Cities across the United States are switching from diesel to fully electric or hybrid-electric buses [1], in part, to decrease transportation emissions. King County Metro Transit has the fifth largest electrified bus fleet in the US [2], and they report battery maintenance protocols that are reactive, rather than predictive, due to limited state-of-health information provided to the fleet operator [3]. Because real-time GPS data is available for all King County Metro buses through the OneBusAway API [4], we are developing a battery maintenance predictor that combines physics-based models for the real-time power cycle on each route, and data science approaches for the cumulative impact on battery degradation and maintenance. The physics-based model utilizes real-time locations and geographic information system (GIS) data to estimate instantaneous variables such as local road grade, estimated vehicle weight, velocity, acceleration, age, and weather conditions during operation to evaluate the real-time electric motor output for vehicles configured with the BAE Systems Hybridrive or Proterra’s electric drivetrain. A power system model calculates instantaneous battery pack power demand for each unique bus on the road. The data science models convert instantaneous data for the fleet into a cumulative state of health estimation for each vehicle, as the foundation for predictive maintenance. This presentation includes an outline of the battery-vehicle model, methods of measuring or determining variables of the model, and preliminary load profiles. We show how the dynamic forces on each bus are used to estimate the instantaneous battery pack power and appropriate drive cycle for different Metro routes. Accurate road elevation and grade data, a critical variable for estimating gravitational forces on the bus, surprisingly do not already exist as ground-truthed roadbed shape files in Seattle. This is true in many areas, so advanced filtering schemes are often needed for use with raw data [5]. We process high resolution Light Detection and Ranging (LiDAR) elevation data for multiple bus routes in the Seattle area and use that, along with typical instantaneous location data, to estimate the drive cycle and corresponding load profiles for different routes [6]. With this data, our first efforts have been to compare different Metro bus routes and rank them in order of expected degradation rate on the battery. We show some of the next steps in incorporating this model into a fleet management system, which can help transportation companies better predict when batteries need to be replaced and help create more efficient routes.