 Research
 Open Access
Enhancing power quality in electrical distribution systems using a smart charging architecture
 Ammar Alyousef†^{1}Email author,
 Dominik Danner†^{1},
 Friederich Kupzog^{2} and
 Hermann de Meer^{1}
 Published: 10 October 2018
Abstract
The electrification of the mobility sector comes with multiple challenges such as the lack of information on when, where, how long and how fast charging processes of electric vehicles will take place. In order to keep up with increasing power demand of charging processes, besides better predictions also the active control of charging processes will be necessary to minimize infrastructure costs. This work deals with a realtime mechanism for supporting the Power Quality (PQ) in electric distribution grids in terms of congestion and voltage management. In the paper, we propose a distributed smart charging approach that considers realtime conditions of the distribution grid provided by an eventdriven architecture that collects data from different points in the grid. Our approach adopts the traffic light model, which allows smooth changing of the charging power to avoid drastic changes of the grid state. In order to be ready for realworld application, the algorithm is validated by a series of experiments on two setups: Pure software (co)simulation and Power Hardware In the Loop (PHIL) where physical charging stations and electric cars are controlled in a laboratory setup.
Keywords
 Smart charging
 Power quality
 Electric vehicle charging
 Charging station
 Voltage control
 Traffic light model
Introduction
New requirements on low voltage distribution grids have to be fulfilled due to an increased number of renewable producers, but also due to electric vehicles as new network participators. This goes along with a paradigm shift. It can be expected that with increasing connection of electric vehicles supply equipment in distribution networks, infrastructure dimensioning can no longer be based on worstcase conditions in all cases. Stateoftheart 22 kW charging power (Longo et al. 2016) by far exceeds the 4 kW estimate for a residential grid connection in central Europe. Consequently, more online monitoring and even active interventions during grid operation will be necessary to maintain critical boundary conditions such as line voltages and asset loading within safe limits.
Undoubtedly, guaranteed immediate and fast charging can only be realized with sufficient grid capacities at the connection point, which are e.g. required for public Charging Station(s) (CSs). However, for the expected large number of private or home CSs, a “smart approach” that makes use of currently available excess capacities can help to reduce grid connection costs. Therefore, this paper proposes a solution for an active network operation at the low voltage level.
While a wide spectrum of charging management algorithms have been proposed in literature already, e.g. (Lopes et al. 2009; Cortés and Martínez 2016; Pipattanasomporn et al. 2012; Fan 2012; Li et al. 2014), most approaches do only consider line loading and the resulting scheduling problem. The algorithm developed in this work is intended not only to avoid asset overloading but also to improve PQ parameters such as node voltage variations. This is analyzed in a field test region in Bavaria, Germany, where the system is practically operated by a Distribution System Operator (DSO).
For achieving a reliable implementation for field use of the algorithm in short time, we follow a rapid prototyping approach for networked smart grid systems based on cosimulation and hardwareintheloop testing (Faschang et al. 2013). This approach allows seamless and stepwise migration from a simulated environment to a laboratory evaluation with physical charging stations and ecars, and finally closedloop field operation. Key to this approach is the message passing middleware AIT Lablink^{1}. The developer of the algorithm is always using the same interface to the physical hardware, while Lablink routes messages to simulated, emulated or real components. In this paper, simulation and laboratory results are presented.
The remainder of this paper is structured as follows: In “Related work” section we discuss related work. The proposed architecture is described in “Architecture” section, after that, we introduce the designed algorithms in detail in “Algorithms” section. The results of different scenarios using pure (co)simulation and PHIL are presented in “Evaluation” section. Finally, we highlight the future work and conclude the paper in “Conclusion and future work” section.
Related work
Potential impacts of introducing a large number of Electrical Vehicle(s) (EVs) to the power distribution network have been studied extensively in the literature and many ideas have been introduced to use the (EV) penetration for supporting the grid stability and power quality.
Approaches of charging management

Challenges in terms of power quality are tackled by the design of a new charging connector or a completely new design of a charging station with power quality compensation for electric vehicles as in Tanaka et al. (2012), Restrepo et al. (2018), Vahedi and AlHaddad (2016), Zhong et al. (2017), Yong et al. (2015). In contrast, we solve the problem using the offtheshelf hardware and software and validate our proposed architecture with hardware in the loop simulation.

Scheduling algorithms have been proposed to shift the EV charging load to offpeak hours, thereby avoiding branch congestion and voltage drop in the distribution network. Most existing work suggest a centralized controller. For example, authors of Chung et al. (2014) propose masterslave control scheme for Plugin Electrical Vehicle(s) (PEVs) smart charging in purpose of increasing the number of PEVs that can be plugged into a single circuit avoiding grid bottlenecks. Other centralized solutions are investigated by the authors of Lopes et al. (2009),Deilami et al. (2011). However, as discussed in a white paper (Taft and Martini 2013), coordinated control at different levels of a hierarchical distributed system such as the power grid becomes infeasible with such centralized control.

Instead of using a centralized approach, some authors propose a decentralized or hierarchical charging schedule (Cortés and Martínez 2016; Rivera et al. 2015; Alonso et al. 2014; Kong et al. 2016). Most solutions are offline algorithms that make decisions based on collected grid data 24 h ahead of time. Furthermore, they consider realtime load balancing as the only grid stability constraint and completely ignore voltage control.

