A Proportional and Weight Based Decentralized Charge Controller of Electric Vehicles for the Improvement of Local Voltage Profile

The market for electric vehicles has been gradually increasing. Despite several advantages, incorporating electric vehicles (EVs) on a large scale into the distribution grid is a challenging task. Feeder overloads, system power losses, and voltage violations could all arise because of the unregulated charging of EVs. A centralized and coordinated control approach can monitor and manage the large number of EVs; however, it requires high bandwidth and a reliable communication channel. Also, proportional based decentralized EV charging control approach automatically reduces charging current to improve the grid voltage profile, downstream EV encounters higher reduction of charging current due to their position in the grid. As a result, this study developed a decentralized modified proportional-based control approach for EV battery charging that can improve the grid voltage under heavy load conditions and ensure fair charging among EVs, regardless of their position in the grid. The battery’s state of charge (SOC), constant current (CC), and constant voltage (CV)-based charging are all considered by the proposed controller. The proposed method was also compared with proportional and voltage-sensitivity-based decentralized EV charging control methods. A SiC-MOSFET switch-based H-bridge converter is utilized to experimentally validate the grid-connected EV system. The proposed system was initially developed in MATLAB Simulink, and a laboratory prototype was constructed to verify the results experimentally.

Nomenclature

AbbreviationExpansion

SOC	State of Charge.
CC	Constant Current.
CV	Constant Voltage.
$I_{EV}$	EV charging current.
${\mathrm {I}}_{\max }$	Maximum charging current at nominal grid voltage.
$\mathrm {k}_{\mathrm {P}}\mathrm {}$	Scaling factor.
$\mathrm {V}_{\mathrm {r}}\mathrm {}$	Nominal grid voltage.
$\mathrm {V}_{\mathrm {k}}(\text {t})$	Measured grid voltage at time t.
$I_{char}$	EV charging current.
$I_{min}$	Minimum charging current.
$\gamma$	Scaling factor.
$\mu _{V}$	Voltage sensitivity due to load change.
$V_{n}$	Voltage at node n.
$V_{th}$	Threshold voltage.
${\Delta \text {V}}_{\max }$	Maximum voltage deviation.
$\mu _{\mathrm {k}}$	Weight.
$\mu _{\max }$	Maximum weight.
b	Scaling factor.
${SOC}_{0}$	Initial state of charge.
AH	Nominal ampere-hour of the battery.
$I_{Battery}$	Battery charging/discharging current.
$V_{g}$	Grid voltage.
$V_{d}$	d components of grid voltage.
$V_{q}$	q components of grid voltage.
$I_{g}$	Grid current.
$i_{d}$	d components of grid current.
$i_{q}$	q components of grid current.
L	Inductance.
R	Resistance.
$\omega _{0}$	Nominal angular grid frequency.
$V_{dC}$	DC link voltage of converter.
Vnom	Nominal battery voltage.
Voc	Maximum charging voltage.

SECTION I. Introduction

Solar and wind energy are abundant, and harvesting technologies are improving day by day. Renewable energy resources (RER) are difficult to match consumer demands due to their high degree of intermittentness. Energy storage systems (ESS) have been identified in this context as a critical technology to address such intermittency and enable efficient RER penetration into the electrical grids [1], [2], [3]. Utility companies are becoming more accepting of renewable energy generation methods as more energy storage systems are being installed alongside them to reduce or even out the fast transients caused by uncontrollable circumstance changes like temperature, wind speed, and solar radiation. A supercapacitor and flywheel energy storage provide a fast-acting power source with a quick dynamic reaction [4], [5]. However, the high cost and significant energy waste of both technologies restricts their practical use. The battery energy storage system (BESS) boasts prolonged charge retention, coupled with high energy density and cycle efficiency. Additionally, there has been a significant decrease in the price of BESS over time. Besides large-scale battery energy storage for grid applications, electric cars (EVs) have drawn a lot of interest lately, and adoption of these vehicles is anticipated to rise in the near future. Due to the uncertainty of renewable energy, it is always a big challenge for the grid operator to balance generation and load demand. Therefore, an EV battery could be utilized for energy storage as enabling technology of decarbonization and help to support grid voltage and frequency. Even though electric vehicles (EVs) have many advantages, improper charging may result in problems. The charging load is significantly increased by the concurrent large-scale integration of EVs into present distribution power systems, which could result in under voltages, higher losses, and larger load peaks [6], [7], [8]. In contrast, the efficient control strategy for the battery energy of EVs could regulate the voltage profile of the renewable energy connected to the distribution grid, provide ancillary services like reactive power and real power support as a group of EVs [9], [10], [11]. Although the negative effects on the large power system, or system-wide, are probably limited to high levels of EV penetration [9], [10], [12], localized effects on the distribution system are anticipated to be even at modest penetration levels, more noteworthy. Still, research demonstrated that with sufficient management, the detrimental effects of EVs can be mitigated, and the extent to which EVs are gaining traction could be raised [13], [14], [15]. A number of tactics have been proposed to regulate EV charging to avert detrimental effects on the grid. Centralized control techniques have been the subject of many EV charge management strategies [16], [17]. In centralized scheduling and control, all EV status and owner data, together with other system variables including market prices, system loading, and constraints, are gathered centrally before the EV management algorithm is executed. An aggregator or the system operator can function as the central controller. A schematic of the centralized control approach is shown in

