Hybrid AC/DC control techniques with improved harmonic conditions using DBN based fuzzy controller and compensator modules

Hybrid AC–DC microgrid provides highly distributed generation receiving capacity by reducing multiple power conversions in individual AC or DC microgrids. Thus, the control of grid variables in a hybrid AC–DC microgrid is complicated due to erroneous frequency transients and the impact of negative sequence components. Hence, a novel Synchronous Reference Frame Phase-Locked Loop (SRF-PLL) with Self-curing decoupling network is introduced in which this network provides synchronization and distinguishes positive and negative sequences in discrete blocks. To eliminate fluctuation of grid variables in ac and dc sides, a novel Switched Tuned Arm Filter (STAF) and Variable Voltage Stabilization Compensator (VVSC) Multi-loop controller including central Deep Belief Network (DBN) with Fuzzy controller is presented. In which STAF Multi-loop controller use tuned circuits to suppress higher-order harmonics. Also, a central DBN with a fuzzy controller provides a control signal to an ac/dc converter to regulate power transmission. Additionally, in order to stabilize the dynamic voltage and compare the DC bus voltage with the reference value, the VVSC Multiloop controller uses a PWM to compare the DC bus voltage. Then the proposed techniques are simulated in Simulink platform and the results achieved stabilized voltage, low RMSE, and low delay.


Introduction
Reduced conventional resources, environmental concerns, and rising fossil fuel prices, among other factors, have resulted in the astonishing rise of decentralized electricity generation which now constitutes the framework of today's power networks.Distributed generation (DG) systems based on renewable energy sources (RES) (Al-Shetwi et al., 2020;Brodny & Tutak, 2020;Lu et al., 2020) such as wind, solar, and hydro are becoming increasingly popular ways to reduce the world's reliance on fossil fuels, which are known to deplete sooner or later.The ability of energy storage systems (ESS) (Stecca et al., 2020) to compensate for the intermittent nature of renewable energy sources (RES) and load disruptions in microgrids(MGs) (Espina et al., 2020;Hu et al., 2021;Vasilakis et al., 2020; CONTACT Yanna Reddy yannareddy.p@gmail.comDepartment of Electrical Engineering, National Institute of Technology (NIT), Silchar, Assam 788010, India Zia et al., 2020) has been demonstrated.Modern MGs are made up of a variety of power electronic equipment, as well as multiple energy sources and local loads, and they constitute a viable alternative to the traditional grid.The connectivity of numerous MGs regions has been recommended in the case of sustainable development to combine the rotational inertia to assist one another during crucial scenarios and allow each unit to run efficiently.
A microgrid is a small-scale electrical system that combines distributed generation (DG) and energy storage devices (ESD) technologies to fulfil local load demand.The microgrid can function in both linked and independent modes of these technologies to work together.Such features expand the system's diversity and power flow options, improving the energy supply's efficiency, dependability, and quality.Microgrids are classed as AC microgrids (Lai & Lu, 2021;Sahoo et al., 2020), DC microgrids (dos Santos Neto et al., 2020;Gallo Julian et al., 2020), or hybrid AC/DC microgrids (Wang et al., 2020) based on the sources and loads linked to them.In comparison to traditional single AC microgrids or DC microgrids, AC/DC hybrid microgrids (Naderipour et al., 2020) have the advantages of convenient DC load access, high distributed generation acceptance capacity, flexible and diverse networking form, and control mode, all of which can improve system security and stability while reducing network loss.Optimal dispatch, as one of the key microgrid technologies, is critical for dealing with the volatility of renewable energy and load, as well as ensuring the microgrid's economic and dependable operation.The optimum dispatching of an AC/DC hybrid microgrid is more difficult than that of a standard AC microgrid because of the features of AC/DC operation zoning, energy coupling, source load diversity, and operation mode diversification.An islanded hybrid ac/dc microgrid (IHMG) is a small-scale power system made up of interlinking converters (ICs) that connect ac and dc microgrids (MGs).Grid feeding, grid forming, and grid supporting converters are the three types of converters used in microgrids (Matute et al., 2021).The grid-feeding converter is a current source with a high parallel output impedance that may inject both real and reactive power into the utility grid.The utility grid/ grid forming or supporting converter controls the amplitude and frequency of the voltage source converter's (VSC) output voltage.In the islanded mode of microgrid operation, the gridforming converter is used as an AC voltage source with low series impedance, which establishes voltage and frequency references for the microgrid.However, the existing control techniques have problems in obtaining synchronization, stabilization of voltage, harmonic distortion, and issues in load power-sharing.Hence to solve the problems in existing ac/dc hybrid control techniques, a novel solution has to be developed.The main contributions of this paper are as follows: • Synchronous Reference Frame Phase Locked Loop (SRF-PLL) provides synchronization and estimates the frequency and phase of grid voltage.• Self-curing decoupling network eliminates the impact of negative sequence components on the positive sequence component.• Switched Tuned Arm Filter (STAF) Multi-loop controller in the AC side suppresses the higher-order harmonics at a frequency above or below-tuned values.This research achieves linear load power characteristics and stable voltage without higher-order harmonics in hybrid ac/dc control techniques by providing synchronization and enhancing the controller operation on both AC and DC sides.
The content of the paper is organized as follows: Section 1 represents the introduction; Section 2 presents the related work; the novel solutions are presented in Section 3; the implementation results and their comparison are provided in Section 4; finally, Section 5 concludes the paper.Buduma et al (2021) provided islanding detection, control, power-sharing, and grid synchronization approaches in this study for the flawless functioning of an AC microgrid in Grid-Connected Mode (GCM), Islanded Mode (IM), and the transition between the two modes.A Master-Slave approach is used to distribute electricity across microgrid sources.Each source's control strategy is built on a Robust Linear Quadratic Regulator (RLQR) and a Mixed Optimal (MO) controller.Synchro-Extraction Transform (SET) and Optimal Support Vector Machine (OSVM) are used to create the detection method.For the reconnection of the Microgrid to the utility grid, a Phase-Locked Loop (PLL) based grid synchronization mechanism is used.However, during synchronization double frequency oscillation to the phase error signal is introduced with a negative impact on the performance.Lv et al. (2020) presented to provide power restriction and state of charge (SOC) protection, a decentralized generation storage subgrid coordination control.A modified droop mechanism is used in the BES control strategy to send the storage's SOC and output power signals without the need for communication connections.Meanwhile, in BPC, a fuzzy logic controller is used to avoid misuse of BESs in both subgrids and to restrict BPC's maximum output power.However, this technique requires trial and error methods adopted for the membership functions which is a time-consuming process.

