Table 5:  SAE J1634 Simulation Test Results

Donated Vehicle:

The University of Kansas

School of Engineering Design Project

A Sustainable Approach to Automobiles and Energy Infrastructure

EcoHawk II and Solar/Wind Energy Filling Station

This page contains the information regarding the second sustainable vehicle built at the University of Kansas intended for use by the KU Libraries in daily delivery activities.

Figure 1.  What the SUV looked like when it was donated

Data Analysis of KU Libraries Route:

As illustrated by the plots below, since the KU Libraries vehicle follows a similar profile each day and it only travels approximately 25 miles, this makes it the perfect route for an Electric Vehicle.  The route is very consistent from day-to-day allowing for an optimal vehicle to be built to suit the route. The vehicle is stored in the same location every night eliminating the majority of grid concerns and providing an optimal charging location.  Electric drivetrains are very well suited for delivery vehicles.  As the vehicle continues on the route throughout the working day, the capacity of the batteries declines, but at the same time the weight of the load is also declining as packages are being delivered to their destinations.  Electric vehicles also provide a reduced maintenance cost.  In addition to the lack of fluids and a reduction in moving parts, the operational components of an electric vehicle are more suited for delivery vehicles. The starters and batteries for the KU Libraries traditional I.C.E. delivery vehicles are under constant strain from continual starting and stopping.  This additional maintenance cost is completely eliminated through the use of an electric vehicle.  The KU libraries currently burn more than a half of a gallon of fuel per week idling.  This is due to their constant starting and stopping both on campus and at delivery locations.  This adds up to over 130 gallons of fuel per vehicle per year on idling costs alone!

Table 1:  I.C.E. model predictions versus collected data

Using data collected from an Auterra DashDyno and Garmin GPS18x, the driving profile of the Chevy S-10 truck of the KU Libraries (velocity and elevation) versus miles driven was generated as shown in Figure 2.  From this information, a model has been developed by the students in order to calculate the miles per gallon of the vehicle and compare it against the data collected from the DashDyno.  The model prediction via Table 1 illustrates that the simple model, using ME 636 computer programs with Dr. D’s help, does a decent job predicting the miles per gallon.  This validates the correct inclusion of the physics (drag, rolling resistance, brake specific fuel consumption, etc…) in the model and indicates that it is ready for Electric Vehicle model incorporation.  Please note that the I.C.E. model predictions have been refined through calibration of the brake specific fuel consumption map.

EV Model Validation:

Figure 2.  Data collected from the KU Libraries truck

The next step is to validate the model for Electric Vehicles.  To this end, we used the specifications for a 1997 Chevy S-10 EV that was actually produced and tested (ref 1 and ref 2).  Unfortunately, it is impossible to find the complete specifications on the powertrain on this older vehicle with respect to the batteries and motor.  Hence, we did the best we could.  In these documents it was mentioned that the S-10 uses a three-phase AC motor with a maximum power of 85 kW.  Looking online, we found a three-phase Azure Dynamics AC motor AC90 at 97kW that is intended for EV applications.  From the information presented on their website, we were able to create a 3D plot of motor efficiency for model use as seen in Figure 3.  With respect to the batteries, it seems that two types were used in the S-10 EV - Delphi and Panasonic.  However, the only data available about these batteries were the capacity at certain levels as shown in Figure 3 below.  Using the knowledge gained from EcoHawk I lead-acid batteries, a couple curve-fits of capacity as a function of amp draw were calculated based on a scaling of Discover EV31A-A capacity curves.  While these are guesses per se, they are educational assumptions based on the available knowledge.  This will lead to some errors in the model predictions, but the results should be in the ballpark and reasonable for our later predictions based on the KU Libraries truck drive cycle.

Figure 3.  Estimated motor efficiency contour plot and lead-acid battery pack capacity as a function of current draw

The first validation of the EV model is a prediction of the constant speed range of the 1997 Chevy S-10 EV at 45 and 60 mph.  Through a comparison in the energy usage, the rolling resistance, drag, electric motor and battery incorporation in the model can be validated.  As seen in Table 2 below, the model does a reasonable job predicting the experimental results.  The difference can be related to the fact that the weather conditions (temperature, wind speed) during the test were unknown and from the educational assumptions of the students above.

Table 2:  EV model predictions versus published values while vehicle undergoes constant speed range testing

The second validation of the EV model is a prediction of the acceleration of the vehicle.  The model input was enhanced in order to include a maximum current draw allowed from the batteries.  Again, since only a little information about the batteries is known, an educated guess is again needed.  From the experimental data taken during the acceleration test involving the Delphi battery in a reference above, a maximum power draw was measured.  From this and the nominal voltage capacity of the pack, a maximum allowed amperage of 320A was utilized for both battery simulations.  This falls in line with our EcoHawk I understanding as we sized the battery cables for 400A based on the Discovery battery specifications and the motor controller.  As Table 3 and Table 4 illustrate below, the model again does a reasonable job predicting the results.

