This paper proposes an optimal power-sharing algorithm (OPSA) devised to regulate the power distribution among grid-forming inverters (GFMIs) in modern power systems powered by distributed energy resources (DERs). Unlike traditional power-sharing techniques that assume an infinite or constant power supply, OPSA continuously modifies virtual impedance based on resource availability, such as the battery capacity and state of charge (SOC). The supervisory controller integrating OPSA calculates the demanded power to adjust the virtual impedance, which is optimally distributed among the GFMIs, minimizing battery stress and promoting balanced discharge. Furthermore, the proposed approach improves grid resiliency by dynamically adapting to changing load variations and mitigating potential instabilities. Finally, the effectiveness of the OPSA is validated by the simulation results of multiple test scenarios across varying operating conditions.

The integration of solid-state transformers (SSTs) with grid-forming inverters (GFMIs) improves flexibility and control in modern microgrids while simultaneously introducing new cyber-physical vulnerabilities. The strong coupling between SSTs and GFMIs creates dynamics that result in the diffusion of localized cyberattacks throughout the microgrid. This paper provides a comprehensive impact analysis of four representative cyberattack scenarios: side-channel noise intrusion (SNI) compromising the integrity of SST measurement feedback, tampering with the voltage-regulation loop of the SST, a virtual governor frequency-restoration attack on GFMI, and cascaded tripping of breakers simulating insider access to protection coordination. The attacks are targeted towards specific control or protection layers and are evaluated through simulations on a modified IEEE 14-bus system. Results demonstrate that even brief attacks can result in prolonged instability, including frequency deviations, power oscillations, and voltage regulation failures, due to the low inertia and damping of power electronic-based systems. Furthermore, the coupling between the SST and the GFMIs results in aftereffects even well beyond the removal of the attack, degrading overall system. The findings underscore the importance of recognizing control-layer cyber threats in power electronic-dominated microgrids and motivate the need for more resilient control architectures.

Recently, incidents of the sub-synchronous oscillation (SSO) occurrence have significantly increased in the power system integrated with renewable energy resources. In the current power system, as more distributed renewable energy resources replace conventional synchronous generator-based sources, the grid is transformed into a low-inertia power system. The low inertia and distributed nature of renewable energy systems contribute to weak grid conditions. In the type-4 wind turbine generators (WTG) SSO can originate due to the interaction of fast dynamics of power converter’s control and weak AC grid. If SSO is not detected and mitigated in a timely manner, it can cause severe damage to the WTG’s shaft, turbine structure, and can pose severe type of instability in the power system that may lead to cascaded tripping. Therefore, this paper presents a machine learning (ML) based scheme for fast and accurate detection of SSO and localizing the WTG’s control parameter that triggers SSO under the influence of weak AC grid. The proposed SSO detection and localization scheme features a dual neural network structure (DNNS) based on long short-term memory (LSTM). The first NN structure detects SSO and triggers second NN structure when SSO is detected. The second multiclass NN identifies the control parameter of WTG that triggered this SSO and provides a recommendation/ reference for the SSO mitigation scheme. The effectiveness of the proposed ML based SSO detection and localization scheme is verified via confusion matrix and simulation that analyzes different cases related to the ML scheme validation.

Mass penetration of renewable energy sources (RES) has drastically triggered the integration of clusters of VSG-based grid-forming inverters (GFMIs) to enhance grid stability and facilitates the transition towards power electronics dominated grid (PEDG). Although high number of GFMIs provides a smoother frequency response due to higher emulated virtual inertia, considering the complex dynamic of the PEDG, estimating the emulated virtual inertia by the primary control layer of the GFMIs introduces extensive mathematical modeling and computational burden. To address the shortcomings of conventional model-based approaches, this paper proposes a data-driven methodology based on convolution neural network (CNN) to estimate effective virtual inertia by analyzing the explanatory frequency response of a cluster of GFMIs. The neural network (NN) is trained offline and evaluated on a sample cluster of 100% GFMIs through multiple case studies to validate the effectiveness of the proposed methodology. The application of the proposed data-driven tool can further facilitate network analysis by considering the corresponding aggregated model to enhance the stability analysis of the PEDG.