Fig. 1. Although a centralized and coordinated control approach can manage the EV fleet and might lead to the best possible use of system resources, they come with several drawbacks like requirements of communication infrastructure, high communication bandwidth, safety precautions, high computing power, and consumer privacy issues etc [18], [19]. Alternative approaches for charging electric vehicles have concentrated on decentralized control, which employ reduced reference decisions for charging and discharging, which are not generated by the system operator or the aggregator. Rather, the operator produces a signal of some kind to encourage the involved EVs to take a specific action [20]. On the other hand, no personal data is returned to the aggregator or the operator; instead, each EV determines its own charging/discharging rate [21]. Multi-agent control techniques are used in other works [22]. Two or more physical entities or virtual that collaborate and communicate with one another to accomplish certain goals pertaining to their surroundings make up a multi-agent system [23]. In [21], an electric car for home charging and discharging was taken into consideration. This study evaluates how price-based demand response tactics affect changes in smart household load patterns. To precisely account for the price elasticity of demand, a hybrid wavelet transforms (WT)—ANN forecasting technique was taken into consideration. The goal of acquiring the household load statistics is to optimize the scheduling of appliances while taking into account an hourly variable price tariff scheme. Recently, a number of studies [24], [25], [26], [27] have employed game theory to coordinate the charging of electric vehicles (EVs) by establishing the Nash equilibrium, in which no player stands to gain by altering their own strategy. Fuzzy logic controllers (FLC) have been used to demonstrate control of energy flow between EVs and the grid, primarily for voltage correction and load flattening, in [28] and [29]. The suggested method is predicated on the EVs being able to be charged and discharged. Additionally, it assumes that every EV in a certain location will charge at a specific station, which will very certainly act as an aggregator. A communication-free decentralized control strategy is being investigated in [30] and [31] for improving the voltage profile of the renewable energy integrated distribution grid under heavy loading conditions. However, battery state of charge (SOC), overvoltage issues under light loading conditions, and dynamic fair charging have not been investigated. In addition, results were provided based on computer simulation, which requires experimental validation. Furthermore, voltage sensitivity does not significantly change with loading, which in turn makes the proposed algorithm as a proportional controller. Therefore, the expected fairness among the EV owners requires further modification. To manage the charging of EVs in a residential distribution system, a proportional and weight-based decentralized controller is proposed in this work. There is no communication between the EV charger and the central controller. This controller dynamically regulates the charging rate of EVs to improve the voltage stability of the system. Since charging between EVs is dependent on the location in the distribution grid, a local voltage-based weight was added to the controller to anticipate the large charging current reduction due to the location of the strong voltage node. In addition to addressing issues related to overloading and undervoltage in the distribution grid, the charging process also takes into account constant current and constant voltage charging algorithms to extend battery lifespan. The complete system has been modeled and simulated on the Matlab Simulink platform. Then, a hardware setup was built for practical validation. In summary, the following study has been carried out:

• Limitation of voltage sensitivity based decentralized EV charging control approach as presented in [30] has been thoroughly presented both in simulation and experimental.

• Shortage of proportional based decentralized EV charging as presented in [31] has been addressed.

• Modified proportional based decentralized EV charging control has been proposed and thoroughly presented. Local voltage-based weight was added to the controller to anticipate the large charging current reduction due to the location of the EV.

• The controller’s back and forth action of proportional based decentralized EV charging technique has been removed by adding a tolerance in the proposed control algorithm.

• Constant current and constant voltage-based charging algorithms have been incorporated into the final EV charging current decision.

This paper is organized as follows: Section II presents the decentralized EV charging control algorithm. Section III provides the mathematical modeling of charging current control. The simulation and experimental findings are shown in Section IV. The conclusion is finally given in Section V.

FIGURE 1.

Centralized control.

