**1. Introduction**

The energy policy challenges facing the European Union are greater than ever. One such challenge is the intelligent management of electricity at all levels, from production to final consumption [1,2]. Distribution efficiency and the reliability of service delivery are two indicators that can be used to evaluate the performance of distribution systems [3,4]. The increasing deployment of smart grids has improved their monitoring and controllability [5]. An example of a modern smart grid feature is the ability to assist with load shifting, limit peak demand and automatically identify malfunctions or outages [6]. These issues are all the more controversial as the number of microgrids, i.e., small-scale independent power systems, continues to grow. Two main modes of operation characterize a microgrid: the grid-connected mode and the off-grid mode [7]. In off-grid mode, electricity is either

**Citation:** Aouichak, I.; Jacques, S.; Bissey, S.; Reymond, C.; Besson, T.; Le Bunetel, J.-C. A Bidirectional Grid-Connected DC–AC Converter for Autonomous and Intelligent Electricity Storage in the Residential Sector. *Energies* **2022**, *15*, 1194. https://doi.org/10.3390/en15031194

Academic Editors: Massimiliano Luna and Nicu Bizon

Received: 2 November 2021 Accepted: 4 February 2022 Published: 7 February 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

unavailable due to a grid failure or outage, or the power grid is in islanding mode, i.e., completely inaccessible. The performance of standalone microgrids has been regularly analyzed in the literature for several years. These standalone microgrids typically consist of two key components: photovoltaic (PV) arrays and/or wind turbines and energy storage systems, such as flywheels, supercapacitors or batteries, which are used to implement intelligent voltage regulation and load tracking systems [8,9].

Energy management strategies are currently playing an increasingly important role in regulating the power quality of microgrids [10]. One of the challenges is to control energy flows to achieve various operational objectives, such as cost minimization, guaranteed delivery or security. One method to achieve this is to adjust the flow of energy to and from the main grid, the distribution of energy resources and the controllability of loads. In the residential sector, this type of strategy, now known as a home electricity management system (HEMS), is typically implemented to stimulate the integration of renewable energy and to protect the electricity distribution system from potential outages [11].

With the advent of Internet of things (IoT)-based smart buildings, in which all smart appliances are connected, as well as smart meters, we now have not only real-time usage data but also the ability to remotely operate various appliances in a home [12,13]. Much of this data can be used by an HEMS to improve the synchronization of energy transfers between the storage system, power generation and its consumption.

A wide range of techniques is already available in the literature and these techniques can be classified into two main categories [14–18]. Indeed, these approaches are based either on information processing methods or on optimization tactics:


As shown in Figure 1, Bissey et al. recently proposed an example of HEMS using fuzzy logic control (FLC) [23]. This approach used machine learning and data collected from different houses (i.e., time, day, past power consumption and indoor/outdoor temperature data) to find optimized criteria for the configuration of an HEMS. Although the approach described in this paper is effective, its validity only makes sense if a large experimental database is available to build the fuzzy rule base and membership functions needed to make this HEMS work. Recently emerged deep learning (DL) methods are proving to be effective tools, even surpassing traditional approaches as they greatly reduce the need for human interaction [24,25].

A storage system is required for the installation of such an HEMS. Alternating loads can be powered either directly from the AC grid or through an inverter in case the storage system would relieve the AC grid. To implement these two modes of operation, a necessarily bidirectional DC–AC converter must be operational [26]. Its metrics in terms of compactness, efficiency and output signal quality must be as high as possible. In this context and as illustrated in Figure 1, this paper focuses on the sizing and implementation of such a bidirectional converter.

Multilevel DC–AC converters, introduced in the 1980s, that are widely deployed today and whose performance has been discussed in the literature for many years allow the output voltage to be varied in steps by generating levels [27–29].

**Figure 1.** The principle of the smart HEMS implemented in this study [23].

By significantly increasing the number of levels, the output signals become more similar to sinusoids, which drastically reduces their total harmonic distortion (THD) and thus, improves the power quality of the converter and a fortiori of the microgrid. To increase the number of levels, two approaches are classically implemented: increasing the number of voltage sources on the DC bus side or multiplying the number of semiconductor devices to be controlled [30]. Multilevel converter topologies are still widely used for medium and high voltage applications, such as electrical motor drives or grid connected converters, because they generate very low harmonics [31].