Virtual impedance-based power sharing between several interconnected grid-forming inverters (GFMI)s in a power electronics-dominated grid (PEDG) does not consider the available power reserves. This non-optimal power-sharing may result in the overloading of nearby inverter-based resources (IBR)s exhibiting lower effective line impedances in the network. This paper proposes an optimal and controllable power-sharing scheme based on a modular artificial neural network (ANN) that can consider various factors like power reserve, line losses, incentives from the grid, etc., and increases the system's scalability. An optimum power allocator (OPA) mechanism that monitors the output active powers and line losses of all the IBRs is proposed to address the issue of uncontrolled power sharing. The OPA dynamically adjusts power allocation weights using linear programming to optimize the overall loss minimization cost function. Moreover, an artificial neural network-based virtual impedance optimizer (ANN-VIO) is also proposed, which estimates the corresponding virtual impedance (VI) value for each GFMI unit in the PEDG. The primary controller of GFMI uses these estimated VI values to adjust its point of common coupling (PCC) voltage reference and attain optimum proportionate power sharing within the overall PEDG. Multiple case studies are performed to validate the effectiveness of the proposed GFMIs power-sharing scheme.

Power electronics dominated grid (PEDG) consisting of multiple inter-connected distributed generation (DG) units are getting more popular because of their benefits in reducing the stress on main transmission lines, limiting line losses, and improving the efficiency and stability of the power system. Often a PEDG is comprised of several energy sources such as solar, wind, and energy storage devices. For that reason, optimal power sharing between all the sources is crucial to enhance the reliability and resiliency of the network. Hence, this paper provides comprehensive insights into the operation and working principles, limitations, and challenges of the power-sharing schemes employed in PEDG. More specifically, control laws of the droop, virtual synchronous generator (VSG), and Lienard and Consensus-based oscillators are briefly overviewed. In addition, the inherent drawbacks and challenges of existing power-sharing schemes are explained for inductive, resistive, and complex line impedances. Moreover, their performance under dynamic variations of the source availability, inverter power reserve, line losses, and generation costs are discussed in detail.

The power grid is rapidly changing into the Modern Power System due to the increased penetration of power electronics. As more distributed energy resources are added, detecting anomalies and faults in the grid is increasingly important. This works proposes an artificial intelligence system built upon deep learning approaches to detect line-line faults in the modern power system. This two-layer approach uses both long short-term memory along with graph neural networks to classify and localize all line-line faults in a fourteen-bus grid. This approach creates a framework which can learn the features from three-phase data and utilize them for the detection of anomalies across the modern power system.

Grid-forming inverters (GFMIs) are promising solutions for voltage and frequency support in upcoming power electronics-dominated grids (PEDG). However, current state-of-the-art decentralized control schemes for GFMIs are designed for operation under normal conditions. These decentralized control schemes could result in an adverse dynamic response if a cluster of GFMIs network disconnects from the rest of the grid due to a fault. The dynamic response of GFMIs could be improved with coordinated control schemes which require communication and make the system more complex. This paper proposes a decentralized Artificial Intelligence (AI)-based method for online fault detection and localization to enhance the dynamic response of GFMIs without using a communication layer. Each GFMI will detect and localize the line tripping based on its output active and reactive power measurement. Then, according to the topology of the grid after line tripping, each GFMI updates its nominal power and inertia to suppress the frequency transient caused by the line tripping. The inertia and damping factor of the GFMIs are then recalculated to ensure optimal operation of the network of inverters after line tripping. Several case studies are presented to validate the effectiveness of the proposed method in the timely detection, localization, and mitigation of adverse frequency transient after line tripping in a communication-less manner.

Speed sensor faults are very common in electric vehicle (EV) applications, often disrupting system performance due to the reliance on speed sensor data by the electric drives. Consequently, fault-tolerant control approaches require robust observer structures capable of functioning effectively during faults, swiftly identifying and reconfiguring failures. Thus, this paper presents a novel fault-tolerant control design leveraging a synergetic observer (SO) to substitute the speed sensor in the event of fault. Comparative analysis with a super-twisting sliding mode observer (ST-SMO) highlights the proposed observer's superior performance in terms of tracking accuracy and ripple reduction. The presented simulation results demonstrate promising performance of the proposed observer for fault-tolerant control in EV applications.

This paper proposes a resilient control scheme for a Solid-State Power Transformer (SST) based power substation (SSPS). The proposed control scheme consists of a centrally located secondary aggregator and primary controller that govern the operation of corresponding SST cells. A cross-correlation-based Intrusion Detection (CC-IDS) mechanism estimates the cross-correlation coefficient between the global voltage feedback and a reference signal. As the estimated cross-correlation drops below the preset threshold, an intrusion on the SST’s global voltage feedback is detected, and a Convolutional Neural Network (CNN) enabled attack mitigation mechanism is triggered. The CNN estimator utilizes the SSPS’s attacked global input voltage feedback to predict the desired global input voltage. Estimated global input reference is then substituted with noisy global input voltage feedback in the SST cell’s primary controllers to mitigate the impacts of the intrusion. To ensure high prediction accuracy, the CNN estimator is trained using datasets encompassing various side-channel noise intrusion (SNI) attack scenarios. Several scenarios are emulated in MATLAB Simulink environment to validate the CNN intrusion mitigation approach under the influence of different attack vectors.

