Abstract:This paper tries to summarize the attempts to apply artificial intelligence (AI) to power systems, particularly power system planning and operations which require significant computer analysis. Although the term AI was coined earlier, this paper considers the beginning to be in the 1980s when the first expert systems were applied to power engineering. Of course, many of the analytical techniques applied can be traced to earlier statistical analysis and pattern recognition. The concept of expert systems was very much in line with the concept of AI. The various methods for applying AI to power systems are traced here. The historical journey in this paper closes with the great explosion of AI applications in the last decade when almost all power system analysis is trying to utilize AI techniques to help the transformation of the power system into a more efficient and carbon-free system. This proliferation of research in the application of AI is covered in the other papers in this series.

Abstract:Recent progress in artificial intelligence (AI) is powered by three key elements: algorithmic innovations, specialized chips and hardware, and a rich ecosystem of software and data toolboxes. This paper provides an analysis of these three key elements, tracing the evolution of AI from symbolic systems and small, labeled benchmarks to today’s large-scale, generative, and agentic models trained on web-scale corpora. We review the hardware trajectory from central processing units (CPUs) to graphics processing units (GPUs), tensor processing units (TPUs), and custom accelerators, and show how the co-design of chips and models has unlocked improvements in throughput and cost by orders of magnitude. On the algorithmic side, we cover the deep learning revolution, scaling laws, pretraining and fine-tuning paradigms, and multimodal and agentic architectures. We map the modern software stacks, i.e., open-source AI frameworks, end-to-end toolchains, and community datasets, that make model development reproducible and widely accessible. Given the environmental and infrastructural impact of scale, we emphasize the trade-offs in energy, datacenter, and governance. Finally, we identify emerging trends that reshape how AI is developed and deployed.

Abstract:This paper provides an overview of the application potential of artificial intelligence (AI) in power systems and points towards prospective developments in the fields of AI that are promised to play a transformative role in the evolution of power systems. Among the basic requirements, also imposed by regulation in some places, are trustworthiness and interpretability. Large language models, foundation models, as well as neuro-symbolic and compound AI models, appear to be the most promising emerging AI paradigms. Finally, the trajectories along which the future of AI in power systems might evolve are discussed, and conclusions are drawn.

Abstract:Large language models (LLMs) have achieved remarkable progress in recent years. Nevertheless, the prevailing centralized paradigm for training generative artificial intelligence (AI) is increasingly approaching its structural limits. First, the concentration of large-scale graphics processing unit (GPU) clusters restricts the access to the pre-training stage, confining the fundamental model development to a small number of resource-rich institutions. Second, the economic and energy costs associated with operating massive data centers render this paradigm progressively less sustainable. Third, the hardware gatekeeping narrows the participation to computer science specialists, limiting the involvement of domain experts who are essential for high-impact applications. Finally, small- and medium-sized enterprises remain dependent on expensive application programming interface (APIs) or shallow fine-tuning methods that are insufficient to modify the core knowledge of a model. Together, these constraints impede innovation and hinder equitable access to next-generation AI systems. Model fusion offers a scalable alternative by integrating multiple specialized models without retraining from scratch. This paper analyzes the current landscape of model fusion, outlining the strengths and limitations of existing methods and discussing future directions. We highlight recent advances such as InfiFusion, InfiGFusion, and InfiFPO, which improve the alignment and scalability through techniques like top-K logit selection, graph-based distillation, and preference optimization. These techniques demonstrate substantial efficiency and reasoning gains, pointing toward a more accessible and resource-aware paradigm for large-scale model development. Finally, we discuss the practical applicability of model fusion, using the energy domain as an illustrative example.

Abstract:An electricity market is a complex, dynamically operated network encompassing multiple participants under defined rules, thereby ensuring real-time supply-demand balance and system reliability. However, the inherent complexity and dynamism of the electricity market pose significant challenges to conventional modelling approaches, which often rely on expert knowledge and manual processes informed by market regulations. This reliance frequently leads to inefficiencies and elevated risks of error. To address these limitations, this paper proposes a framework for automated electricity market modelling and simulation centered on a large language model based agent, termed the modelling and simulation system agent (MSS-Agent) framework. The proposed MSS-Agent framework employs the hierarchical chain-of-thought (HCoT) method to more accurately extract essential information from relevant documents, thereby enhancing modelling fidelity. Moreover, it integrates tool usage and reflexive debugging to optimize the code generation process, ensuring reliability in automated electricity market modelling and simulation. Experimental results demonstrate that the proposed MSS-Agent framework significantly improves both mathematical model extraction accuracy and code execution reliability. Consequently, the proposed MSS-Agent framework not only increases simulation efficiency but also provides more precise and dependable tools for informed decision-making in electricity markets.

