Overview
After launching the RechargeIT initiative, we noticed that employees primarily used our plug-ins for short trips close to our headquarters, so the data wasn’t truly representative of typical U.S. driving patterns. While the data from the GFleet plug-in hybrid electric vehicles (PHEVs) showed significant MPG improvements over our standard hybrid cars, we wondered what the data would look like if the vehicles were used for more typical trips. We were also curious to know how our plug-in vehicles would compare to other hybrid and non-hybrid vehicles that are typically found in the US fleet. Hence, the RechargeIT Driving Experiment was born.
In total, it took just over seven weeks to complete the driving test for all our vehicles. The PHEVs performed significantly better than the standard hybrids. And they greatly outperformed the average American fleet fuel economy of 19.8 MPG, with the Priuses getting more than 90 MPG and the Ford Escape SUV Plug-in Hybrid conversion getting 49 MPG during the test.
Setting up the Experiment
We started out by mining all personal vehicle trips from the National Household Transportation Survey done in 2001 and ended up with a database of around 547,000 trips. We used this data to create a representative set of trip lengths and frequencies that model actual driving behavior in the US for trips under 100 miles. The next step was to generate a set of routes corresponding to each trip length. The routes were generated using Google Maps, using the same starting and ending address (Google’s Mountain View headquarters) for each route. A total of 38 separate routes were generated. In the interest of modeling different types of routes, we created 12 city routes ranging from just over a mile to 30 miles, 14 combined city/highway routes ranging from 7 miles to 60 miles, and 12 highway routes ranging from 15 miles to 92 miles.
Using these routes, we then generated a set of 257 driving trips covering a total distance of 2228 miles. Our goal was to cover enough total distance to allow a minimum of three fueling stops for all the vehicles and minimize the effect of variability in fillups. This is a more significant factor in Priuses since they have a flexible fuel bladder rather than a conventional metal gas tank.
The final step was to set up a scheduling application for the drivers to use. We wanted to create a simple web-based application to make it easy on them, but we also didn’t want to spend a lot of time developing an application from scratch. After a bit of investigation, we settled on using Google App Engine to implement the scheduling application.
To gain additional insight into how the different vehicles would perform in city, combined city/highway and predominantly highway driving, we configured the scheduling application to group the city, combined and highway trips together so that all vehicles completed the set of city trips first, then the combined city/highway trips, and finally the highway trips. The trip scheduling application was also configured so that it would randomly select any available trip of the correct class. It also randomly assigned different vehicles to different drivers and avoided assigning drivers consecutive trips in the same vehicle whenever possible.
The Vehicles
Though the main focus of the driving experiment was to obtain “real world” use data for our PHEVs, we thought it would be interesting to do an apples-to-apples comparison of the plug-ins not only with their non-PHEV brethren, but with typical conventionally powered vehicles found in US households. We settled on a 2008 Ford Expedition SUV with a 4.6 liter V8, a 2008 Toyota Sienna minivan with a 3.5 liter V6, and a 2007 Toyota Corolla compact car with a 1.8 liter 4 cylinder to round out our eight vehicle test fleet. The remainder of the vehicles were pulled from the Google GFleet and consisted of a 2008 Ford Escape Hybrid, a 2008 Ford Escape Hybrid converted to a PHEV using a Hymotion 8kWh Li-Ion pack, a 2006 Toyota Prius, and two 2006 Toyota Priuses converted to PHEVs using a Hymotion 4.7kWh Li-Ion pack.
Though we would have liked to have all of the vehicles in our fleet outfitted with dataloggers to allow trip-specific data collection, the homegrown datalogger systems already present in our GFleet Escapes and Priuses wasn’t easily adapted for use in the Corolla, Expedition or the Sienna so we had to settle for just collecting GPS tracks and fueling data for those vehicles.
Driving the Trips
In the interest of maintaining a predictable schedule and for some measure of driver consistency, we elected to hire five professional drivers from a local shuttle service to carry out the actual driving of the trips. The drivers were instructed to drive all the vehicles in the same manner, e.g. modest acceleration and braking and adhering to the speed limit on all roadways. Although we briefed the drivers about hybrid features like regenerative braking, we did not train them on how to optimize their driving style for the hybrids since we felt this would give the hybrids an unfair advantage over the conventional vehicles, so we just emphasized the need to drive consistently across all the vehicles. We realized that there would be some variance in driver behavior, but since the drivers were rotated amongst the vehicles with each trip we felt that the effect of individual driving styles would be minimal. We also spent a few weeks in training mode, where the drivers got acclimated to the different vehicles, the GPS units and the different routes.
After the training phase was over, we established a production driving procedure for completing our set of 257 trips per vehicle. The procedure started with having a driver check out a trip in a particular vehicle using the web-based scheduling application. For redundancy, they’d first record the check out time and odometer reading on a paper log sheet. They’d then proceed to the designated vehicle, select the appropriate route in the Garmin GPS and proceed to drive the trip. After completing the trip, the driver would record the time in and the odometer reading in the paper log, and then enter the final data in the trip scheduling application. Driving the PHEVs required an additional step, wherein the driver would use a handheld instrument to check the state of charge of the Hymotion battery pack before starting the drive. If the pack wasn’t sufficiently charged, the driver would cancel the trip in the PHEV and check out a trip in another vehicle.
In total, it took just over seven weeks to complete all the trips in all the vehicles. Some of the vehicles ended up with unplanned trips for service that were included in the overall MPG results since it was impossible to exclude them, but these unplanned trips had a negligible impact in overall fuel economy so we don’t feel that they materially affected the resulting data.
From the “Your Mileage May Vary” Department
Though we did our best to control the variability in our experiment and have produced the most impartial/objective results in our apples-to-apples comparison as possible, we acknowledge that the results might have been different if we’d done the experiment outside the rather ideal weather and terrain conditions we have at the Google headquarters in Mountain View. In particular, the weather is generally pretty mild and thus the vehicle heaters were never used, and the A/C was only used moderately. In addition, the terrain in Silicon Valley is very flat and it wasn’t possible to include any variability in elevation/terrain except on the longer routes.
Due to the need to run the experiment in a finite amount of time, we also didn’t attempt to control the number of MPG-robbing cold starts done with each vehicle on a daily basis. If anything, the conventional and non-PHEV hybrids were at an advantage here, as they were more likely to get scheduled for a trip while the PHEVs were being charged.
Lastly, the experiment was designed so that the PHEVs were only scheduled for a trip if they had a sufficient charge to complete the trip with power from the Hymotion pack. This was done so we could show the full benefit of the plug-in vehicle for each trip, but in the “real world” PHEV owners probably wouldn’t always have their vehicles fully charged for each trip. That said, despite our planning the vehicles sometimes did end up driving trips without a sufficient charge, and some of the routes were long enough to completely deplete the Hymotion pack.