SECTION II. Decentralized EV Charging

The system consists of two sets of EV battery, where one set is placed near the source point and another set is placed far away from the source point. The first point of connection is defined as upstream node and the second point is defined as downstream node as shown in Fig. 2. Fig. 2 is a subset of Fig. 1. It is possible to have both controlled and non-controllable loads in a distribution transformer. Since the voltage profile of the system is directly correlated with its loading levels, it can be improved by regulating the loading level. It is assumed in this article that the EVs are the sole controllable loads. To charge the batteries, the EV charger transforms the grid’s AC electricity into a regulated DC current. As a result, the grid views the EV as a current source. In situations where the bus voltage is close to the minimum threshold voltage (0.95pu), it is not desirable for the EV charger to increase charging. One of the objectives of this work is the local voltage based autonomously control the charging current of battery to minimize low voltage problems in the system. Therefore, it automatically controls the charging current with the system voltage.

FIGURE 2.

EV charging.

A. Proportional Based Decentralized EV Charging

Typical proportional based charge current reduction control strategy is given by the following equation [31]

EV charging currrent,

$$\begin{equation*} {I_{EV}=\mathrm { I}}_{\max }-\mathrm {k}_{\mathrm {p}}\mathrm {}\left ({{ \mathrm {V}_{\mathrm {r}}-\mathrm {V}_{\mathrm {k}}\left ({{ \mathrm {t} }}\right) }}\right) \tag {1}\end{equation*}$$

Here, $I_{max}$ is the maximum charging current at nominal grid voltage, $\mathrm {V}_{\mathrm {r}}$ is the nominal grid voltage (1pu), $\mathrm {V}_{\mathrm {k}}\left ({{ \mathrm {t} }}\right)$ is the measured grid voltage at time t, and $\mathrm {k}_{\mathrm {p}}$ is scaling factor to convert voltage to suitable current. Although this control strategy automatically reduces charging current to improve the grid voltage profile, downstream EV encounters higher reduction of charging current according to (1).

B. Voltage Sensitivity Based Decentralized EV Charging

As observed in the preceding proportional-based EV charging method, electric vehicles (EVs) situated in the downstream grid encounter more significant reductions in charging current, owing to the greater voltage differential compared to EVs in the upstream grid. Therefore, a voltage sensitivity based autonomous EV charging was proposed in [30] to improve fair charging among EVs irrespective of their location in the grid. The aim of the suggested charge controller is to regulate the charging current of an electric vehicle (EV) according to the voltage at the point of connection (POC), aiming to prevent voltage violations within the system. This adjustment can result in decreased power losses along the lines and prevent overloads. Moreover, fairness constitutes a fundamental element of every EV charging strategy. In this context, fairness denotes the equal distribution of the limited system capacity among all electric vehicles (EVs). Essentially, EVs with similar state of charge (SOC) should be charged at nearly identical rates, regardless of where they are charging within the system. Consequently, the proposed charge control method utilizes the interdependent relationship between voltage and voltage sensitivity at the POC. This is grounded in the observation that nodes with lower voltages tend to be more sensitive to fluctuations in load power compared to those with higher voltages. As a result, EVs located at downstream nodes typically experience lower voltages but higher sensitivities compared to those at upstream nodes.

The charging current is given by the following equation [30].

$$\begin{align*} I_{char}=\begin{cases} \displaystyle I_{min}+\gamma \left ({{ e^{-\mu _{V}\left ({{ t }}\right)} }}\right)\left ({{ V_{n}\left ({{ t }}\right)-V_{th} }}\right) & if~V_{n}\left ({{ t }}\right)\ge V_{th} \\ \displaystyle 0 & else \end{cases} \tag {2}\end{align*}$$ Here,

$I_{min}$ is the minimum charging current.

$\gamma$ is the scaling factor.

$\mu _{V}$ is the voltage sensitivity due to load change at node n.

$$\begin{equation*} \mu _{V}\left ({{ t }}\right)= \frac {V_{n}\left ({{ t }}\right)-V_{n}\left ({{ t-1 }}\right)}{P_{n}\left ({{ t }}\right)-P_{n}\left ({{ t-1 }}\right)} \tag {3}\end{equation*}$$ $V_{n}\left ({{ t }}\right)$ is the voltage at node n.

$V_{th}$ is the threshold voltage.

As can be seen from (2) that the charging current of EV is a function of node voltage sensitivity and local voltage. Usually, the sensitivity of downstream node voltage is greater than the upstream node voltage. Therefore, the charging current automatically reduces with the local voltage drops. The node voltage sensitivity is used to track the position of EV, and improved fair charging among EVs irrespective their position in the grid.

