The problem of optical flow drift

Hello everyone. I encountered a problem while debugging optical flow, and look forward to your answer!

The optical flow module I am using is based on the optical flow output calculated by OPENMV, and relevant mavlink messages can be viewed in the optical_flow_rad.

What I am puzzled about is that in HITL simulation, the x and y outputs of the optical flow are manually set to 0, but no one has the opportunity to drift in the - y direction. I want to know why