The distributed control algorithm proposed in Ardakanian et al. (2013) adapts the charging rate of EVs to the available capacity of the network while ensuring that network resources are used efficiently and each EV charger receives a fair share. Their algorithm requires a heavy and synchronous communication overhead and only considers the stability of the grid in terms of load balancing, ignoring voltage control completely.
Approaches of Charge Control
Another way of increasing the penetration of EVs into the power grid is establishing a controlled charging process that reacts in realtime on changes of the different local or global parameters of the grid. Authors in Lehfuss and Nöhrer (2017) discuss three different types of charge control approaches: localvoltage driven, centralpower driven and a combination of both. In contrast, our architecture considers a dynamic change of the charging power regarding to different situations of the grid in a more advanced way. A (PEV) charging policy is proposed in Foster et al. (2013) that considers transmission and distribution integration issues and reacts to market signals across time scales and systems. Furthermore, voltage support for the distribution network is introduced in terms of allowing increased penetrations of distributed Solar Photovoltaic (PV) solar arrays. The authors consider only the local voltage near to the CS and ignore the overall state of the low voltage grid and the fairness between running charging processes, which are the main concerns of our proposed architecture. Other solutions propose local smart charging algorithms based on a droop controller (Martinenas et al. 2017; Álvarez et al. 2016) and mitigate line voltage drops and voltage unbalances, without relying on any vehicletogrid capability. These solutions are limited on estimating the voltage locally without considering the situations at other critical points in the grid which probably need different reactions at certain times.
Grid constraints versus energy markets

Green  Market Phase: No critical network situations exist and no intervention of the DSO in the market.

Amber  Interaction Phase: Potential or actual network shortage in the defined network segment exists and the DSO utilizes the flexibility offered by market participants to mitigate the damage.

Red  Network Phase: The DSO must intervene directly to remedy the direct risk to the stability of the system.
The authors of Deutsch et al. (2014), Medved et al. (2016) propose an implementation of the yellow state based on forecasts. In case of a predicted power quality problem, the market mechanisms are used to buy flexibility for this time window. An updated version of the traffic light concept is introduced in Medved et al. (2016) which may be used by the DSO to control Demand Response (DR) units. The proposed approach depends on information from power flow calculation based on the joint load schedule of DR units and the residual loads.
The proposed approach in this paper can be seen as an implementation of the amber state of the aforementioned traffic light model, that depends on realtime conditions and creates its own colored states by predefined thresholds. In this way, the flexibility introduced by emobility sector can be used more efficiently considering the requirements of both the grid and the running charging processes.
Contribution
The large majority of related work studies scheduling for peak power reduction. However, besides line and transformer loading, voltage constraints play a significant role in hosting capacity restriction of European distribution grids (Varela et al. 2017). Therefore, our work differs from the aforementioned categories in the following points: We propose a completely distributed smart charging approach by considering the realtime conditions of the grid using an eventdriven architecture to collect data from the grid. Additionally, our approach considers a smooth change of the charging power capacity to avoid drastic grid state changes. As input parameters of our smart charging solution we consider both the overloading of grid elements (specifically the transformer and feeder lines) and voltage magnitude at certain points in the grid, e.g. at the charging station or at critical points in the grid. Furthermore, a rapid prototyping approach for networked smart systems is followed (Faschang et al. 2013) in order to test the proposed architecture and ensure a safe deployment in realworld environments.
Architecture
The objective of the proposed architecture in this paper is to stabilize the grid and its power quality. The proposed mechanism complies with two design criteria. Firstly, it needs to be scalable in terms of number of involved CSs. Secondly, it is based as much as possible on locally available data at the CS, such that it can even react in case there is a communication problem with the monitoring mechanism of the grid. Hence, the proposed architecture is distributed and located on the actuator side which in our case is the CS.
In order to monitor the power quality, it is essential to measure voltage, current, frequency, harmonic distortion and waveform at different points of the grid (Sankaran 2002). In this paper, a monitored point is referred to as a Measurment Point (MP). In this regard, power quality is indicated by Key Performance Indicator(s) (KPIs), e.g. voltage level or overloading of grid elements, such as the transformer or feeder lines. Furthermore, these (KPI) classes are computed/measured directly at MPs in realtime, e.g calculation of Root Mean Square (RMS) values, or are computed using multiple measured KPI values at different measurement points, e.g. the minimum voltage on a certain feeder line.
However, the collected KPI values are forwarded to controller components, that are located at CSs. The responsibility of theses controllers is to indicate the present status of the low voltage grid and to choose appropriate actions in order to mitigate stress on the power grid arising from emerging power quality issues. As depicted in Fig. 1, power quality estimation is performed by a component called PQIndicator (In this paper, we use PQ as an abbreviation of the Power Quality), which responds to triggered events from Kafka. For example in case of voltage fluctuations, which refer to degradation of the power quality, it gradually estimates the power quality and asks the socalled Smart Charger (SC) to decreases/increases the charging rate in order to counteract the voltage fluctuation and, hence, improve the power quality. This power quality indication, called ’PQIndic’, is defined within the range of [−1,1]. Within this normalized range, (1) corresponds to either a complete shutdown of the charging process or a reduction to the minimum required power in order to be able to control the EV later. In contrast, the value (+1) represents the maximum power capacity of the CS. The Smart Charger applies a smooth or drastic change on the used charging power capacity depending on value the PQIndic. The reason behind separating power quality indication and control logic is due to different interests of the involved parties. From the Charging Service Provider (CSP) perspective, the PQIndicator is a black box, which is configured by the DSO depending on the characteristics of each low voltage grid individually, e.g. applying different thresholds for voltage boundaries. In contrast, the CSP configures the Smart Charger according to its business model.
A main requirement of this architecture is continuous charging power capacity limitation at the charging station during a charging process. The Open Charge Point Protocol (OCPP) in version 2.0 supports communication between SC and CS using smart charging profiles (Open Charge Alliance 2017). This profiles can set constraints to the maximum amount of power that is delivered during the charging transaction and enable dynamic charging profiles for smart charging purposes. Hence, charging stations are able to react on specific behaviors directly without further control signals. Furthermore, the concept of using such charging profiles is seen as a promising direction for better power planning of charging processes in the future since these profiles are generated based on the power constraints of both the vehicle and the grid.
Algorithms
In this section, we describe the main components of the proposed architecture in detail. The logic of each component is introduced, including the used algorithms.
PQindicator
The goal of the PQIndicator is to estimate the grid status based on different KPI values at different MPs in the grid. Furthermore, the output of the PQIndicator is used by the Smart Charger in order to decide about suitable actions of the charging station based on realtime measured KPI values. Since most low voltage networks are built as 3phase systems, the PQIndicator estimates the status of each phase individually. In this regard, the grid status estimation process is equal for each phase and the output is a normalized value within the range [−1,1] called PQIndic.
where \(v_{j,k} \in \mathbb {R}\) is the value of KPI class K_{k} at MP P_{j}. In case the input data does not include a KPI value at a MP, the value is set to v_{j,k}=⊥.
In the remainder of this paper we do not distinguish between composed KPI classes \(\overline {K_{k}}\) (2) or \(\widehat {K_{j}}\) (3) and the original KPI classes K_{k}, but always refer to them by K_{k}.