C. Proposed Proportional and Weight-Based Decentralized EV Charging

If the system voltage is at minimum threshold voltage, then the charging current is set minimum to reduce the overloading issue. On the other hand, charging current is set maximum if the system voltage is at nominal value. The charging current is set in between maximum and minimum current if the system voltage remains within nominal and minimum threshold voltage. However, this approach creates fairness issues among the EV due to their different location from the strong node voltage. Thus, maintaining equity is a key goal when the EVs are being charged. Each EV should be charged to achieve appropriate bus voltages, but the charging process shouldn’t be designed to make one EV charge more quickly than another based solely on where the EVs are located in the grid. It is inappropriate for EVs connected to the lower voltage downstream load bus to have much slower regulated charging rates than EVs connected to the higher voltage upstream load bus. Therefore, the proposed system improves the shortcoming of proportional-based EV charging method by incorporating an additional dynamic weight against significant charging current reduction of EVs in the downstream grid. It is given by the following equation.

Proportional and weight-based EV charging current,

$$\begin{equation*} {I_{EV}=\mathrm { I}}_{\max }-\mathrm {k}_{\mathrm {p}}\mathrm {min (}{\mathrm {\Delta V}}_{\max },\left ({{ \mathrm {V}_{\mathrm {r}}-\mathrm {V}_{\mathrm {k}}\left ({{ \mathrm {t} }}\right) }}\right)+\mathrm {\mu }_{\mathrm {k}}\mathrm {(t) } \tag {4}\end{equation*}$$ Weight, $$\begin{equation*} \mu _{\mathrm {k}}\left ({{ \mathrm {t} }}\right)=\mathrm { min (}\mathrm {\mu }_{\max }\mathrm {,b}\left ({{ \mathrm {V}_{\mathrm {r}}-\mathrm {V}_{\mathrm {k}}\left ({{ \mathrm {t} }}\right) }}\right) \tag {5}\end{equation*}$$ Here, $\mathrm {k}_{\mathrm {p}}$ and b are constant or scaling factor to convert voltage to suitable current, ${\mathrm {\Delta V}}_{\max }$ is maximum voltage deviation from the nominal grid voltage, $\mathrm {\mu }_{\max }$ maximum weight that can be added against the proportional reduction. However, this weight is also a function of voltage difference that means higher voltage difference yields higher weight whereas lower voltage difference yields lower weight. Two scaling factors and discrete limits provide more flexibility of controlling charging current to improve local voltage profile as well as the improvement of fair charging among EVs irrespective of their position in the grids. Besides grid voltage profile improvement, and fairness among EVs, the proposed control strategy also ensures efficient charging mechanism for the EV battery. Usually, it is preferable to charge the battery using a combination of constant current (CC) and constant voltage (CV) method for the longevity of battery life. Battery voltage is maintained at the battery’s max terminal voltage when the battery is charged with a fixed current. The capacity (C) of the battery determines the constant current needed for charging. This charging mode typically charges the batteries at a pace of 0.5C to 0.8C. This state allows the battery to charge very quickly. Fast charging state is another name for this state for this reason. Once the battery is eighty percent charged, constant voltage (CV) mode is used to charge the remaining twenty percent. The charging current will begin lowering and the CC status is to be turned off once the battery is 80% charged. In order for the battery to be fully charged, the constant voltage needs to match its maximum voltage. The CC and CV methods are shown in Fig. 3. For the successful realization of CC and CV method, the information of state of charge (SOC) of battery is needed. The following relation defines the state of charge of the battery. $$\begin{equation*} SOC={SOC}_{0}-\frac {1}{3600AH}\int _{0}^{t} {I_{Battery}dt} \tag {6}\end{equation*}$$ where, ${SOC}_{0}$ is the initial state of charge, AH is the nominal ampere-hour of the battery, and $I_{Battery}$ is the battery charging/discharging current. The coulomb counting method, also known as the ampere-hour counting and current integration approach, is the method used to calculate the SOC in equation (6). In order to determine SOC values, this method uses battery current readings that have been numerically integrated across the usage period. The remaining capacity is then determined by the coulomb counting method, which just adds up the charge that is moved into and out of the battery. This method’s accuracy mainly depends on a precise measurement of the battery current and an accurate estimation of the starting SOC. The SOC of a battery can be computed by integrating the charging and discharging currents during the operating periods, given a predetermined capacity that may be remembered or first estimated by the operating conditions. By including the open circuit voltage approach, the coulomb counting method’s precision or tuning is greatly enhanced. In this work, initial SOC is calculated based on open circuit voltage measurement, and then SOC is dynamically calculated using coulomb counting method. For mitigating under voltage issue, ensuring efficient charging mechanism for the EV battery and providing fairness among the EVs irrespective of their location in the distribution grid, the proposed proportional and weight-based charging current control algorithm is shown in Fig. 4. It measures two voltages at interval $\Delta t$ , which is used to calculate the minimum tolerance $\Delta V_{th}$ for updating the charging current. This tolerance is needed to prevent the back-and-forth condition of charging current. If the charging current is increased, then the grid voltage may decrease, which further decreases the charging current and increases the grid voltage. This back-and-forth condition is prevented by incorporating voltage change tolerance. That means, charging current is not updated until a certain amount of voltage change ($\Delta V_{th}$ ) is observed. The algorithm takes the battery SOC, battery voltage, grid voltage and grid voltage changes to check the next condition. If the SOC is below 80% and battery voltage is also below the maximum battery voltage, then it goes to the next condition.