Table 3:  EV model predictions using Delphi batteries while vehicle undergoes acceleration

Table 4:  EV model predictions using Panasonic batteries while vehicle undergoes acceleration

Figure 4.  Vehicle acceleration at 100% SOC

While the magnitude of the results are slightly off, the trends of model prediction versus SOC follow similarly.  This illustrates that the model is able to capture the influence of the decreasing capacity on acceleration.  When plotting the model results against the measured values taken with the Panasonic batteries in Figure 4, the model follows the shape and magnitude of the results quite well with an R² value of 0.994.  Since the students are making assumptions for motor efficiency, battery capacity, maximum battery amperage draw and a myriad of other input parameters, we feel that the model is performing correctly.

EV Model Prediction (Lead-Acid Batteries) - KU Libraries Driving Cycle with Chevy S-10

Before implementing LiFePO₄ batteries (discharge and charge characteristics) into the model, a few simulations were run using the Chevy S-10 EV vehicle with lead-acid batteries to see what would be expected as to the final State of Charge (SOC) at the end of the KU Libraries driving cycle.  Since they are currently using a Chevy S-10 in their daily delivery of materials, it reasons that the validated model in the previous section will provide a good estimate of what to expect in real life.  There are two caveats to the model predictions presented in Figure 5.  One is that all accessory draws, like heating and cooling, are not included in the model predictions.  Hence, if everything else was correct, the model will over predict the final SOC of the vehicle.  Finally, we are assuming that the vehicle will be driven exactly the same using the EV powertrain as the I.C.E. powertrain.  Obviously, this will not be the case; however, until we actually build the vehicle and measure all important parameters, this is the best we can do at this state of the project.  The model will give us a good idea of what to expect and will help us size the appropriate LiFePO₄ pack.  Without model predictions, we would just be stabbing in the dark as to the capacity (A-h) of our pack and either undersize it (bad for KU Libraries) or oversize it (bad for economics).

As the figures show, the Panasonic batteries perform much better than the Delphi batteries which makes complete sense according to Figure 3.  They have a higher capacity indicating that they have more energy stored and will not drain as much during the driving cycle.  Finally, during the day of the longer driving cycle (09/13/10), the Delphi batteries will be drained completely illustrating that they would not be suited for this usage.  As previously shown for the SAE J1634 driving cycle, using Delphi batteries should last 43.8 miles.  A couple points of interest: believe it or not KU is quite hilly and the Libraries driving cycle undergoes significant stop and go driving; hence, we believe it is a more rigorous usage of the batteries over SAE J1634 and we should see a lower range for the Delphi batteries.  And this is reason why models are important - otherwise we would just go on manufacturer specifications and undersize the pack for our needs.  At this stage, we would estimate that a Chevy S-10 EV would fall in-between the two battery lines given our assumptions.

Figure 6.  Model predictions of lead-acid battery based Chevy S-10 EV for KU Libraries driving cycle.

Up to this point, the model has been performing admirably with respect to battery discharge for power.  However, one of the significant benefits regarding hybrid and electric vehicles is the ability to use regenerative braking.  This requires including the specifics of battery charging; for example, lead acid batteries have three phases of charging of which the last phase is not required to be modeled for EV applications on the road (float charging).  Moreover, modeling regeneration involves estimates of the fraction of braking (friction vs. regeneration), rectifier and DC-DC converter efficiency.  From the data found and through other research, estimates on the conservative side were made and the model was run through the SAE J1634 Electric Vehicle Energy Consumption and Range Test Procedure that consists of repeating the UDDS driving cycle.

Figure 5. State of Charge over SAE J1634 for Delphi battery pack.

As illustrated in Table 5, the model again performs well when regenerative braking is included in the calculation.  As the figure shows, regenerative braking does help slightly; however, lead-acid batteries must charge much slower than their lithium brethren; hence, regenerative braking only provides a few miles of extra travel.  Note that exact replication is not the intended goal of the model as not enough data exists on the components utilized.  Instead, if the results of the model are “in the ballpark” of the experimental data found, the model can be effectively used in a predictive manner later.

EV Model Prediction (LiFePO₄) - KU Libraries Driving Cycle with GMC JimmE-V

The model is now ready for examination of the KU Libraries driving cycle using the GMC Jimmy vehicle that we are converting.  The model input was changed to reflect the Jimmy (curb weight, aerodynamic drag) and a number of assumptions were made:

· Driver of standard weight (200 lbs).

· Three-phase motor considered (Azure Dynamics AC55) is approximately same weight as internal combustion engine that has been removed.

· Battery pack consists of 100 LiFePO₄ batteries + 1.10 scaling factor on weight (so, battery pack weight was added to vehicle with 10% extra to account for components).