Green: The situation of the grid is stable. Increasing or decreasing the CSs’ capacity is possible, but not required. The charging station determines the best reaction. The value of the PQIndic is within the range of (YG,GY), where YG=−0.3 and GY=0.3.

Yellow: A slight change of the charging power is required, since the power quality of the grid is not optimal but still above or below a certain critical threshold. This change can be in form of proactive increase or decrease of the charging power. In the yellow state, the grid has high priority and requirements of the charging process can be taken into account only to a certain degree. The PQIndic value is within (RY,YG]∪[GY,YR), where RY=−0.7, YG=−0.3, GY=0.3 and YR=0.7.

Red: The situation of the grid is critical and the load at the charging station must be reduced or increased according to a certain factor. The red color is defined for a PQIndic value within [ER,RY]∪[YR,RE], where ER=−1, RY=−0.7, YR=0.7 and RE=1.

Overloading an element of the grid
Power distribution equipment, such as transformers or cables, have an upper thermal limit which should only be exceeded for a short time. The specific thresholds can vary with the equipment type. An example that depends on the maximum allowed apparent capacity S_{max} is given in Table 1. In case of a transformer the values are chosen in a way, such that the transformer is operated below its maximum apparent power and optimally with highest efficiency.Table 1The thresholds of KPI class from criteria A_{1}
E R _{ k}
R Y _{ k}
Y G _{ k}
G Y _{ k}
Y R _{ k}
R E _{ k}
S _{ max}
y·S_{max}
z·S_{max}
w·S_{max}
x·S_{max}
0

Voltage level
Different load and generation scenarios can cause the voltage level to increase or decrease in certain areas of the low voltage distribution system. According to EN 50160 standard, this variation should be in the boundaries of ± 10% of the nominal voltage U_{N} during 95% of the week measured by 10min mean RMS voltage values. As we are estimating the grid phasebyphase, the voltage level is measured between the phase and the neutral (conductor). Generally defined thresholds for the voltage KPI class are shown in Table 2.Table 2Thresholds of the voltage KPI class from criteria A_{2}
E R _{ k}
R Y _{ k}
Y G _{ k}
G Y _{ k}
Y R _{ k}
R E _{ k}
0.9 U_{N}
0.95 U_{N}
0.99 U_{N}
1.01 U_{N}
1.05 U_{N}
1.1 U_{N}
In the second level, the PQIndicator will use Algorithm 1 to indicate the state of the grid regarding A_{2}. In case \(PQ  Indic_{A_{2}}\) is colored yellow or red, this value is directly returned. Otherwise, the output is calculated by the third level taking the situation at the critical point into account. Hence, if the PQIndic at the critical point is colored yellow or red, \(PQ  Indic_{A_{2}}^{Critical}\) is used as output of the PQIndicator, otherwise, the PQIndic at the charging station \(\left (PQ  Indic_{A_{2}}^{CS}\right)\) defines the return value.
Smart Charger
According to OCPP 2.0 (Open Charge Alliance 2017), charging stations can handle different types of the charging profiles. The ChargingStationMaxProfile is usually based on a 24 h forecast from the DSO, TxProfiles are used for single charging transactions and ChargingStationExternalConstraints profile represent limits set from external systems. This different profiles are stacked and used by their prioritized stack level. The Composite Schedule combines the different profile types by calculating the minimum in each time interval. During the charging process, our smart charging algorithm controls the charging process using external profiles at a high stack level.
where PQ_{i} is the ’PQIndic’ value of phase i. We intend to use a conservative aggregation when phases are in different colored states since phase balancing is performed by the CS. In contrast, aggressive aggregation is used when the grid is asking for load increase on one or more phases while other states are at least green. In that case a charging increase is allowed on all phases, hence we choose the maximum in order to mitigate the biggest problem first. In the case of green states on all phases, the average perfectly reflects the situation.
Finite state machine