FIGURE 3.

Constant current and constant voltage charging of battery.

FIGURE 4.

Proposed charging current algorithm.

If the grid voltage is below or equal to the threshold ($V_{th}$ ), then charging current is set to minimum and exit the loop. However, if it is above the threshold voltage then it checks the next condition. Again, charging current is set to maximum if the grid voltage is above or equal to nominal voltage. If the statement is false and grid voltage changes is above the minimum threshold voltage ($\Delta V_{th}$ ) then it calculates the charging current using (4) and (5) and exit the loop. However, the charging mechanism switches to constant voltage mode if the SOC is above 80% and battery voltage is below maximum charging voltage. If the battery is fully charged, then it stops charging and exits the loop.

SECTION III. EV Charging Current Control

Fig. 5 shows the two EV battery sets connected through the full bridge AC/DC converter separated by the feeder impedance. Grid voltage, grid current, battery voltage and battery current are fed to controller, which generates gate pulses to control the required charging current. The control for single converter is explained in the following paragraph. However, the control for the second converter is the same as the first converter. The control is developed in dq frame instead of stationary frame. Therefore, the stationary signals of the grid voltage and grid current are converted into dq signal with the grid angle, and all control actions are performed in dq frame. At steady-state, the dynamics of current controller of the converter is described by the following equation [32].

$$\begin{align*} L\frac {di_{d}}{dt}+Ri_{d}& =L\omega _{0}i_{q}+V_{dC}-V_{d} \tag {7}\\ L\frac {di_{q}}{dt}+Ri_{q}& =-L\omega _{0}i_{d}+V_{qC}-V_{q} \tag {8}\end{align*}$$ The dq components of grid voltage ($V_{g}$ ) are denoted $V_{d}$ and $V_{q}$ , and grid current ($I_{g}$ ) in Fig 6(a) are denoted $i_{d}$ and $i_{q}$ . The inductance and resistance of the reactor are L and R. The dq components of the converter output are $V_{dC}$ and $V_{qC}$ . The grid current control is presented in Fig. 6(c). A 90-degree phase-shifted signal ($\beta$ - components) is obtained by utilizing quadrature transport delay block (T/4) for single-phase signal as opposed to a three-phase grid. The two components in the $\alpha \beta$ frame are finally transformed into dq rotating reference frame via the park transform. The PLL-derived grid angle ($\theta$ ) serves as the rotating frame’s reference angle. The value of the current used to charge and discharge batteries is controlled by the reference real current (i$_{\mathrm {d}}$ ). Grid to battery power flow is indicated by a positive sign, while battery to grid power flow is shown by a negative sign. Fig. 6(b) shows the reference current generation, which has been discussed in section II. The reference charging current is either generated directly from the CC mode or after the voltage control loop from the CV mode. The error from the reference current is processed through the PI controller, which combined with feedforward and decoupling term to produce $V_{d} ~and ~V_{q}$ . Finally, it converts into modulating signal, which further generates gate pulses for the full bridge converter. Fig. 6 (c) presents current control in dq frame.

FIGURE 5.

The block diagram of the complete system.

FIGURE 6.

Grid connected EV battery (a) single line diagram (b) reference charging current of battery (c) charging current control in dq frame.

