Improvement of Power Quality for Grid-Connected Microgrid Based Multi-Source Power Systems Using an Integrated Load-Following – Artificial Neural Network Strategy

The proposed DC/AC microgrid is shown in Figure 1. The entire system complements are a solar PV system, ESU DC/AC converter, EV load profile, three phase AC load, and utility grid. The PV system is connected to the DC bus via DC/DC boost converter, which adjusts the MPP via MPPT tracking method. The ESU is controlled using a bidirectional DC/DC converter that adjusts the charging or discharging of the battery energy storage system (BESS). The EV load profile used in this work is based on the ECE-15 Urban drive cycle, which represents the vehicle speed. In addition, the system is connected to the grid via 0.4 KV/33 KV transformer via 33 KV, 50 Hz utility grid.

Figure 1. Proposed system-based microgrids (MG)

2.1 Solar photovoltaic system modeling

Figure 2 shows the single diode model of a PV module or array. This model describes the electrical behavior of the solar system. The equation that describes the current-voltage (I-V) characteristics of a solar cell/module using the single-diode model is as follows [23]:

$I_{p v}=N_p I_{p h}-N_p I_o\left[\exp \left(\frac{V_{p v}+R_s\left(\frac{N_S}{N_p}\right) \cdot I_{p v}}{\alpha V_{T h} N_S}\right)-1\right]-$$\frac{V_{p v}+R_s\left(\frac{N_S}{N_S}\right) \cdot I_{p v}}{R_{s h}\left(\frac{N_S}{N_P}\right)}$     (1)

where,

• $I_{p v}$ is the output current (photocurrent).
• $V_{p v}$ is the output voltage.
• $I_{p h}$ is the photocurrent generated by the cell.
• $I_o$ is the reverse saturation current of the diode.
• $\alpha$ is the diode ideality factor.
• $R_s$ is the series resistance.
• $R_{s h}$ is the shunt resistance.
• $V_{T h}$ is the thermal voltage, approximately 25.85 mV at room temperature (T = 25 ℃).
• $N_S$ is the number of series modules.
• $N_P$ is the number of parallel strings of array.

Figure 2. Solar array model

2.2 EV modeling

The mathematical equation of the total motor force of the vehicle can be expressed as follows:

$F_T=F_m+F_r+F_{a d}+F_U$     (2)

where, $F_m$ represents the motor force, $F_r$ is the rolling resistance force, $F_{a d}$ represents the aerodynamic force, and $F_U$ represents the grade ability force. These forces can be expressed as follows [17]:

$F_m=\left(M_v+J_{e m} \frac{\rho^2}{R^2}\right) \frac{d v(t)}{d t}$     (3)

$F_r=C_r M_v g \cos (\alpha)$     (4)

$F_{a d}=0.5 \rho v^2 A C_d$     (5)

The parameters of the above equations can be defined as:

• $M_v$ is the vehicle mass.
• $J_{e m}$ is the inertia of the motor.
• $\rho$ is the air density.
• $R$ is the tire radius.
• $C_r$ is the rolling friction coefficient.
• $g$ represents the gravity acceleration.
• $\alpha$ is the road slop.
• $v$ is the vehicle speed.
• $A$ is the frontal area.
• $C_d$ is the aerodynamic drag coefficient.

However, the load power produced by the traction engine on the DC link ($P_L(t)$) is written as follows [17]:

$P_L(t)=F_T(t) \times v(t) \times \eta_t$     (6)

where, $\eta_t$ is the mechanical transmission efficiency.

The abbreviated form of the urban driving cycle, i.e., the ECE-15, is used to represent the EV load. The ECE-15 model was shown in Figure 3. Moreover, the driving cycle has been carefully crafted to capture the inherently dynamic nature of urban driving conditions, involving numerous phases of acceleration as well as deceleration, and segments of idling. All of such driving activities translate into time-varying power requirements that must be properly managed by the system for peak performance. By using the ECE-15 driving profile along with system operating conditions, the authors of this paper model their EMS response to real variations in the power demand of an electrically powered vehicle (EV). Moreover, the used EMS can perform under heavy load changes, keep the DC-bus voltage stable, and manage the power flow between the PV source, BESS, and the grid.