Operational states: low red, low yellow, green, high yellow, and highred state representing the different ’PQIndic’ color ranges.

Standby state: The gray state models the charging state after the desired State of Charge (SoC) is reached.

End state (blue state with horizontal lines): With maximum SoC or unplugged EV it is not longer possible to control the charging operation.
The transitions in the FSM are labeled by two parts: Event and Guard. In the proposed FSM we have three kinds of events that can trigger the state transition: Input of a new PQIndic value, unplugging of the vehicle and changing of the SoC of the battery. Each state transition can have a prerequisite, which is modeled by the logical condition of a guard. If the condition of a guard does not match, the FSM remains in the last state. Finally, each state transition can have an action which specifies the output of the Smart Charger. In our case, the action defines the new charging power of the ongoing charging process.
The lowred state is considered as the start state, since the charging operation will start slowly and it adapts a conservative approach concerning the grid stability. The transition to the end state occurs whenever the driver unplugs the vehicle or the battery is fully charged (equals SoC = 100). If the desired SoC_{end} (defined by the end user) is reached, the FSM transits to the charging standby state (gray state). Within this state the Smart Charger can react on critical grid situations using the still plugged EV, only in case the ’PQIndic’ value is positive redcolored, hence requires increase of charging power capacity.
Transitions and actions
In this section, the used capacity of an active charging process at charging station C_{i} at time t is denoted as \(\mathcal {U}_{i}(t)\), the maximum capacity is written as \(\mathcal {M}_{i}\) and the users’ charging profile is denoted as \(\mathcal {C}_{i}(t)\). In all cases, the action of transitions, which reveals the new charging power, should not be bigger than \(\mathcal {M}_{i}\). A safety upper margin is defined by μ in order to stay aligned with the users charging profile regarding the battery state of health and the charging duration. This safety margin is used as a buffer to compensate grid problems that may lead to a short reduction of charging power. Other way round, the minimum charging power needs to be set to a value \(\mathcal {C}_{min}\) higher than zero to avoid disconnection of the vehicle.

∗→ lowred state
If the new PQIndic value is colored lowred, the Smart Charger needs to reduce the charging power. Since this state is considered to be highly critical for the grid, the resulting action is defined in a polynomial way. We calculate δ as the percentage of change in the currently used power.$$ \begin{aligned} \delta &= (\mathrm{PQ\text{}Indic} + 1)^{\alpha}\\ \mathcal{U}_{i}(t+1) &= \max\big(\delta \cdot \mathcal{U}_{i}(t), \mathcal{C}_{min}\big) \end{aligned} $$(6)The parameter α in Eq. 6 needs to be greater than 1 in order to match a polynomial decrease. Therefore, δ∈[0,0.3) because PQIndic∈[−1,−0.7]. As a result, in any case the decrease of the charging power is greater than 70% of the currently used charging power.
The parameter α can be defined depending on the source of the transition or by comparing with the last PQIndic value.