Abstract:The modern power systems face challenges, including high proportions of uncertain renewable energy, rapid dynamics of power electronics, and decentralized control among multiple entities. Digital development has enabled power grids to integrate numerous edge devices equipped with sensing and computing capabilities, aiming to exploit edge data to enhance grid observability, controllability, and resilience. However, much of potential value of edge data remains unexploited with traditional architecture and methods. Therefore, we explore the potential of leveraging large language models (LLMs) to fully exploit edge data in modern power systems. An intelligent, scalable, and efficient three-layer architecture is proposed to align the capabilities of LLMs with the constraints of edge scenarios. Supporting technologies are reviewed for each layer, including multimodal data fusion, lightweight collaborative inference, and closed-loop control. To validate the proposed architecture, we provide three representative scenarios for preliminary exploration: virtual power plant (VPP) dispatch, intelligent substation inspection, and contingency management, illustrating how LLMs can unlock the value of edge data. We conclude by identifying key technical challenges and outlining future research directions for building modern power systems by LLM-based exploitation of edge data.

Abstract:In Part I of this paper, we have proposed the new concept of generalized voltage damping (GVD) and derived the system-wise GVD (sGVD) index for the global assessment of voltage stability and system strength. Part II of this paper extends this concept to develop a port-wise index for quantifying the voltage damping characteristics locally. To this end, we decompose the sGVD index into individual ports (or buses), thereby forming the port-wise GVD (pGVD) index, which can be computed using local measurements. By inheriting the interpretation of the system-wise index, we further prove that the average of pGVD indices across all ports is approximately identical to the sGVD index. Moreover, it exhibits favorable properties absent in existing indices based on the maximum Lyapunov exponents (MLEs) of terminal voltages, empowering its application as an assessment metric for the supportive capability of devices to short-term voltage stability. The model-independent feature enables the assessment considering the complex and nonlinear dynamics of inverter-based resources (IBRs) such as wind turbines, photovoltaics (PVs), and battery energy storages. Experimental simulations conducted on a heterogeneous IEEE 39-bus system and two practical power systems with massive renewable resource integration confirm the theoretical results. The influence of voltage control strategies of IBR, control parameters, integration locations, and active power control parameters are also analyzed, providing a new perspective for understanding the individual support of devices for short-term voltage stability.

Abstract:The existing minimum demand inertia (MDI) assessment methods based on time-domain simulation of system frequency response are complex in modeling and time-consuming in computation. If incorporating the load-side resources, it will lead to further computation inefficiency. This paper proposes a fast assessment method (FAM) for MDI in power systems. A full-response analytical model (FRAM) of a multi-resource system considering the load-side inertia is developed. The analytical expression of the mapping relationship between the maximum frequency deviation and system inertia is derived, thus realizing the fast solution of the system MDI under frequency security constraints. Case studies based on the modified IEEE RTS-79 test system and a provincial power grid in China demonstrate that the proposed FAM can solve the MDI in milliseconds without being affected by the system scale while maintaining high accuracy. This can provide an accurate and rapid analytical tool for sensing inertia security boundary in grid inertia resource planning and operation scheduling.

Abstract:Under weak grid conditions, oscillation in wind power integrated system occurs frequently. However, existing oscillation suppression methods face challenges in effectively coordinating control parameters and fail to guarantee excellent dynamic performance. Therefore, this paper proposes an uncertainty and disturbance estimator-based feedback linearization sliding mode control (UDE-FLSMC) method, which can reduce the negative damping region of the impedance phase of wind power integrated system. Firstly, the feedback linearization process of multi-input multi-output (MIMO) systems is derived. Then, the uncertainty and disturbance estimator (UDE) is used to estimate the disturbance in sliding mode control, and the UDE-FLSMC method is proposed. Secondly, the control structure and impedance model of wind power grid-side converter (GSC) are established. The impact of control parameters on the impedance characteristics of the converter is analyzed. It is demonstrated that the impedance phase in sub/supersynchronous frequency band maintains within a significant positive damping region under different operating conditions. Then, a hardware-in-loop experimental platform is constructed to verify the dynamic performance of the proposed UDE-FLSMC method, which is compared with proportional integral (PI) control and phase margin frequency division compensation (PM-FDC) control. The results show that the proposed UDE-FLSMC method exhibits superior oscillation suppression ability and faster response characteristics, which can significantly improve the stability of wind power integrated system under weak grid conditions.