Totem pole topologies are also interesting structures dedicated to bidirectional DC–AC conversion. In this type of structure, bidirectional switches based on MOSFETs, thyristors or Triacs are used. Such a topology is generally suitable for high-power applications. Its two main advantages are the reduction in the number of components to be controlled and the simplicity of the use of the converter itself. Recent studies have demonstrated their energy efficiency for particularly high-power densities [32,33].

Despite their high energy efficiency, totem pole topologies cause electromagnetic interference at high frequencies due to the floating points that exist in these architectures. Multilevel DC–AC topologies, widely used today in high voltage, use particularly bulky filters to limit electromagnetic compatibility problems. Thus, to avoid these problems, we sought to implement a topology more suitable for HEMS applications whilst remaining within the specified power range.

This paper aims to prove experimentally the merits of another type of bidirectional DC–AC converter that is more suitable for HEMS applications. The proposed topology, which has not yet been fully explored in the literature [34,35], is formed by the association in series of a DC–DC stage and a DC–AC stage with these two stages being necessarily bidirectional. In inverter mode, the DC–DC stage generates a full-wave rectified sine wave thanks to its pulse width modulation (PWM) control. This signal is then inverted halfperiod by half-period by the DC–AC stage to create the sine wave signal. In rectifier mode, the DC–AC stage acts as a full bridge while power factor correction (PFC) is provided by the DC–DC stage.

The main contribution of this study is the sizing and experimental validation of the proposed bidirectional DC–AC converter topology coupled with a control strategy for HEMS applications. This converter needed excellent compactness and high efficiency (at least 95%), both at low power (a few hundred watts) and up to 1.5 kW. The HEMS, which is associated with the designed converter, needed to supervise the transition between operating modes, as well as the amount of energy stored or injected into the AC grid.

The remainder of this paper consists of the following sections. Section 2 describes the proposed topology with the above key properties. The sizing strategy of the converter and the main experimental results are presented in Section 3. Finally, a discussion based on the main experimental results that were obtained is proposed in Section 4.

#### **2. Bidirectional DC–AC Converter Topology Proposal and Control Methodology**

In order to validate the fuzzy logic-based control technique implemented in our proposed HEMS [23], we had to design a bidirectional DC–AC converter that was capable of functioning as both an inverter and a PFC rectifier. The transition between these two modes of operation needed to be fully automated and without human intervention in order for our HEMS to autonomously store, produce and supply energy for domestic use.

Before validating its operation by appropriate experimental measurements, it is essential to detail the operating modes of the converter, as well as its control strategies.

#### *2.1. Proposed Topology and Details of Its Operating Modes*

Figure 2 shows the overall structure of the proposed bidirectional DC–AC converter. The energy transfer between a DC voltage source and an AC voltage source, and vice versa, was the basis of this structure. The association in series of a DC–DC stage and a DC–AC stage ensured this principle of operation with these two stages being necessarily bidirectional.

**Figure 2.** Principle diagram of the proposed bidirectional DC–AC converter.

The DC–DC stage needed to generate a rectified sine wave from the PWM command of the power switches (see ❷ in Figure 2). Since the power devices in this stage switched at a frequency of a few hundred kilohertz to optimize the compactness of the whole converter, the inductance, noted *L*1 in Figure 2, was sized so that the ripple of the current was negligible compared to the sinusoidal component at low frequency (in this case, 50 Hz). Therefore, the DC–DC converter acted as a controllable output voltage source.

By changing the voltage of a modulation stage, the output current could be regulated. This is very interesting, especially when the output current is strongly reduced or in the of variable DC voltage. This strategy allowed the output voltage of the DC–DC stage

to be modulated. Specifically, when this modulated voltage was higher than the mains voltage, the output current had a positive value. In the opposite case, the output current was negative.