Literature survey
Bhat Nempu et al (2020) presented the assessment of the operation and control of a hybrid AC-DC microgrid using alternative energy sources.Incorporating fuzzy logic and adaptive neuro-fuzzy inference system (ANFIS) based controllers into the battery control strategy is investigated.In the DC subgrid, a photovoltaic array and fuel cell system are connected to the DC bus, and energy is stored in an electrolyzer-based hydrogen storage system.An inverter feeds the wind energy system's output, as well as the batteries, to the AC bus.The AC and DC subgrids are linked by an interlinking converter, which is regulated to allow power transfer between the subgrids dependent on the frequency of the AC bus and the voltage of the DC bus.This employs coordinated control techniques to ensure appropriate power management inside subgrids and between subgrids.However, in this technique, the DC bus voltage deviation is not eliminated.Jena et al. (2020) presented back-to-back converters to try to solve the problem of power-sharing in networked hybrid AC/DC microgrid clusters.For both AC and DC microgrid clusters (intra-microgrid control), a hierarchical distributed cooperative control approach is used, as well as an inter-microgrid control strategy that uses a back-to-back converter to enable power-sharing across the clusters based on the demands.The AC and DC MGs' distributed secondary control aids in reprimanding voltage decreases caused by cable resistance and droop characteristics.As a result, it aids in the achievement of a regulated voltage at both the AC and DC.It improves voltage and power quality, especially in AC MG.However, this technique causes more signal transmission delays.Wang et al. (2020) presented an adaptive droop control technique for AC/DC hybrid micro-grid interlinking interface to improve the AC/DC sub-grid voltage accuracy of the classic normalized droop control.Not only can the suggested control method eliminate wasteful power exchange while guaranteeing bidirectional power flow between AC and DC sub-grids, but it can also increase power transmission accuracy to provide better voltage deviation correction.The link between the droop coefficient and transmitted power is first determined for the island mode of a low-voltage hybrid AC/DC microgrid.The voltage variation is then used to build the operating criteria that describe the system load status.However, in this technique, the sensor inaccuracy may not detect small voltage differences.An and Yang (2019) proposed a distributed SSE algorithm using distributed nonconvex optimization protocols based on consensus.The approach is implemented over a multi-agent network, where each agent discreetly analyses individual sensing measurements while interacting with its neighbours via a graph topology.Through the employment of a transformation technique and a distributed vote location approach, the combinatorial difficulty brought on by the sparse sensor assaults is successfully resolved.Future research focuses on learning how to create a safely distributed algorithm in the presence of adversarial agents that have the freedom to update its state in any way, as opposed to using a predetermined method.
An and Yang (2021) proposed two distributed state estimation algorithms for byzantine links/nodes produced by hostile attacks in static linear cyber-physical systems.First, a basic secure distributed algorithm is given that uses a local min-switching decision (LMSD) rather than the current coordinate-wise trimmed means (CWTM) to reduce the impact of Byzantine links/nodes.In the presence of a predetermined number of Byzantine links or nodes as well as sensor noise, necessary and sufficient criteria on the network connectivity are given to ensure that all regular nodes asymptotically converge to the least square estimate of the system state.Future work will combine the classic gradient/subgradient descent technique with the proposed min-switching technique to extend the proposed algorithms to nonlinear measurement models.
In Buduma et al. (2021) during synchronization double frequency oscillation to the phase error signal is introduced with a negative impact on the performance and (Lv et al. 2020) requires trial and error methods adopted for the membership functions which is a time-consuming process.In Nempu et al. (2020), the DC bus voltage deviation is not eliminated, and (Jena & Padhy, 2020) results in more signal transmission delays.In (Can et al., 2020) the sensor inaccuracy may not detect small voltage differences.In An and Yang (2019) requires distributed algorithm and in An and Yang (2021) need to combine the classic gradient descent technique.Hence, to overcome the above-mentioned problems, a novel technique has to be developed.