Abstract:The massive integration of intermittent renewable generation and the increasing variability of demand raise concerns about the high level of uncertainty in the security assessment of power systems. In this context, the main contribution of this study is the proposal of a new definition of contingency criticality, which is based on the violation probability of the voltage security margin (VSM) while considering correlated uncertainties in both system loads and wind power generation. From this new definition, a contingency ranking can be derived and used to determine preventive control actions. To calculate this probability for each contingency, a new approach based on the cross-entropy (CE) method is developed and applied. The CE method is well-suited to handle high levels of uncertainty, as it typically provides faster and more accurate results compared with Monte Carlo simulation, particularly for cases with low violation probabilities of the VSM. Another innovative feature of this approach is the consideration of correlated uncertainties through the use of multivariate normal distributions and Gaussian copulas. Furthermore, the proposed definition is implemented using a formulation that is capable of detecting either saddle-node or limit-induced bifurcations to accurately identify the maximum loadability point. A proof of concept is presented for a comprehensive explanation of the proposed definition, followed by an application of this definition to the IEEE 118-bus test system. The findings of this paper highlight the need to carefully select critical contingencies for voltage security assessment in the context of increasing uncertainties.

Abstract:The uncertainty and variability of advancing wildfires present significant challenges to the resilience of power systems. This paper proposes a hierarchical dispatch strategy of multi-type virtual power plants (VPPs) for enhancing resilience of power systems under wildfires, which encompass geographically distributed VPPs (GDVPPs) based on Internet data centers (IDCs) and geographically concentrated VPPs (GCVPPs) that aggregate flexible loads (FLs). The proposed strategy enhances resistance to wildfire-induced uncertainties by facilitating coordinated operations between these two types of VPPs. At the upper level, an improved maximum flow model is introduced to quantify the dynamic changes in the workload transfer capability of IDC (WTCI) under wildfire conditions, and stochastic model predictive control (SMPC) is employed to perform rolling optimization of generator outputs, IDC workload transfers, and load shedding, thereby minimizing the total regulation costs. Based on the load shedding instructions from the upper level, the lower level integrates GCVPPs to provide load curtailment services, effectively offsetting the load shedding power. Subsequently, the lower level feeds back the load rebound (LR) resulting from these load curtailment services to the upper-level strategy, serving as a basis for its rolling optimization. The SMPC integrates an event-driven deductive model to address the fine-grained modeling of the operational state, effectively overcoming challenges posed by discrepancies in simulation time steps arising from power system cascading failures, variations in IDC adjustment capacity, and LR effects. Finally, a modified 39-bus power system, integrated with an 8-bus IDC network, is used as a case study to validate the effectiveness of the proposed strategy.

Abstract:Wide-area measurement systems enable the transmission of measurement and control signals for wide-area damping controllers (WADCs) in smart grids. However, the vulnerability of the communication network makes the WADC susceptible to malicious cyber attacks, such as false data injection (FDI) attack and denial of service (DoS) attack. Researchers develope numerous supervised machine-learning and model-based solutions for attack detection. However, the partially labeled attack data, skewed class distributions, and the need for precise mathematical models present significant challenges for real-world attack detection. This paper introduces the cyber attack-resilient wide-area damping controller (CyResWadc) system framework to address these challenges. The proposed framework leverages semi-supervised generative adversarial network (SSGAN) model to handle partially labeled attack data. It utilizes the support vector machine-based synthetic minority oversampling technique (SVM-SMOT) for data oversampling to manage skewed class distributions. Furthermore, probing signals are used to stimulate the power system, facilitating the generation of synthetic attack scenarios under different operational conditions. If any attack is detected, an alternate pair of measurement and control signals is used for attack mitigation. The performance is validated on a developed hardware-in-the-loop (HIL) cyber-physical testbed built using the open parallel architecture laboratory-real time (OPAL-RT) simulator, industry-grade hardware, Network Simulator 3 (NS-3), and open platform for data collection (OpenPDC).