Thus, two modes of operation were possible:


**Figure 3.** Schematic diagram of the converter in inverter mode.

**Figure 4.** Schematic diagram of the converter in PFC rectifier mode.

The design approach that was chosen allowed the converter to be used in both gridconnected and off-grid modes. Since the microcontroller was synchronized with the AC grid to sensibly drive the DC–DC and DC–AC stages, only the grid-connected mode will be discussed in the remainder of this paper.

The DC–AC stage was responsible for inverting every other sinusoidal half-wave to obtain a full sinusoidal output signal (see ❶ in Figure 2).

Compared to existing voltage source converter topologies, such as multilevel structures, the coupling of a DC–DC stage with an H-bridge has many advantages:


#### *2.2. Modulation of the Output Voltage of the DC–DC Stage*

In a rather classical way, three solutions (see Figure 5) are proposed here to adjust the voltage *Vc* (see Figure 2) of the DC–DC stage. We assume that the whole DC–AC converter operates in inverter mode (see Figure 3) to briefly explain each solution.

**Figure 5.** Solutions to modulate the output voltage of the DC–DC stage.

In the three cases of Figure 5, the output capacitor of the DC–DC stage, noted *C*1, was always used. Its value considered here was 10 μF to avoid a strong ripple of the voltage *Vc*.

The first solution (see Case 1 in Figure 5) was to use the capacitor *C*1 directly. The DC–DC stage then operated as a buck-type step-down converter. Despite the simplicity of this solution since it does not use any power device, it has a major drawback. Indeed, a current could flow inside the transistor, named *T*2 in Figure 2, during the discharge of the capacitor, which could lead to it overheating.

As can be seen in Case 2 in Figure 5, a resistor of a few ohms coupled with a switch could be used to solve the above problem. If the load impedance was too high, then the capacitor used in this case could be discharged through this resistor. This solution was especially interesting when the output power of the inverter fluctuated during operation. However, because of this resistance, this solution strongly penalized the efficiency of the DC–DC stage and, by extension, of the whole DC–AC converter.

The last solution (see Case 3 in Figure 5) was to connect *n*-quadrupoles in parallel. Each quadrupole consisted of a capacitor and a power switch that were in series. Such a solution aimed to fix the ripple of the output voltage of the DC–DC stage. Even though this solution was slightly more complex than the previous ones, adaptation to many more loads could be achieved. Finally, the efficiency of the DC–DC stage was not penalized so much in this case. In the remainder of this manuscript, this solution is implemented with two quadrupoles in parallel with the output capacitor (see capacitor *C*1 in Figure 5) of the DC–DC stage. In conclusion, the values of the three capacitors, *C*1, *C*2 and *C*3, of the modulation stage were equal to 10 μF, 1 μF and 68 nF, respectively (see Figure 5). The value of each capacitor was determined by the ripple of the voltage *Vc* (arbitrarily, we considered 1% here) and the estimated current (see (1)).

$$\mathbf{C} = \frac{I \times \Delta t}{\Delta V} \tag{1}$$

where:


#### *2.3. Control Strategies*

In this section of the paper, we will describe the control strategies of the proposed bidirectional DC–AC converter. We will only give the principles and thus, we will not detail the control circuit or the AC network connection strategy because several patents are pending.

#### 2.3.1. Inverter Mode

In inverter mode (see Figure 3), energy flowed from the storage system to the AC grid. In this type of operation, the most important objective was to control the current injected into the AC grid by regulating the output voltage of the DC–DC stage. Figure 6 shows the overall architecture of the MOSFET control circuit inside the DC–DC stage (see transistors *T*1 and *T*2 in Figure 2). We recall that this stage supervised the generation of a semi-sinusoidal output signal. The possibility of modifying the injected current was the indispensable aspect of this control strategy. Figure 6 shows that the microcontroller was programmed to adjust the duty cycle from zero to one with great precision.

A minor increase or decrease in this variable caused the current injected into the AC grid to increase or decrease. This allowed us to control the selection of the operating mode and provided us with the ability to switch from inverter to rectifier mode, as we will see later.

The DC–AC stage, on the other hand, was controlled using the frequency measurement of the smart meter with a switching frequency of 50 Hz in order to be synchronized with the grid (see transistors from *T*3 to *T*6 in Figure 2).

