A stable supply of power is critical element for reliable traffic signal operation. Battery backup systems are often used to prevent traffic signals from power disruption. Currently, battery backup systems are designed using engineering judgment due to a lack of operational requirements and well-defined performance measures. New high resolution traffic controllers now have the ability to record event based data for power failures and traffic counts at a tenth of a second resolution. Equipped with this high-resolution data, the paper proposes a performance-based investment programming using average annual signal downtime (AASD) over the analysis period as a performance measure. AASD is stochastically estimated using hazard-based duration models developed with power failure data. A volume and functional class weighted stochastic optimization scheme is then presented for a battery backup system project planning by sizing battery capacity to minimize AASD for a network under given budget constraints.