Hybrid AC/DC control techniques with improved harmonic conditions using DBN based fuzzy controller and compensator modules
The hybrid AC-DC microgrid connects a variety of AC and DC-based renewable energy sources and decreases the number of reverse conversions due to the existence of separate AC and DC grids.However, the existing techniques cause spurious frequency transients with the influence of the negative sequence component on the positive sequence component during synchronization which results in a delay in detecting the frequency changes and complexity to perform phase detection.To tackle this issue, a novel Synchronous Reference Frame Phase Locked Loop (SRF-PLL) with Self-curing decoupling network has been proposed to detect the delay in frequency variations and phase detection complications.In this proposed technique, Synchronous Reference Frame Phase Locked Loop (SRF-PLL) provides synchronization by using a Self-curing decoupling network, which eliminates the negative sequence components.This impact the positive sequence component by identifying the positive and negative sequences in distinct blocks without extra load or external operations.Hence, it reduces the delay in detecting frequency changes and the complexity to perform phase detection.Moreover, in the existing techniques, without controlling the frequency of AC and DC bus voltages, grid variables oscillate with unbalanced and

SRF-PLL with Self-curing network decoupling
Hybrid ac/dc microgrids are gaining popularity as an ideal solution since they incorporate the benefits of both ac and dc microgrids.However, conventional approaches create false frequency transients during synchronization due to the impact of the negative sequence component on the positive sequence component, resulting in a delay in detecting frequency shifts and increased phasedetection complexity.Hence, a novel Synchronous Reference Frame Phase Locked Loop (SRF-PLL) with Self-curing decoupling network has been proposed in which Synchronous Reference Frame Phase Locked Loop (SRF-PLL) provides synchronization and estimates the frequency and phase of the grid voltage.The Self-curing decoupling network eliminates the influence of the negative sequence component on the positive sequence component by determining the positive and negative sequence in the separate block without requiring additional load and any external operation.As a result, when the gridconnected state is altered by nonlinear load, precise phase detection is achieved.
Figure 2 depicts the structure of a Synchronous Reference Frame Phase Locked Loop (SRF-PLL) with a Selfcuring decoupling network.The three-phase voltage w abc is converted to a two-phase static coordinate system w αβ , which offsets the zero-sequence component and reduces From Equation (1); ω * T is the phase voltage.To suppress irregularities in the coordinate systems d +1 and q +1 , d −1 and q −1 are turned into a self-curing decoupling network, and the positive and negative sequence voltage after decoupling is achieved in Equation (2).
Where ϕ * is the frequency in the phase.Finally, the Power integrity (PI) controller is utilized to make w * q +1 = 0 synchronized and phase-locked, thereby allowing obtaining the frequency and phase information of the three-phase voltage.Hence, a Synchronous Reference Frame Phase Locked Loop (SRF-PLL) with Self-curing decoupling network removes the spurious frequency transients and provides better performance for phase detection.However, the frequency of ac and dc bus voltage and the higherorder harmonics are regulated in the next subsection.