#### 2.3.2. PFC Rectifier Mode

In PFC rectifier mode (see Figure 4), energy flowed from the AC grid to the storage system. In this kind of operation, the most important objective was to control the absorbed current, which needed a sinusoidal waveform. This objective could be achieved by adjusting the equivalent capacitance of the modulation stage of the DC–DC stage (see Case 3 in Figure 5). As for the inverter mode, we chose the solution with a capacitance *C*2 equal to 1 μF because it offered the best results. The PFC mode, achieved by the DC–DC stage, was essential to meet the requirements of IEC 61000-3-2.

Figure 7 shows the structure of the control circuit of the MOSFETs inside the DC–DC stage. The regulation of the control circuit was made possible by a current sensor and a voltage sensor to realize the PFC strategy.

**Figure 6.** Control strategy of the MOSFETs inside the proposed bidirectional DC–AC converter used in inverter mode.

**Figure 7.** Control strategy of the MOSFETs inside the proposed bidirectional DC–AC converter used in PFC rectifier mode.

The DC–AC stage was controlled by the very same signal as the inverter mode, so it was synchronized with the AC grid at a switching frequency of 50 Hz.

#### **3. Sizing of the Bidirectional Converter and Main Results in Grid-Connected Mode**

*3.1. Specifications, Key Sizing Steps and Selection of the Main Components*

The specifications of the prototype bidirectional DC–AC converter that we realized are listed in Table 1. The electrical waveforms were evaluated based on the topology and control strategies detailed in the previous section.

**Table 1.** The specifications of the proposed bidirectional DC–AC converter.


The approach to the sizing of the converter was classical because the two stages taken separately are now well-known and mastered [36–40]. Of course, this approach is adaptable according to the targeted application and design constraints. Figure 8 presents the key steps for the sizing of the passive and active components of each of the two stages. In this approach, the following parameters were essential to take into account:


As shown in Figure 8, the main components to be sized were the power MOSFETs of both stages [36–38], the inductor of the DC–DC stage [39,40] and the output voltage modulation of the DC–DC stage [26].

For the power MOSFETs, given the high switching frequency of the DC–DC stage (a few hundred kilohertz) and to optimize the thermal management of the components, we chose silicon carbide (SiC) devices. For the DC–AC stage, which switched at the AC grid frequency, silicon substrate components were perfectly suited. Then, whatever the two stages, it was essential to take into account: the voltage withstand; the current flowing in the drain; the switching characteristics, especially the rise and fall times for the components of the DC–DC stage; the gate charge characteristics; the characteristics of the body diodes; and the thermal characteristics, in terms of junction temperature and thermal resistances.

The choice of power MOSFETs also took into account future normative tests, which are essential before the industrialization of the product, such as immunity to burst transients, resistance to electrostatic discharges and electromagnetic compatibility. As shown in Figure 9 and Table 2, the DC–DC stage used two (see *T*1 and *T*2 in Figure 9) SiC MOS-FETs (reference: SCT3080AL; manufacturer: ROHM semiconductor) with, for these two transistors, a nominal drain current and drain-to-source voltage equal to 30 A and 650 V, respectively. Two switching frequencies were considered: 150 kHz and 300 kHz. Switching

the components of this stage at 300 kHz allowed for the drastic reduction in the size of the passive components and a fortiori the optimization of the size of the converter. On the other hand, this penalized the efficiency of the system (the desired efficiency of 95% was difficult to achieve) because the switching losses were very significant (see Section 4.2). Therefore, the switching frequency of 150 kHz was implemented experimentally in order to obtain a compromise between the compactness of the system, its efficiency over the entire power range (i.e., from 100 W to 1.5 kW) and the optimization of its power quality; this last point will be studied in more detail in the near future. As for the DC–AC stage, as shown in Table 2 and Figure 9, it used four MOSFETs (designated *T*3, *T*4, *T*5 and *T*6) on a silicon substrate (reference: IRFPS43N50K; manufacturer: Vishay Siliconix), each with a nominal drain current and drain-to-source voltage equal to 47 A and 500 V, respectively. The power devices were turned on and off with zero crossing of the AC grid voltage to minimize losses.