Partitioning the power system into smaller, manageable units facilitates better grid monitoring and control, thereby improving the grid’s stability and reliability. However, large-scale power networks consist of thousands of nodes and edges, which complicates the process of learning appropriate node embeddings and aggregating information from neighboring nodes. By representing power grids as undirected weighted graphs, this study proposes a novel power network partitioning approach using Graph Neural Networks (GNN). The proposed model simplifies the clustering objective by focusing on a single balancing term, which reduces computational complexity while maintaining competitive clustering performance. The power network is represented as a graph where the proposed GNN uses the normalized graph Laplacian, which effectively captures the complex connectivity of the nodes, instead of the traditional adjacency matrix. Active power levels serve as nodal attributes, ensuring that clusters represent both the physical and operational characteristics of the network. This dual-focused approach promotes a partitioning that is topologically coherent and functionally homogeneous, vital for enhanced grid management. When applied to the IEEE 14, 39, and 118 bus systems, the proposed method has successfully delineated coherent clusters of buses, underlining its potential for improving power grid management. The simulation results confirm the method’s efficacy and applicability.

Due to their better performance in maintaining the frequency and voltage stability, grid forming inverters (GFMIs) are promising solutions for the future grid with high penetration of inverter-based resources. Among the GFMIs’ control method virtual synchronous generator (VSG) gained attention because of its closeness to the behavior of synchronous generator. However, limiting the output current of the VSGs during faults while ensuring stable operating condition is challenging especially when multiple GFMIs are connected to the grid. Although the virtual impedance can be designed to decrease the value of the current, the design procedure is performed based on the presumed operating condition which may change due to location and severity of the fault. This paper proposes, a method based on the dual control of input power and virtual impedance during fault which will ensure stable operation of VSGs during fault. According to the grid voltage and the requirements of fault current limiting for VSGs, the virtual impedance is calculated such that the VSG can tolerate fault current. Then, the input power of the VSGs is controlled to reduce the accelerating energy caused by the sudden decrease of grid voltage and output voltage of the VSG. The simulation results are presented to validate the impact of the proposed scheme on maintaining the stability of the VSGs during fault and suppressing fault current.

The increasing use of renewable energy sources to supply electrical power to the future grid causes the grid to be more prone to changes in frequency and voltage due to load or generation disturbances. Utilizing the electric vehicles inside the parking lots as an energy source is an economical solution to support grids’ frequency and voltage with an ever-increasing number of electric vehicles (EVs) in the upcoming years. However, integrating these parking lots as EV charging stations into the grid to consider both the grid-side requirement and the situation of the EVs inside the parking lot is challenging. This paper proposes an algorithm to integrate the EV parking lots with the grid using a grid-forming inverter to consider both the power capability of the grid and other generation sources as well as the status of the EV inside the parking lot. Based on the departures or arrivals in the parking lot and the available SOC of the parking lot as well as the data received from the other generation sources of the grid, the proposed algorithm regulates the power injected to the grid by the grid-forming inverter interacting with DC charging station formed by the parking lot. The simulation result shows the ability of the proposed algorithm to regulate the power drawn from EVs according to the capacity of other generation resources and the available SOC of the EV parking lot, thus enabling an enhanced grid interaction of the EV parking lots as an auxiliary source for grid services.

The increase in the penetration of renewable energy sources makes the future grid more vulnerable to frequency and voltage fluctuations. Electric vehicle (EV) charging stations can be used as a source of energy to support the grids’ voltage and frequency during load or generation disturbances. However, managing the contribution of each EV in the total energy injected into the grid, considering each EV capacity while maintaining minimal losses, is challenging. This paper presents an algorithm to distribute the total injected energy among the EVs based on their state of charge, which determines the optimal power injected by each EV inside the parking lot. Furthermore, the design procedure of the CLLC converter for the EV charging stations is elaborated to meet the requirements of each converter and DC charging station for grid integration. The control of the CLLC converter for both grid-to-vehicle (G2V) and vehicle-to-grid (V2G) conditions is described. The simulation results for different scenarios for EV arrival and departure show the effectiveness of the proposed algorithm in minimizing the loss according to the SOC.