SECTION IV. Results and Discussion

A network with two sets of EV batteries similar to Fig. 2 was developed in Matlab Simulink for the decentralized charging of EV batteries. Both EV batteries are connected to the AC grid through a single-stage AC/DC converter. Matlab Simulink results have been practically verified through the laboratory prototyping system. Table 4 and Table 5 include the details needed for the system. An experimental setup, shown in Fig. 7, has been created for the purpose of validating the proposed method. A DC link of 160V is created by connecting three lithium-ion battery packs in series. A variable transformer connected with utility grids simulates the AC grid. An AC capacitor is used to filter the high-frequency harmonics from the AC grid connection points. The dSPACE Microlab-Box platform is used to execute the control algorithm. LEM sensors are used to get the necessary voltage and current signals. Next, a single-phase converter based on SiC MOSFET from Taraz Technologies is driven by the PWM output. The dSPACE control desk is used to record and display real-time currents and voltage. The inductor and resitor are used to create the feeder. Similarly, another three sets of batteries were connected in series to form the DC link for the converter. The converter is controlled using the same dSPACE Microlab box. Before testing the decentralized charging controller for the two battery sets, the performance of the current controller is provided to make sure that the reference current is perfectly tracked. It is also needed to maintain unity power factor operation during the charging process, as the battery receives real power. The real and reactive current controller performance of both converters for battery charging has been investigated. The d-axis current or real current command is set to 5A for EV1 and 4A for EV2, and then eventually dropped to 0A for the grid-to-battery power flow operation. Fig. 8 illustrates the performance of the current controller, demonstrating a slight overshoot of the actual current compared to the reference current with zero steady state error. Usually, the battery receives real power, which means the converter needs to transfer current from the AC grid at unity power factor. Therefore, the reactive reference current is set to 0A for both EV converters. Fig. 9 shows that the q-axis current, or actual reactive current, tracks its reference reactive current command. The grid voltage and current waveforms for two batteries shown in Fig. 10 and Fig. 11 demonstrate that the battery is being charged at unity power factor. The experimental performances, as shown in Fig. 12 and Fig. 13, also verify the simulated performances. In the experiments, the voltage wave is scaled down to fit with the current wave. A zero-phase shift between the current and voltage wave indicates unity power factor operation during battery charging. Both current controllers for the two EV sets show almost similar performance.

TABLE 1 Performance Summary of Voltage Sensitivity Based EV Charging

TABLE 2 Performance Summary of Proposed Technique Based EV Charging

TABLE 3 Performance Comparison of Three Techniques

TABLE 4 Level 1 Residential EV Battery and Converter Data for Matlab Simulation

TABLE 5 LiFePO4 Battery Parameters

FIGURE 7.

Experimental set-up.

FIGURE 8.

d-axis current tracking of EV₁ and EV₂.

FIGURE 9.

q-axis current tracking of EV₁ and EV₂.

FIGURE 10.

Grid current and voltage during charging operation of EV battery set 1.

FIGURE 11.

Grid current and voltage during charging operation of EV battery set 2.

FIGURE 12.

Measured dq current when the battery units were being charged (Experimental).

FIGURE 13.

Measured current and voltage waveforms when the battery units were being charged (Experimental).

A. Voltage Sensitivity Based Decentralized EV Charging [30]

This section has thoroughly presented the limitation of voltage sensitivity based decentralized EV charging control approach as reported in [30] both in simulation and experimental. As discussed in Section II-B, the decision on EV charging current depends on node voltage sensitivity. Matlab simulations and experiments measure the node voltage sensitivity, as illustrated in Fig. 14 and Fig. 15. A load change is applied to measure the voltage change, and finally, sensitivity is calculated. In Fig. 14, load is changed at 0.2s, 0.6s, and 0.9s at the upstream node, whereas load is changed at 1.2s, 1.55s, and 1.85s at the downstream node. The voltage changes are measured corresponding to the load change, which yields node voltage sensitivity. Similar measurements are carried out experimentally to determine the node voltage sensitivity, as depicted in Fig. 15. The node voltage sensitivity in both cases is not greatly affected by changes in load, but it is strongly affected by the node position in the grid. While EV is working in constant current mode without a voltage-sensitivity-based current controller, the load of the upstream node is changed at 4.5 s, as can be seen in Fig. 16. Hence, the voltage of the upstream node drops from 0.975 pu to 0.945 pu. In a similar case, a voltage-sensitivity-based current controller automatically lowers the charging current from 5A to 4A, limiting the node voltage decrease to 0.95 pu, as illustrated in Fig. 17. Similar to this, as shown in Figs. 18 and. 19, the EV charge controller downstream limited the node voltage reduction to 0.92 pu, which decreased below 0.9 pu if the controller was not enabled. Although EVs on both nodes changed the charging current to improve the local voltage profile, the controller failed to ensure fair charging among EVs, which was the main claim of the work [30]. Both EVs are supposed to charge at the same reduced rate automatically. The charging current of upstream and downstream EV are found 4A and 3A respectively, which clearly indicates that the downstream EV has disadvantages due to its position far away from the generating source.

FIGURE 14.

Voltage sensitivity measurement (Matlab Simulation).

FIGURE 15.

Voltage sensitivity measurement (Experimental).

FIGURE 16.

Upstream grid voltage, battery, and load current without EV current controller (Experimental).

FIGURE 17.