∗→ (low and high)yellow states
Within the boundaries of the yellowcolored PQIndic, the gird is not stable, but it is not highly critical like in the red states. Hence, the transitions to this state can consider the users’ charging profile. The change in the charging power capacity is calculated by a linear function, which depends on the PQIndic value and can be parameterized by the source state of the transition.$$ \begin{aligned} \delta_{1} &= 1 + (\beta_{1} \cdot (\mathrm{PQ\text{}Indic}+ YG + 0.1))\\ \delta_{2} &= 1 + (\beta_{2} \cdot (\mathrm{PQ\text{}Indic}+ GY  0.1)) \end{aligned} $$(7)The parameter β_{1} and β_{2} in Eq. 7 are taken from \(\mathbb {R}\) and can in example be configured by the source of the transition. However, in any case the new charging power is limited by the minimum of \(\mathcal {C}_{min}\) and the maximum of \(\mathcal {M}_{i}\).$$ \mathcal{U}_{i}(t+1) = \left\{ \begin{array}{ll} \min\left(\delta_{1} \cdot \mathcal{U}_{i}(t), (1+2\mu) \cdot {\mathcal{C}}_{i}(t+1), \mathcal{M}_{i}\right) & \mathrm{PQ\text{}Indic} \geq GY \\ \max\left(\delta_{2} \cdot \mathcal{U}_{i}(t), \mathcal{C}_{min}\right) & \mathrm{PQ\text{}Indic} \leq YG\\ \end{array}\right. $$ 
∗→ green state
Within the boundaries of the greencolored PQIndic, the gird is stable. In this regard, a linear increase or decrease of the currently used charging power is applied until the charging profile plus the safety margin is reached.$$ \begin{aligned} \lambda &= (\mathrm{PQ\text{}Indic}+ GY + 0.1)/2\\ \delta_{1} &= \lambda \cdot (1+\mu) \cdot {\mathcal{C}}_{i}(t+1)\\ \delta_{2} &= ({\mathcal{C}}_{i}(t)  (1+\mu) \cdot {\mathcal{C}}_{i}(t+1))/2 \end{aligned} $$(8)$$\mathcal{U}_{i}(t+1) = \left\{ \begin{array}{ll} \min(\mathcal{U}_{i}(t)+ \delta_{1}, (1+\mu) \cdot {\mathcal{C}}_{i}(t+1)) & \mathcal{U}_{i}(t) \leq (1 + \mu) \cdot \mathcal{C}_{i}(t+1) \\ \mathcal{U}_{i}(t) \delta_{2} & \mathcal{U}_{i}(t) > (1 + \mu) \cdot \mathcal{C}_{i}(t+1). \\ \end{array}\right. $$ 
∗→ highred state
If the new PQIndic value equals the highred state, the grid is in a highly critical status and the Smart Charger must increase the charging power. Hence, a polynomial function is defined for transitions to this state.$$ \begin{aligned} \delta &= \omega \cdot (\mathrm{PQ\text{}Indic})^{\varepsilon} + 1\\ \mathcal{U}_{i}(t+1)&= \min\left(\delta \cdot \mathcal{U}_{i}(t), \mathcal{M}_{i} \right) \end{aligned} $$(9)The parameters ω and ε in Eq. 10 can be defined depending on the source of the transition or by comparing with the last PQIndic value. The parameter ω must at least be lower than \(\frac {\mathcal {M}_{i}}{\mathcal {U}_{i}(t)}\) and bigger than 1 and ε must be lower than 1 in order to match a polynomial increase. Obviously, δ is bigger than one, because ω is bigger or equal to 1 and PQIndic is a positive value.

∗→ gray state
The gray state represents the standby phase of the charging process. In this phase, the Smart Charger only responses to highly critical grid situations by increasing the charging rate. Otherwise, the charging power is reduced continuously until it reaches \(\mathcal {C}_{min}\) again in a linear way.$$ \begin{aligned} \delta &= \omega \cdot (\mathrm{PQ\text{}Indic})^{\varepsilon} + 1\\ \mathcal{U}_{i}(t+1) &= \left\{ \begin{array}{ll} \min\left(\delta \cdot \mathcal{U}_{i}(t),\mathcal{M}_{i} \right) & \mathrm{PQ\text{}Indic} \geq YR \\ \max(\mu \cdot \mathcal{U}_{i}(t), \mathcal{C}_{min}) & \mathrm{PQ\text{}Indic} < YR\\ \end{array}\right. \end{aligned} $$(10)The parameters ε and ω in Eq. 10 are similarly defined like with transitions to the highred state and μ∈[0,1).
Finally, oscillations between two different states (low and high red) needs to be avoided, because this affects the stability of the Smart Charger. Hence, the parameters α and ε which are used as exponential factors in the aforementioned definitions of both states, need to be different.
Evaluation
The objective of this section is to evaluate the smart charging algorithm and the reaction of EVs on control commands. The evaluation is carried out using pure (co)simulation and PHIL. The smart charging algorithm is evaluated based on the achieved improvement of the power quality in a simulated grid. Furthermore, we compare the charging process of a simulated EV with a real and an emulated EV in order to study the reaction of EVs on the control signals from the algorithm.
Setup
For all the following evaluation scenarios, the cosimulation framework AIT Lablink is used (Faschang et al. 2013). With Lablink it is possible to replace software simulation components by different PHIL equipment in the laboratory, e.g. a hardware charging station with a Type 2 charging connector or either a real or an emulated EV based on a Resistor  Indicator  Capacitor (RLC) load model. The smart charging is carried out on the simulation of a realistic low voltage network, which is located in a small city in Bavaria. This grid connects 22 households, 21 industries, three PV systems and four charging stations to a transformer using 64 cables. The maximum distance to the transformer is given by a cable with the length of 485 meters. As power gird simulation tool we use PowerFactory (DIgSILENT 2018) from DIgSILENT. The simulation is fed with BDEW load profiles for industries, realistic load profiles for households (Tjaden et al. 2015) and real PV generation profiles. The EventEngine is simulated by a locally installed Apache KAFKA cluster with only one node and default configuration.
The hardware charging station communicates the charging signal via a IEC 62196 Type 2 connector to the EV emulator and the real EV. This standard supports a Pulse Width Modulation (PWM) current signal, which indicates the amount of current that can be provided by the charging station. Since this value is valid for all three phase, phasebalancing is not possible with that protocol. Furthermore, only integer current values are allowed, hence the charging power capacity can only be controlled in discrete steps. The real charging station supports 3phase charging from 6 to 32 A, resulting in a maximum charging power of 22 kW.
The thresholds of the overloading KPI class (first part) and the voltage KPI class (second part)
E R _{1}  R Y _{1}  Y G _{1}  G Y _{1}  Y R _{1}  R E _{1} 
400kVA  300kVA  150kVA  0kVA  0kVA  0kVA 
E R _{2}  R Y _{2}  Y G _{2}  G Y _{2}  Y R _{2}  R E _{2} 
220.94V  222.94V  223.94V  237.94V  238.94V  240.94V 
Parameter configuration of the smart charging algorithm
C _{ min}  \(\mathcal {C}_{i}(t)\)  \(\mathcal {M}_{i}\)  α  β _{1}  β _{2}  μ  ε  ω 