Stealthy cyberattacks pose a significant threat to modern power systems by exploiting advanced techniques to manipulate system behavior while avoiding detection by traditional security measures. In this study, we focus on the impact of Deep Reinforcement Learning (DRL) based attackers on a sample microgrid and develop robust detectors to mitigate these threats. Leveraging an iterative training process, we enhance the capabilities of successive attackers and detectors, resulting in improved system security. Our experiments demonstrate that DRL-based attackers can effectively disrupt system operations, highlighting the importance of robust detection mechanisms. Subsequently, we develop robust detection mechanisms, making new attacker attempts unsuccessful. We show that detectors developed through our mechanism are more effective in mitigating system impact and quickly identifying anomalies.

This work presents a novel approach to tackle the problem of real-time detection of cyber intrusions at power electronics dominated grids (PEDG) inverters by establishing an analogy between electrical and medical fields. The methodology leverages the significant similarities between electric grid and human body and posits the feasibility of utilizing the large body of knowledge available in medicine, particularly in the area of disease diagnosis, for the benefit of detecting intrusions occurring to the grid, considering the tools introduced by tremendous advances in artificial intelligence (AI). The theoretical foundation of the analogy is discussed and as a first validation step, the grid inverter voltage is classified into three types of normal, disturbed and anomalous by associating it with three classes of heart rate electrocardiogram (ECG) signals of normal sinus rhythm, mild arrhythmia (supraventricular), and severe arrhythmia (ventricular), respectively. This is done via AI-based time series analyzers which establish a mapping between the two signal categories and enable classifying grid voltage based on ECG using the trained analyzer block. Achieved results demonstrate validity of the approach and promise addressing one of the main challenges in developing intrusion detection systems, that is, data scarcity, by tapping at the enormous knowledge base in the medical field.

This paper presents a Bayesian regularization-based artificial neural network (BRANN) for power setpoint correction of grid-following (GFL) inverter with compromised supervisory controller layer. Coordinated control schemes with multi-layered communication infrastructure makes the upcoming power electronics dominated grid (PEDG) susceptible to cyber threats. In particular, malicious power setpoints introduced by a compromised secondary controller can jeopardize the normal operation of PEDG. The main goal of this study is to develop an AI-scheme for the primary control layer of GFL inverter to correct the manipulated power setpoints in real-time independent of the compromised secondary control. A neural network (NN) module is trained via Bayesian regularization algorithm to address the challenges of AI-based schemes deficiencies. The provided case studies validate the performance and effectiveness of the proposed approach.

This paper presents an efficient and optimal model predictive control (MPC) scheme for cascaded multilevel inverters (CMI) interfaced with battery sources. One of the major control challenges with CMI is the equal power distribution amongst the cascaded cells. For the application in hand, the unbalance state-of-charge of battery cells due to unequal power drawn from them, will impact the battery lifetime and eventually the reliable operation of the overall system due to uneven stress on CMI cells. The existing classical control schemes for CMI with power balancing capability are suffering from slow dynamic response for power balancing. In addition, they require substantial tuning effort due to their multi-nested loop control structure. On the other hand, conventional MPC schemes demonstrates promising superiority for CMI control comparing to classical control schemes, but they suffer from uneven power distribution among cascaded cells. This paper proposes an optimal moving horizon predictive control scheme for CMI that addresses the challenges associated with classical control and conventional MPC schemes for battery sources interfaced grid interactive CMI. The proposed approach addresses the computational burden of MPC for CMI as number of level increases while ensure equal power drawn from battery cells with fast dynamic response which improves the lifetime of entire system by distributing the stresses on all cells of CMI. The functionality of the proposed approach is evaluated with several case studies. The complete experimental results will be included in the final paper.

Microgrids in power electronics dominated grid (PEDG) should be able to operate under islanded and grid-connected modes and switch between them seamlessly. During the islanding events severe fluctuations in voltage and frequency of the microgrid may arise that can cause unstable operation and ultimately blackout. Moreover, under this transition the inverter may lose synchronism or point of synchronization because of disconnection from grid. This paper proposes: (i) a predictive control scheme that enables dual-mode (grid-forming and grid-following) operation of inverters and enables seamless transition between these two modes; and (ii) a swift single-point synchronization scheme for the network of grid-following and grid-forming inverters. The proposed control scheme mitigates voltage, current, and frequency fluctuations and enables seamless transition while needing limited time for synchronization between the grid-connected and the islanded modes of operation. Moreover, the predictive controller ensures smooth toggling between grid-following and grid-forming modes of operation and extracting maximum power from photovoltaic energy sources. The effectiveness of the proposed control scheme with its self-synchronization capability is validated by various case studies under grid-tied, islanded and transition from grid-connected to islanded mode of operations.