Upstream grid voltage, and battery current with EV current controller (Experimental).

FIGURE 18.

Downstream grid voltage, battery, and load current without EV current controller (Experimental).

FIGURE 19.

Downstream grid voltage, and battery current with EV current controller (Experimental).

Table 1 summarizes the performance of voltage sensitivity-based EV charging.

After revisiting (2), further study has been done by changing the controller parameters to make sure that all EVs are treated fairly, regardless of where they are in the grid. It is expected that during significant loading conditions, all EVs with similar states of charge would charge at the same rate. The only tuning parameters of (2) are $I_{min}$ and $\gamma$ , which have been changed to different values to improve fairness among the EVs, irrespective of their position in the grid. However, Fig. 20, Fig. 21, and Fig. 22 show no improvement in terms of fair charging of EVs. As illustrated in these figures, the loads in upstream and downstream nodes have been changed from minimum (1.2kW) to maximum (3.5kW) at 0.5s, 1.5s, 2.25s, and 3s to observe the EV charging current controller performance in validating the fair charging. In Fig. 20, $\gamma$ is set to 16000 and $I_{min}$ is set to 4A, whereas in Fig, 21, $\gamma$ is set to 32000 and $I_{min}$ is set to 2A. Neither case resulted in a similar reduction in the EV charging current. Additionally, the charging current shows the same results as without a controller when the $\gamma$ is set to 32000 and $I_{min} =4$ A, as illustrated in Fig. 22. Hence, fair charging cannot be achieved by considering the absence of other tuning parameters because node voltage sensitivity, $\mu _{V}$ functions as a constant for a particular node regardless of loading conditions, as illustrated in Fig. 14, and Fig. 15. Even though the controller was able to improve the local voltage, it was unable to guarantee that the EVs at various nodes would receive a similar charging current with the same state of charge.

FIGURE 20.

Grid voltage and charging current for $\mathbf {I}_{\mathbf {min}} =4$ A, and $\mathrm {\boldsymbol {\gamma }} =16000$ .

FIGURE 21.

Grid voltage and charging current for $\mathbf {I}_{\mathbf {min}} =2$ A, and $\mathrm {\boldsymbol {\gamma }} =32000$ .

FIGURE 22.

Grid voltage and charging current for $\mathbf {I}_{\mathbf {min}} =4$ A, and $\mathrm {\boldsymbol {\gamma }} =32000$ .

B. Proportional and Weight Based Decentralized Charging

The proposed proportional and weight-based charging control is compared with the proportional based charging current control. The upstream and downstream nodes are loaded with EV batteries and resistive loads. The SOC for both EV batteries is recorded below 80%. The upstream and downstream node voltages were found 0.975 pu and 0.957 pu, respectively, as shown in Fig. 23 and Fig. 24. The performance of two types of control techniques for the control of EV battery charging current has been tested for this condition. In a proportional-based charging current control strategy, the EV charging current is reduced proportionally with the decrease in node voltage from the nominal voltage (1 pu). In Fig. 23 and Fig. 24, the charging current for the EV connected downstream was reduced from 5A to 2.2A, whereas the charging current for the EV connected upstream was reduced only from 5A to 3.8A. In the proportional and weight-based charging current control strategy, the EV charging current is reduced proportionally with the decrease in node voltage from the nominal voltage (1 pu). However, additional weight is added so that the charging current of the EV is not reduced by the same amount as the proportional-based charging current control strategy. In Fig. 23 and Fig. 24, the charging current for the EV connected upstream was reduced from 5A to 3.85A, whereas the charging current for the EV connected downstream was reduced from 5A to 3A. The new set point of charging current is greater than the previous proportional-based charging current control strategy. Hence, it has minimized the negative impact for the EV connected at the downstream node. Both strategies improved the voltage profile. However, comparatively less current reduction from the proposed control strategy made almost similar improvement in the grid voltage profile. Fig. 25 shows the performance of charging the current controller without voltage change tolerance. The current changes abruptly when the controller is activated with a 40% overshoot, which changes the voltage with the overshoot. However, the proposed controller with tolerance produced a small overshoot, as shown in Fig. 23 and Fig. 24. Besides, the location of the generating source has changed to test the controller’s robustness. The result in Fig. 26 is obtained after changing the generating source from the upstream node to the downstream node. Now the downstream node is a strong node, and the upstream node is a weak node. The charging current of EVs at the downstream grid is smaller than that at the upstream grid, which indicates the robustness of the charging current control strategy irrespective of the EV location.

FIGURE 23.

Autonomous charging control of EV battery at Upstream node.

FIGURE 24.

Autonomous charging control of EV battery at Downstream node.

FIGURE 25.

