2.1. System Presentation
Isolated bidirectional DC–DC converters [
9,
10,
11] represent a crucial component for connecting storage devices between a DC bus and a battery with high-power uses. By discovering many topologies for the isolated bidirectional DC–DC converters, it was clear that each topology has the owner characteristics and the related limitations. Despite the simple design and implementation, the main drawback of the DAB topology is the limited voltage operation range [
12]. Therefore, resonant topologies, such as the SRC and LLC topologies, are proposed to provide an improved soft-switching range.
The SRC topology [
14] can improve the DAB operating zone, but it still suffers from soft-switching property loss at high input voltage and light load. With the CLLC converter [
28], switches have the capability to operate with a wide ZVS zone in the two operating modes for a wide range of voltage gain. However, it has additional components, i.e., higher cost and sizing as well as more complex design and control. Thus, the trade-off between the control, design and efficiency requirements and the cost should be taken into account in order to define the converter topology. From here, DC–DC LLC resonant converters are frequently chosen [
29,
30,
31,
32] for uses that require a high power density, such as the EV charger.
The use of DC–DC LLC converters is common in numerous industrial applications [
33]. The DC–DC LLC resonant converter is a popular isolated DC–DC topology because of its appealing advantages: Zero-Voltage Switching (ZVS) for power MOSFETs and Zero-Current Switching (ZCS) for diodes in a wide operating zone; simple circuit setup with low component count; and high power density with high switching frequency [
34].
In this paper, the full-bridge DC–DC LLC resonant converter [
16,
35] and its control are investigated to suit the application features of the bidirectional EV charger, which include a wide range of battery voltages and high power densities [
15,
24].
The configuration of the full-bridge DC–DC LLC resonant converter is shown in
Figure 2.
Figure 2 includes the DC bus capacitor, switching network, resonant circuit and battery model. Switches
–
with anti-parallel diodes and snubber capacitors make up the switching network. A resonant inductor
, a magnetizing inductor
and a resonant capacitor
make up the resonant tank. The high-frequency transformer with turn ratios of
n ensures the galvanic isolation. The battery model is represented by the constant DC voltage source. It should be noted that there is an output filter circuit, which is not represented in
Figure 2, before the battery model that aims to reduce the battery current ripple. The voltage
is the DC bus voltage which represents the output voltage of the AC–DC stage of the EV charger, while
is the battery voltage and
P is the converter power.
Pulse Frequency Modulation (PFM) is the most widely used modulation technique for DC–DC LLC resonant converters. To ensure ZVS conditions, the switching frequency feasible zone is established between the minimum and highest authorized values. The EV charger’s software and hardware implementation face a cost minimization challenge as a result of this frequency feasibility restriction. However, a wide operating switching frequency range is necessary to satisfy the system voltage gain requirement when the PFM strategy is used for wide input/output voltage and power ranges in the onboard battery charger in V2X mode. Due to the loss in the soft-switching operation brought on by the broad switching frequency range [
32,
36], the conversion efficiency and control performance are poor.
DC–DC LLC converters with the PFM strategy are not preferred in wide voltage range applications due to the following: large switching frequency variations are necessary that reduce the performance of the converter’s Electro-Magnetic Interference (EMI); furthermore, in the case of low power loads, the PFM strategy causes degradation of efficiency and low control performance due to a wide switching frequency demand. As a result, DC–DC LLC converters that operate at a fixed switching frequency are favorable in applications involving a large voltage gain range and can effectively address the aforementioned problems. According to the existing literature, there are two categories of control techniques for DC–DC LLC converters that use fixed switching frequency: Phase-Shift Modulation (PSM) and Pulse Width Modulation (PWM). These two strategies have been investigated toward improving the control performance within a wide range of the battery voltage and high power density, as well as avoid the PFM strategy limitations (efficiency degradation at low power loads, switching frequency saturation, etc.).