**Figure 9.** Electrical diagram of the power circuit of the converter.


**Table 2.** A selection of the main components.

Concerning the optimization of the DC–DC stage inductance in terms of electromagnetic characteristics, compactness and mass, we took into account the geometry, toroidal or planar and the type of core, as well as the ferromagnetic material used, according to the frequency and the maximum induction [39,40]. Several technologies were studied in order to determine the most suitable inductance for the sized converter. The choice of the 700 μH inductor (see *L*1 in Figure 9) was based on limiting the current ripple in the inductor to less than 20% of the maximum current (i.e., 5 A here) [41]. We also considered its maximum DC resistance (0.12 Ω) and the evolution of its impedance with frequency. Considering all these elements, we chose the WE-FI leaded toroidal line choke from the manufacturer Wurth Electronik.

Regarding the output voltage modulation of the DC–DC stage, the value of the capacitors (see *C*1, *C*2 and *C*3 in Figure 9) was explained in Section 2.2.

Concerning the servo strategy of the proposed converter, a differential measurement was used to determine the AC bus voltage as a function of the phase potential (*VL*) and the neutral point potential (*VN*). This differential measurement was performed with a voltage divider bridge of the same resistance value, i.e., 47 kΩ (see *R*1, *R*2 and *R*3 in Figure 9). To measure the current flowing between the DC–DC stage and the DC–AC stage, as shown in Figure 9, a 15 A closed-loop Hall transducer (reference: LKSR 15-NP; manufacturer: LEM) was used in series with the inductor of the DC–DC stage.

Finally, an STM32F407VG microcontroller unit from the manufacturer STMicroelectronics drove the entire converter and provided automatic AC grid connection and disconnection. The control board with the microcontroller, the MOSFET drivers and the control strategies of the power components are currently under patent and cannot be detailed in this manuscript.

#### *3.2. Experimental Test Setup and Standby Mode*

A comprehensive experimental process was adopted to validate the two modes of operation of the bidirectional DC–AC converter proposed here when connected to the AC grid and in a power range up to 1.5 kW. The converter demonstrator naturally included the DC–DC and DC–AC stages, but also an adaptive filter, as well as a power supply (i.e., +5 V and +12 V) to power the onboard electronics.

As shown in Figure 10, four 12 V, 7 Ah batteries (reference: NP7-12; manufacturer: YUASA) were used to simulate the storage system. These batteries were associated in series to provide the overall voltage of 48 V (see ❶ in Figure 10). On the AC side, the connection to the power supply was provided by a single-pole variable transformer (reference: SEC2; manufacturer: LANGLOIS) set at 110 V RMS. A line impedance stabilization network (LISN) (reference: PD30; manufacturer: EMC MASTER; main features: monophasic, 220 V, 10 A, 150 kHz to 30 MHz) was used to control the network impedance and ensure measurement repeatability (see ❷ in Figure 10).

**Figure 10.** The experimental test setup.

A high-voltage differential probe (reference: P5205; manufacturer: Tektronix) and a 15 A AC–DC current probe (reference: TCP202; manufacturer: Tektronix) were used to measure the output voltage and current, respectively. The output power of the DC–DC stage was measured using a power meter (reference: PX 110; manufacturer: Metrix). For the measurement of the output power of the DC–AC stage, the same type of electronic meter was used.

To connect to the AC grid, the bidirectional DC–AC converter used a patent-pending automated routine. Figure 10 shows the converter in a steady state with the grid at 0 watts (see ❸ in Figure 10) and the batteries powering the system. The power consumption of the converter in sleep mode was 3.6 W (see ❹ in Figure 10). The DC bus was monitored (see ❺ in Figure 10) and displayed the DC voltage (see ❸ in Figure 2), while differential probes were used to monitor the voltage generated between the two stages of the converter, as well as that between the converter and the grid connection (see ❻ in Figure 10, as well as Figure 11).

**Figure 11.** Experimental validation of the standby mode: output signals.