Abstract:The recent growing integration of wind farms (WFs), particularly variable speed wind turbines (WTs), results in several operational challenges to power grids integrated with WFs, such as low grid inertia and the reduced performance of measurement-based fast frequency response. To deal with such challenges, grid operators use WF active power controllers (WFAPCs) to enhance frequency control support from WTs and improve the frequency stability of the grid. However, the operation of WFAPC relies on measurements received through communication networks and cyber layers of WFs, which consequently makes them prone to cyber threats, e.g., false data injection (FDI). On this basis, firstly, this paper analyzes the cybersecurity vulnerabilities of WFAPCs and the possible impacts of exploiting cybersecurity vulnerabilities on the frequency response of WF and frequency stability of the grid. Then, based on the knowledge of intruders, two attacks, i.e., white-box and black-box FDI attacks, are developed against WFAPCs. Afterward, to detect these attacks, a novel bi-level detection and mitigation technique based on support vector machine (SVM)-based technique and long short-term memory (LSTM)-based technique is developed, which is implemented at the control center of the WF (primary detector) and at the dispatch center of the power grid (secondary detector), respectively. These detectors classify real-time measurements into attack and normal operation. Additionally, a hierarichical mitigation technique is proposed to counter the developed cyber attacks by replacing the active power reference signal of WF with new values obtained based on the droop control theory. The impacts of the attacks and the effectiveness of the proposed bi-level technique are evaluated using the modified 39-bus benchmark.

Abstract:Among disasters that may lead to large-scale blackouts of power systems, wind storms introduce spatio-temporal variations in restoration security risks, making large-scale power system restoration more difficult. Power system restoration during wind storms requires coordinated efforts among the regional independent system operators, transmission system operators, and distribution system operators. However, existing research mainly focuses on the coordination between transmission system operators and distribution system operators, which limits its applicability to large-scale blackouts caused by wind storms. Therefore, a spatio-temporal coordinated restoration method based on restoration security risk assessment for multi-voltage-level power systems (MVLPSs) is proposed in this paper. Typhoons, known for their high wind speeds and destructive power, are considered as the disaster scenario. First, a spatio-temporal restoration security risk assessment approach is proposed to reduce additional control costs caused by restoration security risks. Then, a spatio-temporal coordinated restoration framework for MVLPSs is established, and a triple-level optimization model for the spatio-temporal coordinated restoration of MVLPSs is proposed to maximize the net restoration benefits of MVLPSs during the full-stage restoration process. Finally, case studies on an actual 379-bus MVLPS in China are conducted to verify that the proposed method can achieve higher net restoration benefits compared with existing restoration methods.

Abstract:Operational reliability assessment (ORA), which evaluates the risk level of power systems, is hindered by accumulated computational burdens and thus cannot meet the demands of real-time assessment. Recently, data-driven methods with fast calculation speeds have emerged as a research focus for online ORA. However, the diverse contingencies of transformers, power lines, and other components introduce numerous topologies, posing significant challenges to the learning capabilities of neural networks. To this end, this paper proposes a multi-kernel collaborative graph convolution neural network (GCNN) for ORA considering varying topologies. Specifically, a physics law-informed graph convolution kernel derived from the Gaussian-Seidel iteration is introduced. It effectively aggregates node features across different topologies. By integrating additional advanced graph convolution kernels with a novel self-attention mechanism, the multi-kernel collaborative GCNN is constructed, which enables the extraction of diverse features and the construction of representative node feature vectors, thereby facilitating high-precision reliability assessments. Furthermore, to enhance the robustness of multi-kernel collaborative GCNN, the inherent pattern of the load-shedding model is analyzed and utilized to design a specialized supervised loss function, which allows the neural network to explore a broader feature space. Compared with the existing data-driven methods, the multi-kernel collaborative GCNN, combined with supervised exploration, can accommodate a wider range of contingencies and achieve superior assessment accuracy.