1.3 kW  22 kW  22 kW  1.2  1.5  1.2  0.1  0.7  \( \frac {1}{4} \cdot \left (1 +\frac {\mathcal {M}_{i}}{\mathcal {U}_{i}(t)}\right)\) 
Analysis

To which extend can our smart charging algorithm improve the voltage level in the grid using simulation?

Is the result of the Smart Charger also valid using PHIL simulation?
Conclusion and future work
In this work, a smart charging architecture based on realtime data stream for triggering events in the grid is presented. It is intended to be a scalable, decentralized architecture, and it targets the fairness among the different running charging processes as well. The traffic light model is applied to use the introduced grid flexibility to the emobility sector in a convenient way. The Smart Charger shows the ability to drastically increase the quality of power with regard to the voltage level and load at the transformer by controlling the active power used by the charging stations. In the future, other control strategies for voltage control can be investigated, for example power factor correction, stationary batteries and OLTC mechanism of the transformers. However, considering further factors of power quality beyond the voltage such as harmonics and unbalance of load can be seen as a promising direction even so existing hardware does not support this functionalities. Furthermore, the mathematical stability analysis of the proposed architecture is missing and placed on the top of todo list in the future. Finally, the fairness among the charging processes should be investigated more deeply, e.g. optimizing it by a centralized component which has a global view on the grid and instructs the local Smart Charger to react in a way to improve the power quality in the grid in total.
Notes
Declarations
Funding
This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No. 713864 (ELECTRIFIC). In addition, the validation and test of the proposed approach has been performed using the ERIGrid Research Infrastructure and is part of a project that has received funding from the European Union’s Horizon 2020 Research and Innovation Programme under the Grant Agreement No. 654113. The support of the European Research Infrastructure ERIGrid and its partner ’AIT Austria’ is very much appreciated.
Furthermore, Publication costs for this article were sponsored by the Smart Energy Showcases  Digital Agenda for the Energy Transition (SINTEG) programme.
Availability of data and materials
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.
About this supplement
This article has been published as part of Energy Informatics Volume 1 Supplement 1, 2018: Proceedings of the 7th DACH+ Conference on Energy Informatics. The full contents of the supplement are available online at https://energyinformatics.springeropen.com/articles/supplements/volume1supplement1.
Authors’ contributions
AA and DD were responsible for proposing the architecture, carrying out simulations, and reporting on the results. They also wrote the first draft of the paper. FK and HdM provided research direction, supervision, and funding. They also helped to write the final version of the paper. All authors read and approved the final manuscript.
Competing interests
The authors declare that they have no competing interests.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License(http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.
Authors’ Affiliations
References
 Alonso, M, Amaris H, Germain J, Galan J (2014) Optimal charging scheduling of electric vehicles in smart grids by heuristic algorithms. Energies 7(4):2449–2475. https://doi.org/10.3390/en7042449.View ArticleGoogle Scholar
 Álvarez, JN, Knezović K, Marinelli M (2016) Analysis and comparison of voltage dependent charging strategies for singlephase electric vehicles in an unbalanced danish distribution grid In: 2016 51st International Universities Power Engineering Conference (UPEC), 1–6.. IEEE, Coimbra. https://doi.org/10.1109/UPEC.2016.8114062. Accessed on 20 Aug 2018. https://ieeexplore.ieee.org/document/8114062/.
 Ardakanian, O, Rosenberg C, Keshav S (2013) Distributed control of electric vehicle charging In: Proceedings of the Fourth International Conference on Future Energy Systems. eEnergy ’13, 101–112.. ACM, New York. https://doi.org/10.1145/2487166.2487178. Accessed on 20 Aug 2018. https://doi.org/10.1145/2487166.2487178. Accessed on 20 Aug 2018. http://doi.acm.org/10.1145/2487166.2487178.
