Ramp metering is a transportation systems management and operations strategy that utilizes signals installed at freeway on-ramps to dynamically manage traffic entering the freeway. Ramp metering signals (RMSs) are usually activated during peak hours to alleviate recurrent congestion. However, recurrent congestion during peak hours constitutes less than half of all congestion. It is the non-recurrent congestion resulting from traffic incidents, work zones, adverse weather conditions, special events, and so forth, that adversely affects the performance of the freeway. Thus, this study used a three-regime model to develop guidelines to activate and deactivate RMSs during off-peak hours in response to non-recurrent congestion caused by incidents. A 10-mi section of I-95 in Miami, Florida, was used as the case study. The findings indicated that the RMS immediately upstream of the incident location might be activated when the average speed on the mainline drops below 45 mph and deactivated when the incident has been cleared and the average mainline speed reaches 45 mph for a consistent 5-min period. The RMS immediately downstream of the incident location may be activated when the average speed on the mainline drops below 35 mph for a consistent 5-min period and deactivated when the incident has been cleared and the average mainline speed reaches 35 mph for a consistent 5-min period. The proposed guidelines will enable transportation agencies to use ramp metering to improve traffic operations and safety during off-peak hours.