During our most recent Forecasting 101 workshop, one of the attendees expressed concerns about their data on electric vehicles (EVs) and how they do not think it is representative of the true number of EVs in their service territory. EV data is vital to producing accurate load forecasts, and the more representative it is of the true number of EVs, presumably the more accurate the forecasts will be. With many states vowing to mandate 100 percent of new vehicle sales be EVs by 2035, this data becomes even more critical. These current issues, mandates and conversations got me wondering about where to find the most accurate and reliable EV data at the lowest level of granularity. The following is a brief description of my findings.

First and foremost, the National Renewable Energy Laboratory (NREL) is an excellent source for EV stock and energy data. NREL has developed a model called the Transportation Energy and Mobility Pathways Options (TEMPO), which accounts for many factors, including income level, maximum education, dwelling type, and U.S. Census tract type. TEMPO is currently used to produce yearly county-level forecasts from 2018 out to 2050.

After spending some time on NREL’s website, I was able to find two useful EV datasets that leverage TEMPO. The first is an EV energy forecast dataset that contains yearly shapes, titled “Personally Owned Light Duty Vehicle Fuel Consumption”. The dataset contains two shapes: one which aligns with the Energy Information Administration’s (EIA’s) Annual Energy Outlook 2019 (Reference case), and another which aligns with the Electrification Futures Study (High Electrification case). The second dataset provides EV count forecasts under the same two cases. Both datasets contain data for four different vehicle types: battery-electric (BEVs), hybrids with a gasoline engine (HEV-Gasoline), internal combustion engine (ICEV-Gasoline), and plug-in hybrids (PHEV). These datasets are only available for light-duty passenger vehicles and are therefore not applicable to fleet or medium/heavy-duty vehicles.

The datasets can be found on NREL’s State and Local Planning for Energy (SLOPE) platform. From the home page, navigate to the data viewer, and select a dataset of interest from the “Layer Database” tab on the left-hand side of the webpage. The tool also provides geospatial visualizations of the selected dataset. Note that there are a variety of datasets available other than the ones I mention here. The datasets described here are available under the “Transportation” section of the Layer Database.

While county-level data is an excellent starting point and may be sufficient in a lot of cases, higher-resolution EV data may be more desirable, especially if it is believed that there is a higher or lower EV count than the NREL county-level data suggests. This is especially true for small utilities that may operate within a single city or municipality.

I also recently learned that some state departments of motor vehicles (DMVs) release data on vehicle counts by zip code. For example, the California DMV has data available for the past five years. Their datasets not only include EVs but hybrids and internal combustion engines as well. I was able to find zip code-level data through several other state DMVs including Oregon, Maine, and Minnesota, just to name a few. It should be noted that these are raw count data, and are not forecasts, unlike the NREL data. However, they may still be useful if a more granular level of EV data is necessary for your modeling endeavors.