Two neglected findings

There are two findings in the battery-saving algorithm paper from the University of Warwick that I want to write about in this post. Both were mentioned only in passing in the paper. Although these findings are crucial information for those who want to implement such a system in the real world, these were not mentioned in the conclusion nor in the discussion nor in the list of things they want to improve upon.

These are the two findings:

1. The number of electric cars that contribute to the vehicle-to-grid

   Remember, the principal functionality of that smart grid algorithm described in the paper is to minimize battery degradation. This is done by comparing the state of the battery with the state of the grid. If it is calculated that battery degradation would be less when discharging to the grid than when staying idle, then the car will discharge to the grid. Otherwise it will do nothing, not participating in the system.

   This means that not all cars that are connected to that system will participate, only those with batteries that would not degrade more by discharging. But how many cars would participate?

   To answer that question, let’s look at the assumptions that were made about the total number of cars that could participate in the vehicle-to-grid simulation (my emphasis):

   There are 54 car parks on the University of Warwick campus (including bay parking zones) with an approximate total capacity of 5560 car parking spaces. It is assumed that 120 of the maximum possible 5560 cars on campus are electric vehicles (approximately 2.1%, in line with the estimated UK market share for EVs in 2016 [2]) that can be connected to the universities electricity network.

   Basically, there are 120 electric vehicles on campus which seems to correspond with the estimated UK share for EVs in 2016. The simulation was not done with those 120 actual cars, but the researchers used a representative distribution of electric car usage from the Ultra-Low Carbon Vehicle Demonstrator Programme (ULCVDP). This is a program gathering usage data of electric car trials. With that data they ran their algorithm.

   This is the part of the paper where they describe the number of cars that could participate in the scheme (my emphasis):

   Of the 120 EVs connected to the electricity network only 46 participated in V2G; 74 are excluded because the smart-grid estimated that for those EVs V2G will degrade the battery more than if they were left idle. The capacity fade and power fade after 1 year of cycling for each of the 120 EVs is shown in Fig. 13; the smart-grid algorithm was able to reduce capacity fade by up to 9.1% and power fade by up to 12.1%.

   Apparently, there were 74 cars excluded that didn’t meet the condition. That is almost 62% that didn’t contribute because the algorithm decided that the battery would degrade more by discharging than by doing nothing!

   The good news is of course that the smart-grid algorithm was able to reduce capacity fade by up to 9.1% and power fade by up to 12.1% (that is where the 10% comes from). That is very nice when the main purpose of the system is to increase battery life, but whether that is useful for for example real-life grid load balancing is a different matter.

2. When are those cars able to contribute to the vehicle-to-grid

   There is nothing that guarantees that cars are connected to the grid at the time it is needed. This also happened in their simulation using the ULCVDP data (my emphasis):

   Since energy demand is only provided in 30-min time bins and the ULCVDP data suggests the morning peak for journey start times is 8am, the earliest time at which an EV can connect to the universities electricity network is assumed to be 8.30am. Hence, the demand peak on weekdays that commences at 6.30am cannot be suppressed until 8.30am.

   It is not hard to understand why these commuter cars can only partly suppress the morning peak. Those cars are still at home or are driving towards their workplace during the morning peak, so the charge in their battery can not be used. Since the morning peak in the IDL building starts at 06:30, the batteries will not contribute from that time until the cars start arriving from 08:30.

   It is not mentioned, but the same problem will also present itself in the evening peak, although not as pronounced in their simulation.

These findings will have consequences for real-life vehicle-to-grid systems. The main purpose of the battery saving algorithm in the paper is to minimize battery degradation and in that case it is easy to establish quite a gain (although I think that the 10% increase of battery life is an overestimation because of the use of unrealistic assumptions). But what if this algorithm would be implemented in a real-world vehicle-to-grid system having a different main purpose? I already encountered two vehicle-to-grid systems, the CREG model and the WeDriveSolar initiative of Utrecht. What will be the impact of the two findings on those systems?

• The CREG model

  The main purpose is to investigate the possible financial gain of participating in such a scheme and whether it is possible to replace some storage capacity in order to balance the Belgian grid. The researchers found that the electricity price would lower for the participating cars and these could replace first-order storage if there were 1 million electric cars and 85% of them would be connected to the grid at all times.

  If the battery saving algorithm would be integrated in that model, then this certainly would be a huge selling point (10% increase of battery life and gaining financially), but it would also require more cars to be connected to the grid. If roughly 62% of all cars would be excluded because the battery is in the wrong state, then it would require more than double the electric cars (about 2.2 million) to reach the necessary 850,000 participating cars.

  It also means that not all drivers will enjoy a reduced pricing when being connected to the system. If their battery would degrade more by discharging then it would not participate in the arbitrage.

• Smart Solar Charging Lombok in Utrecht

  The main purpose of this system is to optimize the use of the electricity produced by their solar panels. They developed a bidirectional charging station that can charge the vehicles using the electricity of solar panels during the day and discharge the battery to the network at night. Adding an algorithm that saves battery life of the participating cars will complicate matters when for example the algorithm calculates that charging during the day would degrade the battery more than not charging or when discharging at night would degrade the battery more than staying idle. Therefor more cars would need to contribute to have the same effect.

  This will eat away some of the efficiency of transferring the generated electricity from the solar panels to the grid, meaning that the algorithm will have a bad influence on the intended purpose of the system (using as much solar electricity as possible).

That is why I think the promoters of such systems will not be very eager to implement such an algorithm and would rather tolerate battery degradation of the participating electric cars.

Advertisement