performances of the distributed magnetic abnormality detection system through real-time high-precision dynamic compensation.