Look at stephen boyd’s answer at How to handle nonlinear equality constraints?#35 . Despite the thread name, that response addresses inequality constraints “going the wrong direction”. I offer no opinion as to whether that is the best way, or even a good way in your case.