Figure 3. Electric vehicle (EV) profile-based reduced ECE- driving cycle

2.3 Energy storage unit

The ESU under consideration here depends upon battery energy storage systems (BESS) to effectively reduce power swings and supply/output for the microgrid's stableness [24-26]. Batteries possess the answer to combat the intermittency of renewable sources of energy, i.e., the PV sources, by conserving excess generated energy at high-solar-irradiation moments and releasing energy at maximum demand or low-generation moments. This helps significantly to enhance system flexibility and reliability. In this regard, lithium-ion (Li-ion) batteries are chosen due to the high-energy storage capacity or high energy density, higher cycle life, higher efficacy, and fast response to transients of load conditions. Advanced BMS actively monitors and controls SOC, and voltage, and temperature of the battery and thereby ensures safe functioning and maximization of the lifetime of the battery [26].

The battery system functions at pre-defined SOC levels (say, 20% to 90%) and inhibits deep discharges and overcharges and thereby sustains the lifespan of the battery and achieves optimized functioning. The interfacing of the battery ESS with the energy management system surely enhances renewable energy penetration and minimizes dependence upon fossil-fuel-based backup sources and minimizes operating expenses and adverse environmental effect and achieves sound microgrid stableness. The SOC battery equation is expressed as [24]:

$S O C_{b t}(t)=S O C_{b t}\left(t_o\right)+\frac{1}{C_n} \int_{t_o}^t i_{b t}(t)$     (7)

where, $\operatorname{SOC}_{b t}\left(t_o\right)$ is the initial SOC of the battery, $C_n$ is the nominal capacity of the EV battery. Figure 4 shows the electrical model of the lithium battery.

Figure 4. Electrical model of the lithium battery

Figure 5 illustrates the suggested ESU control scheme. The ESU integrates consists of the bi-directional DC-DC converter, which functions under the diligent supervision of a suggested EMS along with a Pulse Width Modulation (PWM) generator. However, the BESS basically depends on functioning as a dynamic energy buffer; it can uniquely absorb the surplus energy that can be unconventionally produced, e.g., from solar panels, or, on the other hand, release power to the whole system when there is a lack of energy. The bi-directional converter plays an enormous and vital role in that it enables two-way power flow, which can be transferred by its two operating modes. In the boost mode operation, the converter enables the battery to be charged by increasing its voltage level when there is a large amount of excess energy available; such operation guarantees that the excess energy is exploited efficiently.

The coordinated control plays a significant role towards ensuring that there remains effective power exchange, as well as stability of the DC bus voltage. Additionally, it benefits the battery through minimizing instances of overcharging or deep discharging of the battery pack. Thus, the EMS ensures improved system reliability as well as responsiveness of the energy system.

Figure 5. Energy storage unit (ESU)-based microgrid system

The simulation results of the proposed work are obtained using MATLAB/Simulink software. The proposed EMS was verified and tested under different case studies to prove the performance of the proposed EMS. The parameters of the system are listed in Table 1. The case studies are discussed as follows.

Table 1. Simulation parameters of the system

4.1 Case study 1: Variable irradiance evaluation with consumption EV load

Figure 10 shows the profiles of solar irradiation, temperature, EV drawn EV current, and SOC of the BESS. As shown in this figure, the simulation time interval was taken for 2.5 seconds. The irradiance was changed during this duration to show the increasing or decreasing the PV generation. The noticeable decrease in irradiance levels is significant, since it drops considerably from its initial measurement of 900 W/m² to 500 W/m², showing a considerable decrease in solar resource availability. Despite these unexpected and abrupt changes, the designed EMS, is capable of taking over and immediately responding to these changes. Hence stabilizing the operation. In this case, study, the PV temperature keeps constant with 25 ℃ degree. The current drawn by the EV load is varied as presented in Figure 10(c). This figure gives a detailed and complete description on the charging current profile of the EV during consumption duration. Moreover, the proposed EMS is responsible for controlling and managing power interchange from the battery to the EV. Thus, a high-magnitude levels up to 200 A can be achieved without instability for the trajectory of the SOC. Therefore, the SOC goes through a highly precise evolution throughout the entire duration of the test situation. Even with the dramatic reductions in irradiance and the intermittent nature of charging for EVs, the SOC is found to hover in an extremely tight and stable band ranging from 69.9995% to 70.002%.