Abstract:Time-domain distance protection shows superior performance for transmission lines integrated with renewable energy sources (RESs). However, in 35-110 kV renewable power transmission systems, the inhomogeneity of the mixed overhead lines (OHLs) and underground cables (UGCs) negatively affects the feasibility of distance protection. This paper proposes a robust algorithm of time-domain distance protection for renewable power transmission system with the mixed OHLs and UGCs. First, based on the time-domain mathematical model, the accuracy and robustness of the conventional algorithm under inhomogeneous line parameters are evaluated. To solve the “0/0” problem caused by weak signals, the singular value decomposition-based least squares method (SVD-LSM) is adopted to avoid calculation outliers and improve the protection reliability. Meanwhile, a weighting method based on Euclidean norm is designed to overcome the problem of computational non-convergence. It also ensures the protection operation speed by using a short time window. In addition, a distance correction method is designed for mixed lines to improve the accuracy of fault location. On the basis, a prototype of the protection device is developed, and extensive hardware-in-the-loop (HIL) tests are performed to verify its feasibility and superiority. In addition, the prototype of the protection device has been applied to actual renewable power transmission systems.

Abstract:High-reliability double-sided ring collector systems have been widely implemented in offshore wind farms (OWFs). It is challenging to achieve a globally optimal network topology and a cable capacity rating for the OWF collector system (CS) simultaneously. This paper proposes an optimal collector system planning (CSP) method for OWF with double-sided ring topology based on bidirectional flow conservation method to minimize cable costs and total power losses. By analyzing the power flow direction after faults, all fault scenarios are summarized into two fault conditions. The bidirectional flow conservation method is developed to reveal the matching mechanism between different cable sequence positions and their optimal ratings, considering the minimal rating requirements. The complex high-dimensional CSP problem, which involves the coupling characteristics of different cable parameters and system power flows, is convexified by equivalent alternative methods into a mixed-integer quadratic programming (MIQP) to guarantee a global optimal solution within feasible computation time, improving the solvability and practicality. The effectiveness of the proposed optimal CSP method has been validated in MATLAB.

Abstract:Integrated energy system (IES) integrates various energy subsystems such as electricity, natural gas, heat, and the dynamic characteristics of different energy networks differ significantly. To realize the coordinated operation of heterogeneous energy flow network of electricity, natural gas, and heat, in this paper, a multi-spatial-temporal-scale coordinated optimal scheduling method of IES considering frequency support ability is presented. The method divides the IES into three layers on the spatial scale and divides IES optimal scheduling into three stages: day-ahead, intra-day and real-time on the temporal scale. In the day-ahead stage, the most economical day-ahead scheduling plan is developed. In the intra-day stage, considering the different response characteristics of the device, the slow, medium, and fast subsystem layers are divided for control, and the device output related to cold, heat, electricity, and natural gas is controlled hierarchically based on distributed model predictive control. In the real-time stage, the supporting effect of IES on power grid frequency is fully explored, and an IES active-frequency-support control method considering frequency regulation cost is proposed. Case studies show that the devices can be fully utilized with different response ability to perform the scheduling plans of each layer, effectively reducing the system operation cost and improving the frequency quality.

Abstract:The offshore-onshore integrated energy system (OOIES) comprises offshore gas production platforms, wind farms, and onshore gas-fired combined heat and power plants, facilitating the integrated operation of multiple energy sources. To address the challenge of optimally configuring the device capacities in carbon capture and power to gas (CC-P2G) amid stochastic fluctuations in offshore gas and wind power outputs, this study proposes a multi-objective approximate dynamic programming algorithm. This algorithm solves the multi-objective stochastic optimal configuration for the device capacities in CC-P2G in OOIES by simultaneously optimizing investment and operation costs, wind power curtailment, and carbon emissions. By leveraging value function matrices for multiple objectives to solve the extended Bellman equation, the multi-objective multi-period model is decomposed into a series of multi-objective single-period optimization problems, which are solved recursively. Additionally, a weighted Chebyshev function is introduced to obtain the compromise optimal solution for multi-objective optimization model during each period. A case study of an OOIES confirms the effectiveness and efficiency of the proposed algorithm.

