Thinking with Data: An Example from Flagstaff

A few years ago, I wrote a post outlining what a data-driven approach to a topical city-related problem might look like. In that case, I was looking at sidewalks in Denver to develop an approach that would leverage data on traffic accidents and known pedestrian routes to prioritize the development of the Denver sidewalk network to improve walkability. I suggested some additional data sources that would provide further insight as a way to illustrate how data could be used to address a pressing problem.

A recent article in FutureStructure on the electrification of the bus fleet in Flagstaff, AZ presents another interesting opportunity to do something similar. In this case, the issue is vehicle-to-grid charging. The Mountain Line, a transit agency in northern Arizona serving Flagstaff and surrounding areas, is beginning the replacement of its hybrid-electric fleet with all-electric buses. Each bus is powered by a 444-kilowatt-hour battery. While the bus is plugged in and not in use, the electrical utility (Arizona Public Service) can pull power from the battery back onto the grid to meet any shortfall in supply. This shortfall often comes from the dramatic reduction in renewable energy generation from solar and wind as the sun sets and the wind dies down.

As the article states:

For utility provider and Mountain Line partner Arizona Public Service (APS), the new electric buses will serve as a vital research opportunity to gather data about the implementation of electric vehicle infrastructure around the state.

The article goes on to specify that APS is interested in exploring:

• The frequency of recharging and the placement of charging stations
• How charging schedules can take advantage of APS excess solar energy during daylight hours
• How the buses might be able to re-supply power to the grid

This is the jumping-off point I’d like to take for this post: how do you go about answering these questions? I won’t have any actual data to inform this exercise, so it’s going to be a notional exercise to go through the process, but I hope it will still be helpful.

The first is to understand the key goal. In this case, the buses need to have enough power to run their routes. The batteries on these buses are advertised to give them a range of between 100-130 miles per charge. The city is 66 square miles and the system has 9 routes cutting across it, so that may seem like a generous amount of range, but we still need to:?

1. Measure the length of the bus routes to estimate how many runs per charge a bus could make. This could be done by mapping the routes in a GIS (hopefully already done) and then measuring the routes. The goal here isn’t to get an exact number but to estimate the range so there’s a comfortable margin based on the advertised performance. The performance of the batteries will change with weather conditions and the age of the battery, as well as other factors, so it's important to be very conservative with this baseline estimate, but it will be helpful for the initial planning of charging stations as well as preparing for field testing.?
2. Test the routes with the buses to validate the estimates. Did the estimates account for idling time at either end of the route or along the way? Are there other demands on the batteries that need to be taken into account (A/C, heater, onboard announcements, etc.)? Ideally, the charge level can be tracked continuously for these tests so it’s clear how the battery level changes along the route. Starting with shorter routes will give time to calibrate the measurements and smooth out the testing process without accidentally running out of power on the bus. The data gathered on shorter routes can be used to better estimate usage for longer routes in the system prior to testing.
3. Determine the charging locations and arrangements for the buses on the system. Buses frequently idle, either at stops or at terminal points on their runs. This is used to maintain service intervals (known as head times) and provide drivers with necessary breaks. These are prime charging times for electric buses. Additionally, buses can be cycled in and out of service for shift changes or other requirements. In the case of electric buses, pairing routes with a high battery draw with a route that draws a lesser amount of power could economize on charging stops and duration. Setting this pattern of bus allocation, service times, and on-route charging will help optimize the usage of the fleet. Measuring the remaining charge as the buses come back to the yard for charging will also help calibrate the estimates, validate the system planning, and indicate where further testing or improved forecasting might be needed. If a bus went out with a 100% charge and was estimated to come back with a 30% charge but actually came back with 5%, then it’s important to understand the discrepancy to improve the estimate and possibly alter the system plan.
4. Use the estimates to set minimum charge levels for buses. Once it’s clear how much battery charge is being used on a given route, Mountain Line can set a minimum charge level on the buses leaving the yard, ideally with a cushion to account for any variance due to the intended route, weather conditions, and battery performance.?

The difference between this minimum charge and the full charge of the battery would be the excess APS could draw on to meet high demand when supply drops. This could be established (with a reasonable cushion for safety) as a fleet-wide standard (no bus leaves the yard at a lower charge level than the most draining route on the system would require), which simplifies fleet management and communication with APS.?

However, if certain buses are designated for each route and it’s possible to know with regularity the minimum charge level necessary for each bus, then it would be possible to do this bus by bus, assuming this could be communicated easily and reliably to APS (possibly programmed into the charging station by the fleet managers).?

Violating the cushion could be established as an emergency reserve that everyone understands could impact service delivery on the system if buses don’t have sufficient charge to run their routes, but a way to meet a serious need that threatens the electric grid.

There are several key things about this worth highlighting:

• It’s essential to remember the ultimate goal and mission of the transit agency: to get people where they’re trying to go. While it is good (and potentially lucrative) to implement vehicle-to-grid charging, providing this service to the power grid is secondary and needs to be structured in a way that ensures the primary mission doesn’t suffer from secondary obligations.
• The data is going to be highly variable and needs constant collection, updating, and alerting. This is like fuel economy estimates: so many things play into real-world performance that it’s almost impossible to do any detailed planning without seeing how things actually operate. The benefit is an optimized system with plenty of slack to ensure no bus becomes stranded for lack of charge but can also be used as a helpful reserve to ensure the reliability of our electrical grid.
• No commitments to APS should be made until this data is collected. Any estimates without real-world testing will be unreliable. That being said, once sufficient data is collected and the estimates start being more accurate (ie the cone of uncertainty starts to close), then there’s an opportunity to work out the agreements once it’s clear how much power Mountain Line has to spare.

Know the goal, test assumptions, and validate your hypothesis. These are at the heart of all data analysis and will service Moutain Line and APS well as they embark on this exciting new chapter in public transit and public utility management.

If you want to understand more about our approach to analytics and how we help public organizations grow their capacity to do this type of work, please drop us a line at [email protected].