LLC resonant converters contain the resonant circuit that consists of a series resonant inductor, capacitor and a parallel magnetic inductor. On the other hand, the LLC resonant converter is a nonlinear system due to the presence of switching frequency harmonics, nonlinear coupling between AC and DC model variables and the LLC resonant circuit. There are two modeling approaches: large signal modeling and small signal modeling. The first harmonic approximation (FHA) [
36,
37] is used to determine the voltage transfer functions. By disregarding the impact of the LLC resonant circuit dynamics, an averaged LLC converter model is demonstrated. Control can be made easy by using small signal modeling [
18,
19] to create an LLC transfer function based on an averaged mode [
38]. It does not, however, take into consideration nonlinear and uncertain effects, such as the DC–DC LLC converter structure, which reduces the operating range of the control law and has an impact on the robustness of the control law in the face of system perturbations.
For the DC–DC LLC converter, models based on the Extending Describing Function (EDF) have been presented [
20,
21,
23]. When an unexpected disturbance arises, these models provide adequate LLC converter dynamic information. With these models, it is possible to create a mathematical model that describes the dynamics of the DC–DC LLC converter variables. The model is still nonlinear and complex, making control law design difficult.
2.2. Modulation System Based Small Signal Modeling
Small signal modeling with the first harmonic approximation (FHA) methodology [
36] is used to examine the dynamics of the DC–DC LLC converter. The FHA is based on the following assumptions:
The input voltage is modeled as an ideal sinusoidal voltage source, in which all higher-order harmonics are ignored and only the fundamental component is reflected;
The capacitor of the output filter, the leakage inductance of the transformer and the effects of MOSFETs are ignored.
2.2.1. PFM
The PFM strategy [
39] consists of varying the switching frequency of MOSFET control signals where the full bridge’s power MOSFETs are regulated in complementary mode at 50% duty. It should be noted that the PFM strategy is one of most frequently adopted modulation strategies for DC–DC resonant converters.
Assuming that the voltage signal can be only represented by the fundamental component in order to simplify the model design, the equivalent circuit model in G2V mode is presented in [
40].
In V2X mode, while the full-bridge diodes on the primary side of the transformer are used for rectification, the power MOSFETs on the secondary side (S5-8) are regulated in complementary at 0.5 duty disregarding the dead time.
The input voltage of the LLC resonant circuit, in
Figure 2, can be expressed in V2X mode, as in (
1) [
24]:
The rectifier full-bridge side is driven by a square output voltage with a fundamental component
expressed in (
2):
Using small signal modeling with FHA, the equivalent model of the DC–DC LLC resonant converter (
Figure 2) can be derived as presented in
Figure 3 [
24].
The equivalent resistor
is defined, based on FHA [
24], in (
3):
Based on the equivalent model of the DC–DC LLC converter shown in
Figure 3, the gain transfer function, based on FHA, can be expressed as in (
4):
where
(
f is the switching frequency).
For the DC–DC LLC converter, the widely adopted modulation strategy is PFM. However, there is low conversion efficiency when the PFM strategy is adopted for wide voltage range application in the reverse operating mode. Therefore, many different modulation strategies such as PWM and PSM strategies are proposed to increase the DC–DC LLC converter feasibility operating with improved efficiency.
2.2.2. PWM
In the Pulse Width Modulation (PWM) approach [
41], a rectangular pulse wave with variable pulse width is employed. The key idea behind the PWM technique is that the duty cycle can change the square wave input voltage of the resonant circuit.
The fundamental idea behind the PWM technique is that the square input voltage of the LLC resonant circuit can be adjusted by modifying the duty cycle while maintaining a constant switching frequency. The gain value is affected by the resonant tank impedance (Equation (
4)), which depends on the switching frequency. The PWM strategy is designed based on FHA. The switching frequency
f is constant and the variable duty cycle
D ensures DC bus voltage control.
PWM offers the benefit of being able to work at switching frequencies that are lower than those of the traditional PFM control method.
The following expression, which relies on
D, describes the fundamental component of the resonant circuit input voltage [
27]:
The following is the fundamental component of the square output voltage:
Based on FHA, the transmission gain of the resonant tank for the PWM mode
[
27] is provided by:
where
G is defined in Equation (
4).
On the other hand, in order to respect FHA and ensure appropriate resonant operation of the DC–DC LLC converter, the duty cycle variation should be limited. In cases of high duty cycles, the ZVS property can be lost.
2.2.3. PSM
The Phase-Shift Modulation (PSM) technique [
42] is a method of modifying the phase-shift angle between two signals with a constant signal frequency. The converter is controlled by varying the phase shift between the MOSFET control signals, with fixed switching frequency and fixed duty cycle.
