A cascaded multilevel converter (CMC) with hybrid energy storage system (HESS) offers a promising solution for high-voltage and high-power hybrid dc-ac systems. However, asymmetric power distribution (APD) across different energy storage systems (ESSs) presents a challenge. This article proposes a dynamic power management strategy for CMC-based HESS, featuring dynamic power sharing to optimize power allocation between batteries and supercapacitors (SCs) based on their distinct response times. In addition, battery state-of-charge (SOC) balancing and SC energy management strategies are introduced to maintain optimal energy distribution and ensure long-term operation. The approach enhances overall system performance by fully leveraging the complementary strengths of batteries and SCs. The effectiveness of this strategy is validated through simulations and experiments, confirming its scalability and adaptability to various CMC-based HESS configurations.

Learning-based distribution system state estimation (DSSE) methods typically depend on sufficient fully labeled data to construct mapping functions. However, collecting historical labels (state variables) can be challenging and costly in practice, resulting in performance degradation for these methods. To fully leverage low-cost unlabeled historical measurement data, this article proposes a Gaussian process (GP)-regularized semi-supervised learning method for DSSE models, aiming at achieving feasible estimation precision using minimal state variable labels while also providing valuable interval estimation of state variables. Firstly, a structure-informed graphical encoder is established to generate appropriate node embeddings. A tailored GP-regularized learning method is then developed to model the intermediate latent space using these embeddings. It constructs the unlabeled embeddings by a weighted combination of labeled space vectors, with the weights determined by a kernel function, thereby forming additional supervision and regularizing the learning process of the network in the latent space. The regularized embeddings are then fed into a decoder to yield estimation outcomes. This procedure enables the proposed method to model intrinsic correlations across measurement data, as well as capture essential patterns related to DSSE even using extremely limited state labels. The trained DSSE models can thus adapt to the domain of new measurements. Lastly, the decoder outcomes and related latent embeddings are processed through a composite GP kernel to further derive the interval estimation of state variables, enabling uncertainty quantification. Experimental results demonstrate the effectiveness of the proposed method in handling extremely limited historical state labels and accurately quantifying the uncertainty of state variables.

The distributed control strategy has been widely employed, facilitating flexible and scalable coordination of different units in DC microgrids. However, the communication networks employed make them cyber-physical systems that are susceptible to cyberattacks. In this paper, a novel cybersecurity vulnerability identification and detection framework based on adversarial learning is developed for DC microgrids. Specifically, a multi-agent reinforcement learning (MARL) algorithm is used to emulate intelligent attackers that exploit system vulnerabilities by generating attacks capable of bypassing the existing detection mechanisms. In response, a data-driven attack detector is developed to complement the detection scheme, enhancing the overall detection capability. The framework utilizes an iterative adversarial learning process, wherein attacker and defender models continuously challenge and improve each other. This dynamic interaction enables the identification of a wider range of potential attacks, resulting in a more robust and adaptable detection mechanism for DC microgrids.

The training process of learning-based distribution system state estimation (DSSE) methods relies on accurate state variables, which typically contain unknown noise and outliers in practice. To this end, this paper proposes an adaptive noise-resistant graphical learning-based DSSE method considering the impact of inaccurate state variables. Specifically, two global-scanning graph jumping connection networks are first developed to capture the regression rules between measurements and state variables considering the structure constraints. To mitigate the negative impact caused by inaccurate labels, a collaborative learning framework is further developed, within which Gaussian mixture model-based discriminators are employed to adaptively select clean samples in each mini-batch. These allow the method to obtain robustness against noisy state labels in historical data, as well as anomalous measurements during online operations. Comparative tests show the superiority of the proposed method in tackling abnormal data in both the training and test phases.

Large-scale hydrogen electrolyzers for hard-to-abate industries, such as steel industry, have the potential to be an essential tool of demand response in low inertial power systems with high shares of renewable energies. Their flexibility comes from the possibility to store hydrogen, decoupling electric consumption from hydrogen demand. Therefore, they can help in the integration of more renewable energies by the provision of grid services, such as frequency regulation. Alkaline electrolyzers (AELs) are the most mature and cost effective technology for large-scale hydrogen applications. However, their slow dynamics do not allow a fast response. Therefore, their combination with energy storage systems (ESSs) into hybrid hydrogen systems (HHSs) enhances their flexibility and fast response for frequency regulation. Supercapacitors (SCs) are suitable ESS technology in this application due to the high power and low energy required. A decentralized dynamic power sharing control is proposed for an AEL/SC HHS to provide frequency regulation with scalability. The control strategy respects the slow dynamics of the AEL, while the use of the SC is optimized by the automatic recovery of the dc bus voltage and SC state of charge (SoC). The decentralized approach of the control strategy enables easy expansion of the system, essential for large-scale hydrogen systems. The effectiveness of the method in large-scale power systems, as well as its scalability is shown in simulation results. The control strategy is validated with experimental results.

