This paper proposes a methodology for the effective maintenance planning of traffic signal controllers, which are important for the safety of vehicles and pedestrians. For the effective maintenance planning, it is necessary to estimate the lifespan of a traffic signal controller. Since a traffic signal controller operates outdoors, the lifespan depends on the environmental factors including temperature, humidity, and salinity. We devise a deep learning model to estimate the lifespan of a traffic signal controller by using environmental conditions. The proposed maintenance planning methodology computes the optimal maintenance period which provides the lowest maintenance cost.