STAF and VVSC Multi-loop controller including central DBN with fuzzy controller
To eliminate the oscillation of grid variables and the unbalanced load power characteristics, a novel STAF and VVSC Multi-loop controller including DBN with the fuzzy controller has been introduced to enhance the higher order harmonics with the improper regulation of the ac and dc bus voltage.This technique, Switched Tuned Arm Filter (STAF) Multi-loop controller used on the ac side to suppress the higher order harmonics.Also, to control the operation of the bidirectional converter and to regulate the frequency of ac and dc bus voltages, Deep Belief Network (DBN) with a Fuzzy controller is utilized.To create the DBN, a fuzzy controller was first executed and then confirmed.Finally, by using the input/output data of the belief network apart from the executed fuzzy layer.To replace the fuzzy controller, the DBN was modified and as a result, it will finally be controlled by the DBN.The Deep Belief Network tool compartment is used to structure the fuzzy logic.Its hidden layer now consists of two layers instead of the original one.Moreover, on the dc side Variable Voltage Stabilization Compensator (VVSC) Multi-loop controller is placed to regulate the non-linear characteristics with the reduction of transient voltage.
Figure 3 depicts the Switched Tuned Arm Filter (STAF) Multiloop controller at tuned frequency has low impedance due to the tuned circuits present in it and effectively shunts the higher order harmonics by reducing the amplitude of current and voltage harmonics in the ac microgrid at a frequency above or below the tuned values.The tuned circuits in the Switched Tuned Arm Filter (STAF) Multi-loop controller are also used to pass the signals.By utilizing three-phase balanced sinusoidal currents with a single power factor in Switched Tuned Arm Filter (STAF), the harmonics and reactive power are corrected, therefore reducing the unwanted effects of the non-ideal primary source.Hence, Switched Tuned Arm Filter (STAF) Multiloop controller eliminates harmonic current and reactive power adjustment with improved power quality.
The bidirectional interlinking AC/DC converter in the hybrid AC/DC microgrid is controlled by using Deep Belief Network (DBN) with a fuzzy controller, which utilizes an additional fuzzy layer in the Deep Belief Network to generate control signals.
Figure 4 illustrates the Deep Belief Network (DBN) with the fuzzy controller.As areas of focus in data collection, fuzzy logic allows values of system variables to be described using language phrases like high, low, and middle.In Deep Belief Network (DBN) with the fuzzy controller, the input layer receives the rotating reference frame current signal, which is termed the input current.After that, the hidden layers process the reference frame current.And the fuzzy logic layer then applies fuzzy rules to control the peak point values by changing the duty ratio to vary the operating point at the time of maximum power output.These fuzzy rules provide the most efficient operating point and a steady output of current on the grid side.The output layer receives control signals from the fuzzy layer to operate the bidirectional converter by the judgment made by the fuzzy logic layer.Hence these control provided by the Deep Belief Network (DBN) with fuzzy controller regulates to allow power transmission across subgrids depending on the frequency of AC bus and DC bus voltage.
In the DC side of the hybrid AC/DC microgrid, Variable Voltage Stabilization Compensator (VVSC) Multiloop controller is employed to enhance the efficiency of DC bus stabilization which regulates the minimization of voltage, current transients, and the non-linear characteristics through the reactor arm which is connected to a stationary parallel capacitor bank in the Multi-loop controller.According to the task ratio and error value that cause the switch to change, the controller performance of the Variable Voltage Stabilization Compensator (VVSC) Multiloop controller compares the DC bus voltage with the reference value and power error in a Pulse With Modulation (PWM).Compared to parallel FACTs devices, this system is more cost-effective and can also be used as a two-pulse low-voltage switch (a diode rectifier).
Figure 5 depicts the VVSC Multi-loop controller structure.In this process, the DC and voltage are processed to their RMS values.Then it is given to a voltage stabilization loop to provide stabilized voltage at the DC side with linear load power characteristics.The resultant global error E G from the voltage, the stabilization loop  is the input to the proportional integral derivative (PID) controller which is the summation of three error values.Then, the Pulse With Modulation (PWM) is used to regulate dynamic voltage by changing the switch based on error.
The global error E G is a three-error PID control input that self-adjusts.The voltage stabilization loop error (E VDC1 ), the current limiting error (E IDC1 ), and the dynamic power loop error (E PDC1 ) are the three issues in the global error.The global error E G is given in Equation ( 3); where γ VDC1 is the voltage stabilization loop, γ IDC1 is limiting the current, γ PDC1 is the power loop. (5) For the PID controller, a time domain adjusts the Pulse With Modulation (PWM) control signal of the VVSC Multiloop controller which is given by the equation, Thus on the DC side, the VVSC Multiloop controller uses the Pulse With Modulation (PWM) to compare the DC bus voltage to the reference value and power error in the Pulse With Modulation (PWM), which changes the switch based on the task ratio and error, reducing the transient voltage state, regulating non-linear characteristics, and stabilizing the dynamic voltage.
Overall the hybrid AC/DC control techniques with improved Harmonic conditions using Deep Belief Network (DBN) based fuzzy controller and compensator modules provide the synchronization by utilizing SRF-PLL with Self-curing decoupling network.And also, to eliminate the phase-detection complexity and by connecting STAF and VVSC Multi-loop controllers on the AC and DC side improves the power quality and regulates the non-linear characteristics and dynamic voltage.Also, the AC and DC bus voltages are regulated by using DBN based fuzzy logic controller which provides the control signal to control the bidirectional converter.The next section explains the result obtained from the hybrid AC/DC control techniques with improved Harmonic conditions using DBN-based fuzzy controller and compensator modules and discusses it in detail.