Figure 10. (a) Irradiance graph, (b) temperature, (c) electric vehicle (EV) current, (d) state-of-charge (SOC) of the battery

The obtained results in terms of DC bus voltage, active power of all sources, injected AC current of grid, and dq-axis currents are illustrated in Figure 11. In this figure, the DC link voltage demonstrates the ability of the presented LF-ANN based EMS strategy to tightly control the measured voltage to its desired value (700 V). Therefore, the presented deviation in the voltage less than 2% even with the high transient consumption EV load, and PV generation fluctuations. This confirms the fast response and optimal power sharing via the applied EMS response. As observed in Figure 11(b), the active power of the grid is changed from positive sign to negative when the consumption power more the generation during the reduction the solar irradiation. The BSU reference power discharges the DC bus based on ots SOC value while the excess power is transferred to grid when the irradiance is high. The difference between the PV power and the consumption is used to charge the BSU or exported to the utility grid via DC/AC converter. As a result, the obtained results show that the recommended EMS ensures a high optimization ability to regulate the bus voltage and controls the RES production. Figure 11(c) displays the AC injected current to the grid during this scenario. The peak RMS current value is 2.3 A.

The EMS controls the export and import operations of the grid based on the PV generation. The smooth changes in current show that the grid-side converters and local resources are working together well without causing harmonic distortion. Figure 11(d) shows the dq-axis currents, when power is exported or absorbed from the grid. As shown in the $I_d$, it is identical to the reference value of $I_d$ = 1 A and its value changes from positive to negative when power is injected into the grid, and the current turns to the negative part when power is absorbed from the main grid. Furthermore, the q -axis of the reactive power $I_q$ = 0 A.

Figure 11. (a) Direct-current (DC) bus voltage (b) active power of system (c) injected current to grid (d) dq-axis currents

Figure 12 shows the active and reactive power of grid, high voltage side of grid and 3-ph load current. The active power value varies depends the PV generation and load power as shown in time interval 0 seconds to 1 seconds, where the active power is positive (100 KW) which means the required power for the BSU and consumption in the 3-ph load and EV model is covered by the PV system. After that, when the consumption load of EV increases at time 1.4 sec, the active power is absorbed from the grid. During this transient, the reactive power of the system is zero, which confirms the reliability of the proposed EMS. The grid voltage at right side of the transformer still constant with 33 KV. The power drawn from the load also constant with peak ac value about 85 A.

Figure 12. (a) Active and reactive power of grid, (b) high voltage side of grid, (c) 3-ph load current

Figures 13 and 14 display the grid current with voltage under scenario 1. The grid voltage is scaled by $V_a / 5000$ for plotting with the corresponding AC current ($I_a$). The zoomed values as seen in Figure 14 during small interval time of (1.93 sec-2.1 sec) are presented to show the current reduction or increasing. The grid current value about $\pm$ 1.6 A with pure sinusoidal waveforms. The phase shift between the voltage and current is zero and this means the power factor of the inverter is unity. This result confirms the bidirectional power flow from the inverter to grid. Also, the sinusoidal shape of current and the absence of harmonic distortion demonstrate compliance with IEEE 519 current quality standards. Table 2 reports the obtained results at case study 1. The achieved results demonstrate that the latest LF ANN EMS can still tightly regulate the DC-link voltage very efficiently despite moderate operating variations. The DC bus voltage is always maintained very close to the setpoint (700 V) with limited overshoot and undershoot; hence, control improvement is confirmed through enhanced damping and faster recovery as compared with the conventional one. Additionally, the grid reactive power is kept nearly at zero level; therefore, the power factor is almost unity (PF > 0.99), which is proof of the power quality improvement objective. The grid energy exchange reveals that the import energy is at a higher level than the export one; thus, the operating condition is stable, the grid is mainly supporting the load, and the EMS is balancing the power flow. Moreover, the high PV self-consumption rate indicates that the renewable generation is being effectively utilized within the microgrid.