Decentralized hierarchical control techniques do not require a communication layer in their secondary control level. These control techniques reduce the vulnerability of microgrids (MGs) to cyber-attacks, reduce data losses, lessen time delays, and mitigate costs for communication infrastructure. In the islanded AC MG, droop control is profusely used at the primary level to achieve accurate power-sharing. However, it may result in steady-state voltage and frequency (V/f) deviations with varying load conditions. The secondary control layer is required to restore these deviations while maintaining droop dictated power levels. The secondary layer supports V/f control at a slower time scale, while the primary droop-based layer supports V/f tracking at a faster time scale. This paper presents a review of the decentralized secondary control systems that interact with droop based primary controllers. The V/f restoration schemes by the secondary control layer are categorized as linear and non-linear technologies. The mathematical formulations for each control scheme are presented, and a tabulated summary of all the control systems. Finally, simulations are conducted for a selected control technique from each category, and the results are presented and discussed.

In this work, a novel LCL filter topology for non-isolated Photovoltaic (PV) applications is developed. This topology has the ability to trap the High-Frequency (HF) Common Mode (CM) current inside the PV inverter. Consequently, suppressing the ground leakage current. Furthermore, the proposed solution is immune to ground leakage current resonance issues (i.e. effective for applications where the utility grid is characterized as a weak grid). Moreover, the theoretical analyses were validated on a 10 kW grid-connected PV system. The results demonstrated that at resonance conditions, the proposed system reduced the leakage current root-mean-square (RMS) value from 1 A to 25 mA. Thus, satisfying the VDE standard's leakage current limit.

Low inertia power generation units make islanded microgrids and nanogrids more vulnerable to voltage and power fluctuations. Smart loads are a possible solution to suppress voltage and power fluctuations in islanded nanogrids. Since smart loads utilize inverters with short-time responses, their dynamics would have a considerable effect on the dynamics and stability of nanogrids, Therefore, the dynamics of smart loads play a significant role for stability analysis of these systems. This paper analyzes the dynamic behaviors of smart loads in nanogrids. Furthermore, a state-space model is developed for smart loads. The stability of smart loads is studied using the developed model, along with circuit simulations. The case study simulations are provided to verify the performance of utilizing smart loads in mitigating voltage and power fluctuations in islanded nanogrids.

Due to variability of solar energy resources, maximum power point tracking (MPPT) of photovoltaic (PV) is required to ensure continuous operation at the maximum power point (MPP) and maximize the energy harvest. Many standards are developed to ensure the safe and efficient power generation under dynamic weather conditions. This paper presents a high efficiency fixed-step model predictive control (MPC) technique to employ the MPPT for photovoltaic applications. The MPP operating point is determined by using perturb and observe (P&O) technique. The proposed fixed-step predictive model based MPPT presents significant advantages in dynamic response and power ripple at steady state. A characteristic of MPC is the use of system models for selecting optimal actuations, thus evaluating the effect of model parameter mismatch on control effectiveness is of interest. In this paper, the load model is eliminated from the proposed MPC formulation by using an observer-based technique. The performance of the proposed observer-based MPC-MPPT is evaluated on the basis of European Efficiency Test, EN 50530 that assesses the performance of PV systems under dynamic environment conditions. The proposed MPC-MPPT technique for a flyback converter is implemented using dSPACE DS1007.

This paper presents a reactive power compensation technique using model predictive control (MPC) of a matrix converter. This technique compensates lagging power factor loads using inductive energy storage elements instead of electrolytic capacitors (e-caps). Although ubiquitous in power electronic converters, e-caps have well-known failure modes and wear-out mechanisms. Therefore, the capacitors used to store energy in a voltage-sourced inverter (VSI) reactive power compensator require continuous monitoring and periodic replacement, both of which significantly increase the cost of the traditional load compensation technique. MPC of the matrix converter provides reactive power compensation by controlling the input reactive power and the output current to the inductive storage elements. Thus, compared to VSI techniques, the proposed reactive power compensation technique is more reliable and has a longer expected service life that is not limited by failure and wear-out modes of capacitors.