Autonomous charging control of EV battery without tolerance.

FIGURE 26.

Autonomous charging control of EV battery when the generating location is changed.

To experimentally validate the simulated system depicted in Figures 23 to 24, two electric vehicle (EV) batteries were connected to the upstream and downstream nodes respectively, with each EV being charged at a constant current of 5A. The experimental platform as depicted in Fig. 7 was utilized for real time validation. In the proportional based charging current control strategy, the charging current for the EV connected at downstream was reduced from 5A to 2.8A, whereas the charging current for the EV connected at upstream was reduced only from 5A to 4.3A, as illustrated in Fig. 27 and Fig. 28. In the proportional and weight-based charging current control strategy, the EV charging current is reduced proportionally with the decrease of node voltage from the nominal voltage (1pu). However, additional weight is added so that the charging current of EV is not reduced the same amount as proportional based charging current control strategy. From Fig. 27 and Fig. 28, the charging current for the EV connected upstream was reduced from 5A to 4.5A. whereas the charging current for the EV connected at downstream was reduced from 5A to 3.5A in the proposed charging current control strategy. The new set point of charging current is greater than the previous proportional based charging current control strategy. Hence, it has minimized the negative impact for the EV connected at downstream node. Both strategies improved the voltage profile. However, comparatively less current reduction from the proposed control strategy made almost similar improvement in the grid voltage profile. As can be seen from Fig. 23 to Fig. 28, both simulated platform and experimental platform have provided almost identical results. In summary, the following investigations have been carried out, and concluded in this section.

• A comparative performance with proportional based decentralized EV charging has been presented.

• The proposed system dynamically minimizes the large current reduction of downstream EVs, as compared to voltage sensitivity based decentralized EV charging control.

• The proposed system was found more straightforward, and easier to implement compared to voltage sensitivity based decentralized EV charging control.

• No back-and-forth action of the proposed controller was found due to the addition of a tolerance band in the proposed control algorithm.

Table 2 summarizes the performance of proportional and weight based decentralized EV Charging.

FIGURE 27.

Autonomous charging control of EV battery at Upstream node (Experimental).

FIGURE 28.

Autonomous charging control of EV battery at Downstream node (Experimental).

Based on the analysis and results of three techniques for EV charging control, proportional-based decentralized EV charging, and the proposed approach are found simple and easier to implement. On the other hand, voltage-sensitivity-based EV charging necessitates phase shift information to determine the actual power flow and requires a complex calculation to determine the node’s voltage sensitivity. Furthermore, proportional-based decentralized EV charging control improved the local voltage profile more than the other techniques, resulting in the largest reduction of EV charging current. This technique, however, has disadvantages for the EV owner positioned at the downstream node. Due to the counterweight-based charging current reference, the proposed technique provides higher benefits to the EVs positioned on the downstream grid. Voltage-sensitivity-based EV charging also provides benefits to the EVs positioned at the downstream grid, but voltage sensitivity becomes almost independent of load change and acts as a constant scaling term instead of a variable term. The proportional and voltage-sensitivity-based charging approaches did not consider CC-CV charging, unlike the proposed approach. Table 3 summarizes the findings of three techniques.

SECTION V. Conclusion and Further Research

This work developed a decentralized local charging control approach for the EV battery by taking into consideration the local voltage, CC-CV charging, state of charge (SOC) of the battery, and fairness among the EVs. Firstly, the EV controller was tested on a MATLAB Simulink platform, and then a small-scale laboratory setup was developed for real-time hardware validation. Both platforms provided almost identical performances. By reducing charging current, the suggested method has improved the local voltage profile under overloaded conditions. However, substantial reduction in charging current for downstream node EVs was prevented by the counterweight-based charging current control approach. The location of source was changed to test the controller’s robustness. The proposed system was also compared with the proportional-based charging current reduction approach and found improved fairness among the EVs.

Furthermore, the constraints of a decentralized EV charging control approach based on voltage sensitivity have been comprehensively outlined, with evaluations conducted via simulation and experimentation, revealing its lower performance in comparison to the suggested EV charging method. The proposed system and hardware platform can be extended to utilize the bidirectional energy flow of energy storage for supporting voltage and frequency. In addition, renewable energy can be added to the system to study its intermittent nature and performance improvement with EV based energy storage. Although the proposed system improves the local voltage profile, optimization and AI could be utilized to find out the optimized static or dynamic value of weight to ensure fair charging among the EVs, irrespective of location in the grid. All EVs should contribute to the improvement of the grid voltage profile during heavy loading conditions without sacrificing significant charging current reduction due to their position in the downstream grid compared to their position in the upstream grid.

Appendix

See Tables 4 and 5.