Abstract:The proliferation of distributed energy resources and time-varying network topologies in active distribution networks presents unprecedented challenges for network operators. While reinforcement learning (RL) has shown promise in addressing network-constrained energy scheduling, it faces difficulties in managing the complexities of dynamic topologies and discrete-continuous hybrid action spaces. To address these challenges, a graph-based safe RL approach is proposed to learn dynamic optimal power flow under time-varying network topologies. This proposed approach leverages graph convolution operators to handle network topology changes, while safe RL with parameterized action ensures policy development. Specifically, the graph convolution operator abstracts key characteristics of the network topology, enabling effective power flow management in non-stationary environments. Besides that, a parameterized action constrained Markov decision process is employed to handle the hybrid action space and ensure compliance with physical network constraints, thereby accelerating the deployment of safe policy for hybrid action spaces. Numerical results demonstrate that the proposed approach efficiently navigates the discrete-continuous decision space while accounting for the constraints imposed by the dynamic nature of power flow in time-varying network topologies.

Abstract:The static var compensator (SVC) is a cost-effective device in flexible AC transmission system (FACTS) family. We introduce an improved artificial hummingbird algorithm (IAHA) for optimal allocation of SVCs in distribution networks to maximize energy efficiency. Three loading levels (low, medium, and high) per day are investigated. The proposed IAHA is evaluated on the IEEE 33-bus distribution network (DN) and 69-bus DN. The proposed IAHA demonstrates notable improvements in cost savings and voltage profile compared with the conventional artificial hummingbird algorithm (AHA). In addition, it enhances energy savings across various loading conditions and outperforms the conventional AHA in both best and average performance metrics. Although raising the compensation limit initially increases cost savings, the benefits decrease beyond a threshold, highlighting the importance of balancing the compensation levels for maximum efficiency.

Abstract:This paper proposes a robust faulted line-section location method based on the normalized quantile Hausdorff distance (NQHD) algorithm for detecting single-phase-to-ground faults in distribution networks. The faulted line section is determined according to the characteristic differences between the zero-sequence currents on the faulted and healthy line sections. Specifically, the zero-sequence currents at both ends of a healthy line section are highly similar to each other, while such is generally not the case on a faulted line section. The NQHD algorithm can disregard extremes or outliers while also providing a normalized scaling in different scenarios. Thus, it can be applied to calculate the robust waveform similarity of zero-sequence current waveforms at both ends of different line sections for identifying reliably the faulted line section even under the interference of outliers. The results demonstrate the good performance of the proposed method in detecting single-phase-to-ground faults under different fault conditions. Comparative tests with the existing methods confirm the advantageous robustness of the proposed method against the impacts of outliers and noises.

Abstract:The coordination of power distribution networks (PDNs) and microgrids (MGs) is challenging due to the abundant resources and their dispersed geographical distribution, making centralized computation inefficient. To address this issue, we propose a coordination framework with single leader and multiple followers that allows limited information exchange. In this framework, the PDN operators act as leaders, while the MG operators act as followers. However, variations in load and renewable energy during MG scheduling intervals can cause variability in power transactions between PDNs and MGs. This variability can reduce the net revenue of MGs and increase the operation costs of PDNs, which makes it essential to consider the worst-case fluctuations. We introduce a multi-agent robust deep reinforcement learning (MARDRL) approach for coordination of PDNs and MGs, accounting for the worst-case scenarios. The numerical results on the test systems verify the effectiveness of the proposed approach in enhancing the coordination of PDNs and MGs.

Abstract:The increasing integration of inverter-based renewable energy sources (RESs) has significantly reduced the power grid inertia, leading to challenges in maintaining frequency stability. Virtual synchronous generators (VSGs), which emulate the behavior of synchronous generators (SGs), can help address this issue by providing synthetic inertia and improving system stability during disturbances. The paralleled operation of VSGs and SGs is particularly important in islanded microgrids, where small SGs are commonly used for power generation. This paper presents a comprehensive dynamic model of a paralleled VSG-SG system and proposes a model predictive control (MPC) strategy for VSG to enhance disturbance rejection and improve dynamic performance. Additionally, an adaptive delay compensator (ADC) is introduced to manage communication delays between the control center and system. Simulation results in MATLAB/Simulink demonstrate the effectiveness of the MPC-based VSG control method in improving frequency control in various disturbance scenarios.