Figure 13. Grid voltage and current at scenario 1

Figure 14. Zoom result of grid current

Table 2. Obtained results at case study 1

4.2 Case study 2: Cloudy irradiance evaluation

In this section, only the solar irradiance is changed as shown in Figure 15. This profile was added to represent the cloudy irradiance during the day. This profile was used to show the performance of the EMS under different weather conditions. Moreover, the achieved results under this scenario are reported in Figure 16. The DC bus voltage has been regulated via the EMS by ensuring a minimum overshot and high dynamic response to the changes in the irradiance. The real power of all sources are varied according to the PV generation due to the consumption EV is still same. The electrical load at time interval (0 sec - 0.5 sec) is 40 kW, ich is more than the generated power from the PV panels, therefore, the required demand is absorbed from the grid while the ESU keeps in charging operation. After this period, the PV power increases due to the increasing the irradiation. The electrical load stays without changing and the EV load is zero, therefore the required power and charging operation for the ESU are absorbed from the PV system. Between intervals (0.5 sec - 2 sec), the PV productivity is high and so the required power is covered, and the excess is exported to the grid as observed in the injected current. During this interval, the BSU can support the EV load based on its SOC limits. Lastly, the EMS successfully optimizes the generation sources, load demand, and charges the BSU by managing the power and controls the reference power conditions. The findings confirm that the LF-ANN can improve the power quality of the grid-tied system via controlled the dq-axis power as shown in Figure 16(d).

Figure 15. (a) Cloudy irradiance profile, (b) temperature, (c) electric vehicle (EV) current, (d) state-of-charge (SOC) of the BSU under scenario 2

Figure 16. (a) DC bus voltage, (b) active power of system, (c) injected current to grid, (d) dq-axis currents under scenario 2

Furthermore, the active and reactive power of the grid under case 2 is shown in Figure 17. In this figure, the grid can support the load by providing a required active power and cancelled reactive power component. This means the EMS of the grid is robustness and the system can share or drawn power from the grid based on the PV generation or load consumption. Figure 18 shows the grid current with its voltage. Figure 19 displays the zoom of the achieved results under this scenario. As observed, the grid current to at a time interval (2.05 sec- 2.16 sec) was injected to the grid. The RMS value of this current ais bout 2 A. At this interval, the reactive power is exported into the grid due to the consumption is low and the BSU is within the same SOC. At same time, the PV power is high and therefore all demand covered by the renewable energy source. Against, the injected current at interval (2.16 sec- 2.27 sec) was shifts by 180° relative to the corresponding grid voltage. This indicates a reversal power flow and the real power consumed from the grid. As observed, the peak RMS current is reduced based on the required absorption value. The transient in the current occurs smoothly with absence high overshoot or distortion, ensuring the accuracy of the grid-tied control scheme and the performance of the LF-ANN method. Table 3 displays the obtained results at case study 2.

Figure 17. (a) Active reactive power of the grid, (b) high voltage side of the grid, (c) electrical load power under scenario 2

Figure 18. Injected current to the grid

Figure 19. Zoom out of the injected current

Table 3 presents a summary of the performances of the proposed LF ANN EMS under case study 2. The DC-link voltage is tightly regulated around the nominal value of 700 V, with the controller limiting the voltage excursions to the upper bound (790 V) and lower bound (640 V), which is a strong indication that the closed-loop system remains stable under the influence of operation disturbances. The overshoot (7.8%) and undershoot (5.5%) reported here illustrate a well-damped dynamic behavior, and the transient deviation is thus sufficiently small for the DC microgrid operation. The grid power factor is kept at a high level (PF ≈ 0.98), which means that there has been an almost total disappearance of reactive power, and thus the power exchanged between the converter and the grid is actually active power only. Furthermore, the current distortion is extremely low (THD ≈ 0.6%), which indicates that the injected current is practically a pure sinusoid and that there is strong adherence to the grid power quality standards. The whole set of results points to the fact that the proposed EMS is very effective in maintaining voltage stability as well as in enhancing the grid-side power quality.