 BDEW Bundesverband der Energie und Wasserwirtschaft eV (2013) BDEW Roadmap, Realistic Steps for the Implementation of Smart Grids in Germany. Accessed on 2 Jan 2018. https://www.bdew.de/energie/bdewroadmapsmartgrids/.
 BDEW Bundesverband der Energie und Wasserwirtschaft eV (2015) Smart Grid Traffic Light Concept. Accessed on 2 Jan 2018. https://www.bdew.de/media/documents/Stn_20150310_SmartGridsTrafficLightConcept_english.pdf.
 Chung, C, Chynoweth J, Chu C, Gadh R (2014) Masterslave control scheme in electric vehicle smart charging infrastructure. Sci World J 2014. https://www.hindawi.com/journals/tswj/2014/462312/abs/.
 Cortés, A, Martínez S (2016) A hierarchical algorithm for optimal plugin electric vehicle charging with usage constraints. Automatica 68:119–131. https://doi.org/10.1016/j.automatica.2016.01.060. Accessed on 20 Aug 2018.MathSciNetView ArticleGoogle Scholar
 Deilami, S, Masoum AS, Moses PS, Masoum MAS (2011) Realtime coordination of plugin electric vehicle charging in smart grids to minimize power losses and improve voltage profile. IEEE Trans Smart Grid 2(3):456–467. https://doi.org/10.1109/TSG.2011.2159816. Accessed on 20 Aug 2018.View ArticleGoogle Scholar
 Deutsch, T, Kupzog F, Einfalt A, Ghaemi S (2014) Avoiding grid congestions with traffic light approach and the flexibility operator. Challenges Implementing Act Distrib Syst Manag CIRED 1:1–4. Accessed on 20 Aug 2018.Google Scholar
 DIgSILENT (2018) Digital Simulation and Network Calculation. Accessed on 20 Aug 2018. www.digsilent.de/en/powerfactory.html.
 Fan, Z (2012) A distributed demand response algorithm and its application to phev charging in smart grids. IEEE Trans Smart Grid 3(3):1280–1290. https://doi.org/10.1109/TSG.2012.2185075. Accessed on 20 Aug 2018.View ArticleGoogle Scholar
 Faschang, M, Kupzog F, Mosshammer R, Einfalt A (2013) Rapid control prototyping platform for networked smart grid systems In: IECON 2013  39th Annual Conference of the IEEE Industrial Electronics Society, 8172–8176.. IEEE, Vienna. https://doi.org/10.1109/IECON.2013.6700500. Accessed on 20 Aug 2018. https://ieeexplore.ieee.org/document/6700500/.
 Fernandez, RC, Weidlich M, Pietzuch P, Gal A (2014) Scalable stateful stream processing for smart grids In: Proceedings of the 8th ACM International Conference on Distributed EventBased Systems  DEBS ’14, 276–281.. ACM Press, New York. https://doi.org/10.1145/2611286.2611326. Accessed on 20 Aug 2018. http://dl.acm.org/citation.cfm?doid=2611286.2611326.
 Foster, JM, Trevino G, Kuss M, Caramanis MC (2013) Plugin electric vehicle and voltage support for distributed solar: Theory and application. IEEE Syst J 7(4):881–888. https://doi.org/10.1109/JSYST.2012.2223534. Accessed on 20 Aug 2018.View ArticleGoogle Scholar
 Foundation, AS (2018) Apache Kafka: A Distributed Streaming Platform. Accessed on 8 Jan 2018. https://kafka.apache.org/.
 Kong, F, Liu X, Sun Z, Wang Q (2016) Smart Rate Control and Demand Balancing for Electric Vehicle Charging In: 2016 ACM/IEEE 7th International Conference on CyberPhysical Systems (ICCPS), 1–10.. IEEE, Vienna. https://doi.org/10.1109/ICCPS.2016.7479118. Accessed on 20 Aug 2018. http://ieeexplore.ieee.org/document/7479118/.
 Lehfuss, F, Nöhrer M (2017) Evaluation of different control algorithm with lowlevel communication requirements to increase the maximum electric vehicle penetration. CIRED  Open Access Proc J 2017(1):1750–1754. https://doi.org/10.1049/oapcired.2017.0265. Accessed on 20 Aug 2018.View ArticleGoogle Scholar
 Li, R, Wu Q, Oren SS (2014) Distribution locational marginal pricing for optimal electric vehicle charging management. IEEE Trans Power Syst 29(1):203–211. https://doi.org/10.1109/TPWRS.2013.2278952. Accessed on 20 Aug 2018.View ArticleGoogle Scholar
 Lopes, JAP, Soares FJ, Almeida PMR (2009) Identifying management procedures to deal with connection of electric vehicles in the grid In: 2009 IEEE Bucharest PowerTech, 1–8.. IEEE, Bucharest. https://doi.org/10.1109/PTC.2009.5282155. Accessed on 20 Aug 2018. https://ieeexplore.ieee.org/document/5282155/.
 Longo, M, Zaninelli D, Viola F, Romano P, Miceli R, Caruso M, Pellitteri F (2016) Recharge stations: A review In: 2016 Eleventh International Conference on Ecological Vehicles and Renewable Energies (EVER), 1–8.. IEEE, Monte Carlo. https://doi.org/10.1109/EVER.2016.7476390. Accessed on 20 Aug 2018. https://ieeexplore.ieee.org/document/7476390/.
 Martinenas, S, Knezović K, Marinelli M (2017) Management of power quality issues in low voltage networks using electric vehicles: Experimental validation. IEEE Trans Power Deliv 32(2):971–979. https://doi.org/10.1109/TPWRD.2016.2614582. Accessed on 20 Aug 2018.View ArticleGoogle Scholar