The PSM strategy is a very effective technique for the DC–DC LLC converter, overcoming the drawbacks of the PFM strategy and avoiding the limitations of the PWM strategy.
By adjusting the phase shift between the control signals of the MOSFETs, which have constant switching frequency and duty cycle, the DC–DC LLC gain can be changed. The input voltage’s fundamental harmonic is impacted by the controlled phase shift.
The duty cycle is set to 0.5 for all MOSFETs, where the controlled phase-shift angle
is specified, as shown in
Figure 4, as the phase-shift angle between the left (MOSFET S6 (inversely S5)) and right arms (MOSFET S7 (inversely S8)) in the identical H-bridge in
Figure 2.
The fundamental component of the input voltage of the LLC resonant tank has the following expression, which relies on
, based on small signal modeling (FHA) and using asynchronous clamped mode (ACM) [
31,
43]:
The fundamental component of the LLC resonant tank output voltage is:
The LLC transmission gain of the resonant circuit
in case of the PSM strategy is then given by:
where
G is expressed in Equation (
4) and the phase shift
is adjusted between 0 and 0.5.
To summarize, the main characteristics and limitations of different modulation strategies, applied to the DC–DC LLC converter for an EV charger application in V2X mode, are shown in
Table 1.
The FHA model based PFM strategy represented by Equation (
4) is used to obtain the controlled switching frequency as detailed in [
15,
24]. In order to ensure the ZVS operation, the switching frequency should be limited. For a wide operating zone in V2X mode, especially under light load conditions (for low power loads), the PFM strategy is not sufficient for reaching all the operating points because the switching frequency is saturated, causing low conversion efficiency. The PWM strategy, represented in Equation (
7), is able to reach more operating points by fixing the switching frequency and varying the duty cycle. However, a wide operating zone requires a large duty cycle variation that is not coherent with the FHA principle and causes reduced efficiency. Then, the model based on the PSM strategy (
10) is proposed to avoid the limitations of PFM and PWM by fixing both switching frequency and duty cycle and varying the phase-shift angle in order to reach the whole operating zone in V2X mode with improved efficiency.
2.3. Large Signal Modeling
2.3.1. PFM
The large signal model can be split into three parts [
48]: write the nonlinear dynamic state equations, use variables’ harmonic approximation and apply the Extended Describing Function (EDF) [
20,
21,
22,
23] for approximation of nonlinear terms.
The equivalent circuit of the DC–DC LLC converter in V2X mode is presented in
Figure 5.
The LLC converter resonant tank receives a square wave voltage created by the full bridge. The output capacitor’s parasitic resistor is ignored [
49].
The nonlinear dynamic equations, based on the LLC equivalent circuit in
Figure 5, can be defined [
20,
23,
49] by applying the Kirchhoff’s laws, as in Equations (
11)–(
13):
where
is a square wave voltage produced by the full-bridge switches and applied to the resonant tank in V2X mode. The resonant current
, capacitor voltage
and DC bus voltage
are also state variables.
The AC state variables are decomposed into sine and cosine components, and the derivatives are equal to zero to yield the steady-state values using the sinusoidal approximation. On the other hand, this decomposition provides two states for each AC variable, resulting in a higher-order dynamic model. The approximation for the series resonant current and its derivative can be presented as follows in Equations (
14) and (
15), respectively [
20,
23,
49].
Likewise,
and
can be divided into sine and cosine components, as shown in Equations (
16) and (
17):
It is worth noting that nonlinear terms such as
and
appear in the dynamic Equations (
11)–(
13). The EDF concept is a strong mathematical tool for modeling and studying the dynamic behavior of resonant converters. By breaking modulated waveforms into sine and cosine waveforms, this method combines time domain and frequency domain analysis to derive the model. The fundamental harmonic terms can be used to approximate the nonlinear terms.
The nonlinear terms could then be expressed by their sine and cosine components using EDF [
20,
22,
23,
49,
49] as presented in Equations (
18)–(
20):
where the EDF parameters are F1 (
d,
), F2 (
,
), F3 (
,
) and F4 (
,
). The variable
is defined as in Equation (
21):
It should be noted that d is the duty cycle, which is fixed at 50%.
