Results

One of the interesting features of the end-user App of ELECTRIFIC is that it allows to introduce easily all the details needed to provide the user with the best solutions, whether if he/she is an electric car sharer or an electric car owner. Car sharers do not always know the specifications of the vehicle they are driving, but it is as easy as to enter the license plate. On the other hand, if they are owners, they can select their vehicle in the list and adapt the details to those of their vehicle.

A direct way to ease grid-pressure, decrease an EV’s CO2 – emissions and increase its range is more energy efficient driving. Driving smooth rather than acceleration-prone saves up to 50% of energy – in EV’s and CEV alike.

To help maintain a smooth style, EV’s are equipped with an energy-efficient driving modes that limit acceleration and discourage high speeds. A trial with more than 200 monitored drives in Renault Zoé’s tested two ways to promote the use of the eco-mode.

Defaults:

When the mode was activated at the beginning of the drive, i.e. when it was the default for the drive, it was active 70% of the drive time. In contrast, when the mode was not the default for the drive, it was active only 10% of the drive time. ELECTRIFIC solutions will therefore be conscious about setting defaults.

In this trial analysis, two EVs of type Nissan Leaf are 

used. They have the same age, mileage and battery health (SoH, State of Health). The homogenous EV usage is guaranteed, as both EVs are used by colleagues of THD (Technische Hochschule Deggendorf) to commute to work daily and each week they switch the EVs between them.

The trial consists in seeing the influence of the type of charging. That is why one of the EVs is strictly charged slow and the other one fast. Slow means power less or equal to 3.65 kW and Fast is power equal or more than 20 kW. Both EVs are equipped with data acquisition systems of THD in order to analyse driving data later on.

0


Hypothesis: Fast and slow charging influence battery degradation differently

It is assumed, that fast charging is worsening the battery health state

On the contrary, slow charging could prolonger or improve the health

0


click to enlarge

FIRST RESULTS

The graph on the right shows the change of the SoH in steps of two-weeks

Difference of SoH between slow charged and fast charged EV (e.g. positive values: SoH of slow charged EV higher then SoH of fast charged EV)

Ambient temperature development in two-week steps
(measured by the EVs)

After a half year, the battery of the fast charged EV reaches its initial health state

During the same period, the SoH of the slow charged EV battery improves by 1 %


TEMPERATURE

SoH decreases, if ambient temperature is high

SoH increases, if ambient temperature is low

0000

0000

0000


CHARGING

If temperature is high, slow charging is improving SoH

Ff temperature is low, fast charging is improving SoH

0000

0000

0000


PREDICTION for second half of trial

Fast charged battery SoH could improve during winter, but drop in the spring

It is assumed that fast charging is more harmful to EV batteries than slow charging

The slow charged battery SoH could also slowly decrease

Dynamic pricing of charging services can help balance the needs of EV drivers, charging station operators and grid operators. As an input, the charging station receives charging requests created as users plans their trips. Based on the predicted demand, existing bookings, availability of power and its cost, the pricing algorithm determines the price for the incoming charging requests. Users then decide, whether they accept the offered price or whether they will charge elsewhere or at different time. The dynamic pricing model is based on solving Markov Decision Process, framework for decision making in stochastic environments.

Dynamic pricing using Markov Decision Processes achieves much better allocation efficiency and revenue than currently most common flat rate pricing methods. Changing price is used to incentivize charging when charging resources are plentiful and discourage charging when they are scarce. Additionally, dynamic pricing can proactively help prevent electrical grid failures with same incentives.

http://transport.felk.cvut.cz

As an input, the EV user needs to define his/her activities. The activities are defined by their time (duration, earliest start, latest end) and spatial attributes.  The planning algorithm then selects the best ordering of the activities together with the locations and routes between the activities. The algorithm also considers the limited range of the EV and selects the best option where, when and how (fast/slow) to charge. The algorithm is a heuristic forward state space search based on multi-objective A* and is able to find the best plan according to one of many possible criteria (time,cost,…).

The whole day planning approach saves time and money in comparison with the naïve single-trip approach and as a consequence also decreases the amount of used energy.

http://transport.felk.cvut.cz

ARCHITECTURE

In the context of ELECTRIFIC, we are developing a decentralized load management controller that takes the power quality of the grid into account. This controller provides decentralized and automated load management at the level of the EV Charging Station (CS). Measurement devices that are installed in the low voltage grid collect data that are used by the controller in order to make decisions about the current situation in the grid in terms of power quality and to determine the best reaction from the CS. This controller can be seen as an extension of any existing Charging Station Management System (CSMS) or as a standalonecomponentsince it can speak the same language between CSMS and the CS (i.e. using the OCPP protocol). We call this controller “Smart Charger”. To achieve this goal, we propose an architecture that contains three components as depicted in Figure:

  1. A real-time handling of big data streams collected by measurement devices (Apache Kafka).
  2. A component that indicates the status of the grid based on the collected data (PQ Indicator).The design of the PQ-Indicator adapts the traffic light model with three colors that specify the electric vehicle charging capability of the grid.
    • Green: The situation of the grid is stable
    • Yellow: The power quality of the grid is not optimal but still above or below a certain critical threshold
    • Red: The situation of the grid is critical
  3. A controller that changes the used charging capacity regarding some recommendation from PQ-indicator (Smart Charger).

The proposed smart charging algorithm is:

  • Decentralized
  • Fair
  • Avoid power quality problems that arise of drastic changes by smoothly adopting the charging power capacity of the charging process.

Testing the proposed architecture is done In the FlexEVLab, which provides all required hardware and software components as well as the AIT Lablink middleware for PHIL (Power Hardware in the Loop), coupling for steady state co-emulation. In this lab, we have a CS that can be controlled using a simple Modbus controller. The possible control options are start/stop charging operation and changing the charging current.

©AIT Austrian Institute of Technology GmbH

The result of our smart charging algorithm is compared against two baseline scenarios:

  1. All charging stations are charging for the whole time period with their maximum charging power
  2. No charging station is charging at all.

We compared both the voltage level at the critical point and the transformer load in the aforementioned scenarios.

The figurebelow shows only the voltage level at the critical point

In ELECTRIRIFC, we develop a flexibility rewarding scheme for grid-friendly charging processes. This part is designed as proactive schedule of charging processes in order to best fit the available power capacity in the power grid.

The concept includes the differentiation between guaranteed power and flexible power. Guaranteed power represents the baseline whereas flexible power can be used to stabilize the grid and will be rewarded or penalized in case of misuse.

The reward scheme is requested and only valid for specific charging spot, which from grid perspective is the grid connection point of a set of charging stations. The operator of that charging spot, e.g. the CSP, is responsible for splitting the demand to the available charging stations/connectors at this charging spot.

Visualized, a diagram with the following properties can be imagined as the output of the Reward Scheme: The x-axis represents the time, and the y-axis shows the available maximum charging rate of single charging operations in kW (forecasted values). Each charging operation is divided into one or multiple grid-friendliness options, which include a specific grid-friendliness factor. The grid-friendliness factor is defined in the range of [−1, 1], where −1 means worst and + 1 best grid-friendliness. A neutral grid-friendliness factor of 0.0 is always assigned to the guaranteed option as it represents the reference value. Within a certain grid-friendliness option (charging power range), the reward stays the same. The maximum charging rate is limited by the remaining grid capacity (e.g. cable, line, transformer, and bus bar) and the power quality (e.g. voltage range, flicker or harmonics) in the grid.

The output of the so called FlexibilityReward Schemecan further be used by the CSP to optimize its charging offers to the end user.

More information about this concept can be found here.