· Motor has regenerative braking capabilities; hence, estimated 30% regenerative fraction was included in model.

· LiFePO₄ charging and discharging curves included & temperature dependency added (cold day will reduce Ah availability).

· Previous Chevy S-10 truck data references used 51 psi for the tires.  This was reduced to 35 psi to reflect standard pressure (yes, the model does have the fidelity to change the results based on tire pressure).

 Some items still not included in model:

· Package weight - we need to measure the standard weight of the packages being delivered on a daily basis from the Libraries.

· Accessory draws - we are currently sizing our HVAC system and should be measuring power draws from the components at some point in the Spring 2011 semester.  This will then be included in the model.

Running the model with these assumptions should provide further understanding of the battery pack capacity needed.  Moreover, it will give us an upper estimate of the State of Charge at the end of the day.

Looking at the above figures, we see that the lithium iron batteries perform better than the lead-acid batteries, but not substantially better than the Panasonic batteries.  This can be attributed to the fact that the students selected 60Ah nominal capacity batteries that are actually quite similar to the Panasonic battery capacity (Figure 3).  Therefore, the difference between these lines is largely a function of the weight saved and the regenerative current that the batteries can handle.  For the Panasonic batteries, regenerative braking saves about 2.7% SOC; whereas, for the lithium batteries we see a savings in 3.3% SOC for the 09/03/10 data.  This is probably conservative given our assumptions; however, better to be conservative rather than excessive.  It appears that the 60Ah, lithium iron batteries will potentially work but more simulations are required with the actual GMC Jimmy parameters while also including accessory draws.  The next option in batteries would be 90Ah that would definitely work; however, that will increase the cost of the battery pack substantially.

Figure 7.  Comparing lead-acid and lithium iron batteries for KU driving cycle using Chevy S-10 EV base.

Table 6:  Comparing current I.C.E. truck with model predictions for 09/03/10 test.

* Values are from coal power plant use during EV charging

Using information from the DoE’s Energy Information Administration, the model can predict the equivalent miles per gallon of the vehicle along with the greenhouse gases emitted.  Since the power plant north of Lawrence uses coal, the corresponding values from EIA’s database were used in predicting emissions from the vehicle.  As Table 6 elucidates, converting the Chevy S-10 of the KU Libraries truck to a modern EV will dramatically increase the efficiency of the vehicle while substantially reducing the emissions.

EV Model Prediction (LiFePO₄) - KU Libraries Driving Cycle with Chevy S-10

At this stage, the model is ready for incorporation of lithium iron batteries.  From our research, we found a motor that will work for the GMC Jimmy and LiFePO₄ batteries with a nominal capacity of 60Ah.  To see how this powertrain performs for the KU Libraries use, we used the same vehicle parameters as the previous section but changed the battery pack to the lithium batteries.  Hence, we subtracted the weight of the lead-acid battery pack and added the weight of the lithium pack.  This assumes that all other weights would be equivalent (like electric motor).  This is one large benefit of lithium batteries, a significant reduction in weight because the batteries are more energy dense.  Moreover, they are able to handle a higher amperage during regenerative braking; so, we should see more energy recovered during regenerative braking.

Figure 8.  GMC JimmE-V predictions using different pack sizes, ambient temperatures and tire pressures.

As illustrated in Figure 8 above, the model works as it should.  As we decrease the tire pressure and ambient temperature, the SOC of the vehicle decreases.  When running the model for the longer day (09/13/10) using a 60Ah pack, we see that the  SOC at the end of the day is nearly zero under our worst case scenario.  Because of this, we simulated a higher capacity 90Ah pack and found that we should still have approximately 35% of capacity.  Figuring package weight and accessories, this should be sufficient and will be checked later.  The next significant battery capacity option would be 120Ah; however, we are reasoning that the added cost would outweigh the benefits.  Plus, as a student project with limited funds, we cannot afford a 120Ah pack; therefore, we will complete the build with the 90Ah packs as they are within our budget.  Moreover, we will attempt to reduce some of the weight of the vehicle by removing the second row of seats and will look into low rolling resistance tires.

Table 7:  JimmE-V model predictions without packages and accessories (0 deg C, 35 psi, 90Ah)

* Values are from coal power plant use during EV charging

From the model predictions in Table 7, we feel that the JimmE-V will achieve around 60 mpge based on the driving cycle, a substantial increase over the current 11.5 mpg that the KU Libraries are getting.  Moreover, even though there is the potential to fill up the vehicle using the coal plant north of Lawrence, we will still achieve a significant reduction in greenhouse gases.  We will further offset this by recharging the vehicle at our solar energy filling station that we will enhance with a wind turbine.  As of 12/5/10, the wind turbine has been selected and a feasibility study is being performed by Design, Construction and Management on KU’s campus in order to ensure that we can put it right next to our barn.