DC microgrids (MGs) are cyber-physical systems (CPSs) prone to cyber attacks which could disrupt the normal operation of DC MGs. Accurate estimation of the attack vector is crucial to recover correct signals from compromised measurements for safe DC MG operation, while it has not been effectively achieved by existing methods and the accuracy is challenged by unmodeled uncertainties in practical power electronic converters. This paper proposes a digital twin (DT)-based cyber attack detection and mitigation scheme for DC MGs. First, the lightweight radial basis function neural network (RBFNN) is adopted to compensate for the mismatch between the ideal model and the real system for accurate converter modeling. Second, a composite descriptor observer-based local DT is designed to achieve accurate estimations of attack signals and correct observations of converter states. In addition, a global DT is developed at the system level to accurately estimate and eliminate cyber attacks in the secondary control. As a result, the proposed method can mitigate attacks by replacing the corrupted signals with estimated true values provided by DT, leading to accurate and stable operation of the system. Finally, simulation and experimental results are given to validate the effectiveness of the proposed method.

The distributed microgrids cooperate to accomplish economic and environmental objectives, which have a vital impact on maintaining the reliable and economic operation of power systems. Therefore a distributed multi-agent reinforcement learning (MARL) algorithm is put forward incorporating the actor-critic architecture, which learns multiple critics for subtasks and utilizes only information from neighbors to find dispatch strategy. Based on our proposed algorithm, multi-objective optimal dispatch problem of microgrids with continuous state changes and power values is dealt with. Meanwhile, the computation and communication resources requirements are greatly reduced and the privacy of each agent is protected in the process of information interaction. In addition, the convergence for the proposed algorithm is guaranteed with the adoption of linear function approximation. Simulation results validate the performance of the algorithm, demonstrating its effectiveness in achieving multi-objective optimal dispatch in microgrids.

The widespread adoption of electric vehicles (EVs) and hydrogen fuel cell electric vehicles (HVs) is tightening the interdependence between power and transportation systems, calling for better coordination between them. To address this challenge, this paper proposed a distributed coordination method for the hydrogen-integrated microgrids and transportation system. First, we introduce energy sharing among microgrids which reduces the overall system cost by 16.2% and analyze how it improves the traffic flow. Additionally, we develop bidding models for microgrids participating in joint energy and ancillary service markets, maximizing flexible resources utilization and increasing revenue by 147%. A mixed vehicle flow transportation system model is then established, including EVs, HVs, and gasoline vehicles. To coordinate the two individual systems efficiently, a distributed algorithm is proposed, incorporating a filtering mechanism that reduces the communication burden by 63% during the iterative process. Uncertainties and nonlinearities are handled using distributionally robust method and linearization techniques. Finally, case studies validate the effectiveness of the proposed method and highlight the mutual impact between the two systems.

Finite-set model predictive control (FS-MPC) appears to be a promising and effective control method for power electronic converters. Conventional FS-MPC suffers from the time-consuming process of weighting factor selection, which significantly impacts control performance. Another ongoing challenge of FS-MPC is its dependence on the prediction model for desirable control performance. To overcome the above issues, we propose to apply reinforcement learning (RL) to FS-MPC for power converters. The RL algorithm is first employed for the automatic weighting factor design of the FS-MPC, aiming to minimize the total harmonic distortion (THD) or reduce the average switching frequency. Furthermore, by formulating the incentive for the RL agent with the cost function of the predictive algorithm, the agent learns autonomously to find the optimal switching policy for the power converter by imitating the predictive controller without prior knowledge of the system model. Finally, a deployment framework that allows for experimental validation of the proposed RL-based methods on a practical FS-MPC regulated stand-alone converter configuration is presented. Two exemplary control objectives are demonstrated to show the effectiveness of the proposed RL-aided weighting factor tuning method. Moreover, the results show a good match between the model-free RL-based controller and the FS-MPC performance.

Deep reinforcement learning (DRL) is a promising solution for the coordination of power systems with high penetration of inverter-based renewable energy sources (RESs). Yet, when adopting the DRL-based control method, the safe and optimal operation of the system cannot be guaranteed at the same time, as the conventional DRL agent is not designed to solve the hard constraint problem. To address this challenge, a deep neural network (DNN) assisted projection-based DRL method for safe control of distribution grids is proposed in this section. First, a finite iteration projection algorithm is proposed to guarantee hard constraints by converting a non-convex optimization problem into a finite iteration problem. Next, a DNN-assisted projection method is proposed to accelerate the calculation of projection and achieve the practical implementation of hard constraints in the DRL problem. Finally, a DNN Projection embedded twin-delayed deep deterministic policy gradient (DPe-TD3) method is proposed to achieve optimal operation of distribution grids with guaranteed 100% safety of the distribution grid. The safety of the DRL training is guaranteed via the embedded Projection DNN in TD3 with participation in the gradient return process, which could smoothly and effectively project the DRL agent actions into the feasible area, thus guaranteeing the safety of data-driven control and the optimal operation at the same time. The case studies and comparisons are conducted in the IEEE 33 bus system to show the effectiveness of the proposed method.