The EDF parameters are defined as in Equations (
22)–(
25).
By making use of EDF terms and the harmonic approximations, splitting the sine and cosine terms, the following Equations (
26)–(
30) are obtained [
20,
23,
49]:
These equations represent the large signal dynamic model of the DC–DC LLC converter in V2X mode based on the PFM strategy.
When a large signal transient disruption occurs, this model guarantees enough dynamic information for the DC–DC LLC converter. However, this model is used with the PFM strategy that needs wide switching frequencies to cover the entire operating zone. A wide switching frequency causes DC–DC LLC converter saturation, preventing achievement of good control performance and reducing the efficiency. For this reason, this model will be rewritten, in the next section, combined with the PSM strategy that can avoid the PFM limitations and cover the whole operating zone. Furthermore, the proposed model based on PSM strategy allows us to design and apply nonlinear robust controllers in order to ensure robustness of the control against disturbances.
2.3.2. PSM
The PSM strategy is proposed to get around the constraint of the PFM strategy in V2X mode in order to improve the DC–DC LLC converter efficiency and control the operating points that the PFM strategy cannot reach in V2X mode [
24]. Based on a small signal model with FHA, the PSM strategy in the last section was implemented based on the gain transfer function in V2X mode. Small signal modeling enables creating a transfer function based on an equivalent model, which makes the control simple to implement. It does not, however, account for nonlinear and uncertain effects, such as the DC–DC LLC converter structure, which has an impact on the robustness of the applied control law with regard to system perturbations and results in a constrained operating zone.
Large signal modeling [
20,
23] has been proposed for the DC–DC LLC converter, based on the PFM strategy, to provide enough dynamic information when large signal transient disturbance occurs. However, this model is used with the PFM strategy that needs wide switching frequencies to cover the entire operating zone, causing LLC converter saturation and reduced efficiency. For this reason, this model has been rewritten and combined with the PSM strategy that can avoid the PFM limitations. Furthermore, the proposed model based on PSM strategy allows us to design and apply nonlinear robust controllers in order to ensure the control robustness against the disturbances.
The equivalent circuit of the DC–DC LLC converter in V2X mode is presented in
Figure 5. Based on
Figure 5, the square voltage
represents the input voltage in V2X mode, which depends on the battery voltage
and the controlled phase-shift angle
in the case of the PSM strategy. Its fundamental component is expressed in Equation (
8).
The nonlinear dynamic equations of the DC–DC LLC converter are defined in Equations (
11)–(
13). Like the large signal model in case of PFM strategy, the resonant tank’s AC variables can be divided into sine and cosine components by using the sinusoidal approximation. The following expression (
31) presents the resonant circuit’s input voltage in V2X mode:
with
where
f is the switching frequency.
is the RMS value of the fundamental component of the resonant tank’s input defined in Equation (
8) and is a function of the phase shift
. The approximation of the series resonant current and its derivative is given in Equations (
14) and (
15). Likewise,
and
can be divided into sine and cosine components, as shown in Equations (
16) and (
17). The terms
and
are approximated by their sine and cosine components as shown in Equations (
19) and (
20).
Using the sinusoidal approximation of each variable as given in Equations (
14)–(
17), (
19), (
20) and (
31) in the LLC dynamic equations (
11)–(
13), and splitting the sine and cosine terms, the following equations, (
32)–(
36), can be obtained:
These equations present an improved model of the DC–DC LLC converter combined with the PSM strategy in V2X mode. This model describes the LLC dynamics and provides enough information about the resonant tank, the DC bus and the switching network. It provides the necessary information of the LLC dynamics (partially known) to design a nonlinear and robust control law.
A comparison between the small signal modeling and the large signal modeling is conducted in
Table 2. The advantages and drawbacks of each modeling approach are highlighted.
A transfer function based on an averaged or linearized mode can be obtained using small signal modeling, making it simple to implement control. It does not, however, account for nonlinear and unpredictable effects, such as the DC–DC LLC converter structure, which has an impact on the robustness of the applied control law with regard to system disturbances and results in a constrained operating range. From here, the large signal model appears important for obtaining enough information about the DC–DC LLC converter dynamics (partially known) and help with the design of nonlinear and robust control laws.