Result and discussion
This section provides a detailed description of the implementation results as well as the performance of the proposed system and a comparison section to ensure that the proposed system performs valuable.The simulated output of the proposed system is depicted in Figure 6.The process flow of the simulated output begins with synchronizing the grid voltage using SRF-PLL with Self-curing decoupling network module to eliminate spurious frequency transients and difficulty to determine the phase of voltage signals by determining the negative and positive sequence in separate blocks.Then, the STAF Multiloop controller module is included in the ac side with tuned circuits to eliminate higher-order harmonics above or below the tuned values.In the hybrid microgrid, to convert AC to DC voltage, a bidirectional interlinking converter is present which is controlled by DBN based fuzzy controller that provides control signals to the converter based on the decision taken through the fuzzy logic layer.Finally, on the DC side, the VVSC Multiloop controller module is placed which regulates voltage through the Pulse With Modulation (PWM).
The voltage of the microgrid is shown in Figure 7.The voltage of the grid lies in the range of −2-2 volts and the time taken ranges from 0.001 to 0.06.The phase and  frequency of the grid voltage are synchronized by using SRF-PLL with Self-curing decoupling network by eliminating the influence of the negative sequence component over the positive sequence component.
The current at the grid is shown in Figure 8.The current value ranges from −25 to 25 amperes under the time limit of 0.001 to 0.06 s.A Switched Tuned Arm Filter (STAF) Multi-loop Controller, which passes signals using tuned circuits, is placed on the ac microgrid.This process eliminates higher-order harmonics and decreases the amplitude of current harmonics in the ac microgrid at a frequency above or below the tuned values by making use of tuned circuits in the controller.
Figure 9 shows the voltage on the DC side.The voltage ranges from a minimum of 320 volts at 0.06 volts to a maximum of 750 volts at 0.001 s.By comparing the DC bus voltage with the reference value and power error using a Pulse With Modulation (PWM) module, the Variable Voltage Stabilization Compensator (VVSC) Multiloop controller checks the DC bus voltage with the reference value and regulates the dynamic voltage.
The output voltage obtained from the hybrid ac/dc microgrid is shown in Figure 10.The voltage attains a maximum value of 350 volts when the time is 0.05 s and   Table 1 shows a comparison of the proposed techniques' root mean square error (RMSE) with existing strategies such as Deep Belief Networks with a Fuzzy Neural Network (DBFNN) (Wang et al., 2019), Temperature based Deep Belief Networks (TL-GDBN) (Xing et al., 2021), Deep Belief Echo State Network (DBESN) (Zhang et al., 2021), Spiking Convolutional Neural Network (SCNN) (Zhou et al., 2020), and Growing Echo State Network (GESN) (Li & Li, 2019).According to Table 1,