 Medved, T, Prislan B, Zupančič J, Gubina A (2016) A traffic light system for enhancing the utilization of demand response in lv distribution networks In: 2016 IEEE PES Innovative Smart Grid Technologies Conference Europe (ISGTEurope), 1–5.. IEEE, Ljubljana. https://doi.org/10.1109/ISGTEurope.2016.7856303. Accessed on 20 Aug 2018. https://ieeexplore.ieee.org/document/7856303/.
 Open Charge Alliance (2017) OCPP 2.0 Specification. https://www.openchargealliance.org/uploads/files/protected/OCPP2.0.zip.
 Pipattanasomporn, M, Kuzlu M, Rahman S (2012) An algorithm for intelligent home energy management and demand response analysis. IEEE Trans Smart Grid 3(4):2166–2173. https://doi.org/10.1109/TSG.2012.2201182. Accessed on 20 Aug 2018.View ArticleGoogle Scholar
 Restrepo, M, Morris J, Kazerani M, Cañizares CA (2018) Modeling and testing of a bidirectional smart charger for distribution system ev integration. IEEE Trans Smart Grid 9(1):152–162. https://doi.org/10.1109/TSG.2016.2547178. Accessed on 20 Aug 2018.View ArticleGoogle Scholar
 Rivera, J, Goebel C, Jacobsen HA (2015) A distributed anytime algorithm for realtime ev charging congestion control In: Proceedings of the 2015 ACM Sixth International Conference on Future Energy Systems. eEnergy ’15, 67–76.. ACM, New York. https://doi.org/10.1145/2768510.2768544. Accessed on 20 Aug 2018. http://doi.acm.org/10.1145/2768510.2768544.
 Sankaran, C (2002) Power Quality. The electric power engineering series. CRC Press, Boca Raton.Google Scholar
 Shyam, R, Ganesh HBB, Kumar SS, Poornachandran P, Soman KP (2015) Apache spark a big data analytics platform for smart grid. Procedia Technol 21:171–178. https://doi.org/10.1016/j.protcy.2015.10.085. SMART GRID TECHNOLOGIES.View ArticleGoogle Scholar
 Simonov, M (2013) Eventdriven communication in smart grid. IEEE Commun Lett 17(6):1061–1064. https://doi.org/10.1109/LCOMM.2013.043013.122798. Accessed on 20 Aug 2018.View ArticleGoogle Scholar
 Standard, EN (1994) 50160: Voltage characteristics of electricity supplied by public distribution systems. Eur Stand CLC, BTTF 68–6.Google Scholar
 Taft, J, Martini PD (2013) Ultralarge scale control architecture In: 2013 IEEE PES Innovative Smart Grid Technologies Conference (ISGT), 1–6.. IEEE, Washington, DC. https://doi.org/10.1109/ISGT.2013.6497906. Accessed on 20 Aug 2018. https://doi.org/10.1109/ISGT.2013.6497906. Accessed on 20 Aug 2018. https://ieeexplore.ieee.org/document/6497906/.
 Tanaka, T, Sekiya T, Tanaka H, Hiraki E, Okamoto M (2012) Smart charger for electric vehicles with power quality compensator on singlephase threewire distribution feeders In: 2012 IEEE Energy Conversion Congress and Exposition (ECCE), 3075–3081.. IEEE, Raleigh. https://doi.org/10.1109/ECCE.2012.6342512. Accessed on 20th of Aug. 2018. https://ieeexplore.ieee.org/document/6342512/.
 Tjaden, T, Joseph B, Quaschning V (2015) Repräsentative elektrische Lastprofile für Wohngebäude in Deutschland auf 1sekündiger Datenbasis, 8.. HTW Berlin. https://doi.org/10.13140/RG.2.1.5112.0080.
 Vahedi, H, AlHaddad K (2016) A novel multilevel multioutput bidirectional active buck pfc rectifier. IEEE Trans Ind Electron 63(9):5442–5450. https://doi.org/10.1109/TIE.2016.2555279. Accessed on 20 Aug 2018.View ArticleGoogle Scholar
 Varela, J, Hatziargyriou N, Puglisi LJ, Rossi M, Abart A, Bletterie B (2017) The igreengrid project: Increasing hosting capacity in distribution grids. IEEE Power Energy Mag 15(3):30–40. https://doi.org/10.1109/MPE.2017.2662338. Accessed on 20 Aug 2018.View ArticleGoogle Scholar
 Yong, JY, Ramachandaramurthy VK, Tan KM, Mithulananthan N (2015) Bidirectional electric vehicle fast charging station with novel reactive power compensation for voltage regulation. Int J Electr Power Energy Syst 64:300–310. https://doi.org/10.1016/j.ijepes.2014.07.025. Accessed on 20 Aug 2018.View ArticleGoogle Scholar
 Zhang, SS, Xu K, Jow TR (2006) Charge and discharge characteristics of a commercial LiCoO2based 18650 Liion battery. J Power Sources 160(2):1403–1409. https://doi.org/10.1016/j.jpowsour.2006.03.037. Accessed on 20 Aug 2018.View ArticleGoogle Scholar
 Zhong, Y, Xia M, Chiang H (2017) Electric vehicle charging station microgrid providing unified power quality conditioner support to local power distribution networks. Int Trans Electr Energy Syst 27(3):2262. https://doi.org/10.1002/etep.2262. Accessed on 20 Aug 2018.View ArticleGoogle Scholar