Table 3. Obtained results at case study 2

4.3 Case study 3: Variable weather and step increasing in AC load

In this section, both irradiance and cell temperature are varied with step changes conditions as shown in Figure 20. Moreover, the goal of this case is to validate the suggested EMS under different operation conditions related to the PV generation and load consumption.

Figure 21 introduces the obtained results under varying weather and load conditions. AS observed, the DC voltage regulation via the EMS is achieved. The ripple content or fluctuating in the measured voltage stay within limits. The deviation in the voltage less than $4 \%$, and this confirms the EMS working by controls the derivative of the bus voltage $\dot{V}_{\text {dc,bus}}$ using ANN controller. the PV power from time 0.5 sec to 2 sec is very high and the BSU charging, EV load, and electrical load are covered. Before this interval the grid is covered the loads. At time t = 1 sec, the electrical load is increased from 40 kW to 80 kW.at this moment, the EMS ensures the generation and compute the load consumption to provide the decision. As a result, the PV power is high and this load was converted. After some seconds at time t = 1.2 sec, the load profile of the EV becomes very high as well as the electrical load is high, so the PV and grid are worked together to covers the load and charge the ESU.

Figure 20. (a) Irradiance graph, (b) temperature, (c) electric vehicle (EV) current, (d) state-of-charge (SOC) of the battery under scenario 3

Figure 21. Obtained results under varying weather and load conditions

Figures 22 and 23 display the grid power, volatge and current of the elctrical load. Also, the zoom of this result was shown in Figure 23. As observed, at time t = 1 sec, the electrical load current is increased from 80 A to 175 A. The grid voltage keeps constant with 33 KV. The increasing in the load may cause the power flow changes from the positive to negative. At him moment, the grid power keeps positive (export power) due to the PV power is high. Moreover, this can be seen in the injected current as shown in Figures 24 and 25.

Figure 22. (a) Grid power, (b) volatge of gird, (c) electrcal load current

Figure 23. Zoom of the results

Figure 24. The injected current to the grid

Figure 25. Zoom of the injected current

Table 4 reports the obtained results at case study 3. Although the voltage of the DC bus is showing a bit more overshoot/undershoot, this variation is due to the very rapid changes in load and renewable generation. However, it is still within perfectly acceptable limits and is very fast in coming back to the reference value. The grid power factor stays very close to unity (PF > 0.99), which is evidence that the reactive power exchange has been kept at an absolute minimum even during a very dynamic operation. The grid actively participates in maintaining system stability, as evidenced by the grid import/export energy figures. Meanwhile, the PV self-consumption ratio of approximately 69% indicates that a significant portion of locally generated renewable energy is consumed locally, demonstrating the potential of EMS for coordinated energy dispatch.

Table 4. Obtained results at case study 3

4.4 Comparison with the classical load-following method

To prove the effectiveness of the proposed EMS, the comparison between the results achieved in the classical LF-PI method and the suggested EMS is done. Therefore, the DC bus voltage curves under the third case was selected to compare the performance. In this case, the irradiance, temperature, and electrical load are varied as shown in previous section (case 3). Figure 26 shows the response of the proposed EMS and the classical LF-PI method. The comparison in terms of the DC bus response for the three cases is presented. Then, the overshoot, settling time, and rise time are shown to verify the comparison and displays the best performance as displays in Table 5. As observed, the proposed EMS offers less overshoot and fast time convergence especially when the irradiance changed in time 0.5 sec, while the classical method severs from the unwanted high spikes in the DC voltage, which presents energy loss in the system and lowered the overall efficiency.

Figure 26. DC bus voltage curve under both classical and proposed energy management strategy (EMS)

Table 5. Comparison between the proposed energy management strategy (EMS) and the classical method

Table 6. Comparison with recent state-of-arts energy management strategy (EMS) methods

Table 6 reports a comparative analysis of the proposed EMS with state of arts methods. The comparison was conducted in terms of the complexity, degree of the voltage stability, and type of the configuration structure. From this table, the suggested LF-ANN EMS allows for very simple low-complexity for real-time energy dispatch. It requires less inputs, thus achieving fast DC-link stabilization, near-unity PF with very low THD, and stringent performance under PV/EV/grid operating scenarios without the need for forecasting or communication infrastructure.