Conclusion
In this research, hybrid AC/DC control techniques with improved Harmonic conditions using a DBN-based fuzzy controller and compensator modules have been proposed to remove the spurious frequency transients and complexity in determining phase.This proposed technique achieves stabilization of dynamic voltage and regulates the power transmission across the hybrid ac/dc microgrid using SRF-PLL with Self-curing decoupling.Again, introducing, the STAF and VVSC Multi-loop controller including central DBN with Fuzzy controller, eliminates the influence of the negative sequence component on the positive sequence component and suppresses the higher order harmonics at a frequency above or below-tuned values.The proposed system achieves a stabilized voltage of 350 volts at 0.05 s.The performance of the proposed hybrid AC/DC control techniques with improved harmonic conditions using DBN based fuzzy controller and compensator modules outperforms the existing techniques with a low RMSE of 0.1 and low execution time of 10 s.

Figure 1 .
Figure 1.Block diagram of the proposed system.

Figure 6 .
Figure 6.Simulated output of the proposed system.

Figure 7 .
Figure 7. Voltage of the grid.

Figure 8 .
Figure 8.Current of the grid.

Figure 9 .
Figure 9. Voltage at the DC side.
× 10 −3 SCNN 1.7 × 10 −3 GESN 1.2 × 10 −3 Proposed 0.1 × 10 −3 attains a minimum value of 230 v when the time is 0.001 s.The stabilized output voltage is achieved by using a central Deep Belief Network (DBN) with a fuzzy controller.This regulates the bidirectional interlinking AC/DC converter by utilizing the fuzzy logic layer which controls the peak point values and the output layer gives the control signal to the converter, as a result, based on the frequency of AC and DC bus voltage, it regulates to allow power transmission across subgrids.
the root means square error (RMSE) of the proposed model is minimum, whereas the root means square error (RMSE) of the Temperature-based Deep Belief Network (TDBN) is high.The proposed model's root means square error (RMSE) is reduced by adopting a Variable Voltage Stabilization Compensator (VVSC) Multiloop controller which calculates the error and minimizes the error with the Pulse With Modulation (PWM).Table 2 shows a comparison of the proposed techniques' mean actual error (MAE) with existing strategies such as Deep Belief Networks with a Fuzzy Neural Network (DBFNN), Temperature based Deep Belief Networks (TL-GDBN), Deep Belief Echo State Networks (DBESN),

Figure 10 .
Figure 10.Output voltage obtained from the hybrid ac/dc microgrid.
reduced by adopting an SRF-PLL with a Self-curing decoupling network which reduces the complexity of determining the phase and frequency of signals since it does not require any additional load and external operation.

Table 2 .
Comparison of MAE.

Table 3 .
Comparison of execution time.Spiking Convolutional Neural Network (SCNN), and Growing Echo State Network (GESN).According to Table 2, the MAE of the proposed model is minimum, whereas the MAE of the Temperature-based Deep Belief Network (TL-GDBN) is high.The proposed model's MAE is reduced by Deep Belief Network (DBN) with fuzzy controller regulates to allow power transmission across subgrids depending on the frequency of AC bus and DC bus voltage.Table 3 depicts a comparison of the overall execution time of the proposed system with existing strategies such as a Deep Belief Network with a Fuzzy Neural Network (DBFNN), Temperature based Deep Belief Network (TL-GDBN), Deep Belief Echo State Network (DBESN), Spiking Convolutional Neural Network (SCNN), and Growing Echo State Network (GESN).From Table 3, the execution time of the proposed model is minimum, whereas the execution time of the Spiking Convolutional Neural Network (SCNN) is high.The proposed model's execution time is