Abstract:Liquefied natural gas (LNG), recognized as the primary form for natural gas transportation, can release substantial cold energy during gasification. To make efficient use of this cold energy, this paper proposes a data-driven stochastic robust (DDSR) energy management method for the multi-stage cascade utilization of LNG cold energy in a multi-energy microgrid (MEMG) of an LNG receiving terminal. Firstly, a general scheduling model considering the flexible coupling between adjacent stages, energy losses, and electric power consumption for the cascade utilization of LNG cold energy is introduced. This model is applied to carbon capture, cryogenic power generation, and direct cooling, which are sequentially associated with the deep, medium, and shallow cooling zones of LNG cold energy, respectively. Moreover, a two-stage energy management framework is proposed to coordinate the cascade utilization of LNG cold energy with other energy resources in the MEMG. To tackle the uncertainties of renewable energy generation and various loads, a DDSR-based solution method is developed, aiming to achieve both economic benefits and solution robustness by identifying the worst-case scenarios and the corresponding worst-case probability. Accordingly, a Benders decomposition-based solution algorithm is proposed to divide the original problem into a master problem and a slave problem, which are solved iteratively. The simulation results verify the effectiveness and high efficiency of the proposed DDSR energy management method for multi-stage cascade utilization of LNG cold energy.

Abstract:The growing electricity demand, combined with the increasing integration of photovoltaic (PV) generation into the distribution system, requires higher flexibility from the demand side. This paper proposes a customized scheduling approach for demand response (DR) of customers with dispatchable inverters in distribution-level PV facilities. Based on the Chilean context, the proposed approach enables these energy resources to provide flexibility in the technical and economic management of the distribution system operator (DSO). Specifically, a bi-level optimization model is introduced. At the upper level, the DSO minimizes distribution system costs by determining daily price signals for customers based on their response profile classes (RPCs) and active and reactive power set points for PV facilities. At the lower level, customers aim to reduce their electricity bills. In addition, the proposed approach ensures the reliable operation of the distribution system with high probability by addressing uncertainty through chance constraints (CCs). Incorporated CCs in the distribution system modeling include the squared magnitude of nodal voltage, complex power flow in lines, and apparent power of inverters. Finally, two case studies are presented, involving 420 residential and commercial Chilean customers with two distribution-level PV facilities using real-world market prices and daily consumption profiles on the IEEE 37-node test feeder. Results demonstrate how the proposed model enables the customized scheduling of customers and PV facilities, highlighting its effectiveness over the uniform price scheme.

Abstract:Accurate load profile data are essential for optimizing energy systems. However, real-world datasets often suffer from low resolution and significant missing values. To address these challenges, this paper introduces physics-informed loss generative adversarial network (PIL-GAN), a model that combines generative adversarial networks (GANs) with physics-informed losses (PILs) derived from physics-informed neural networks (PINNs) that are integrated directly into the generator. High-resolution load profiles are generated that not only fill in missing data but also ensure that the generated profiles adhere to physical laws governing the energy systems, such as energy conservation and load fluctuations. By embedding domain-specific physics into the generation process, the proposed model significantly enhances data quality and resolution for low-quality datasets. The experimental results demonstrate notable gains in data accuracy, resolution, and consistency, making PIL-GAN an effective tool for energy management systems. The PIL-GAN also has broader applicability in other fields such as generating and inpainting high-resolution datasets for energy systems, industrial processes, and any domain in which data must comply with real-world physical laws or operational requirements.

Abstract:Distributed photovoltaic (PV) entities can be coordinated to provide reactive power for voltage regulation in distribution networks. However, integrating large-scale distributed PV entities into reactive power optimization makes it difficult to balance the individual benefit of each PV entity with the overall economic efficiency of the system. To address this challenge, we propose a market-oriented two-stage reactive power regulation method. At the first stage, a long-term multi-layer reactive power capacity market is created to incentivize each PV entity to provide reactive power capacity, while ensuring their financial interests are guaranteed. At the second stage, a real-time multi-layer reactive power dispatch mechanism is introduced to manage the reactive power generation of distributed PV entities, prioritizing the dispatch of lower-cost PV entities to maximize system-wide economic efficiency. Simulation results based on a real Finnish radial distribution network demonstrate the effectiveness of the proposed method in optimizing reactive power for large-scale distributed PV entities.

Abstract:The transformer inrush current has been a potential threat in wind farms connected modular multilevel converter based high-voltage direct current (WF-MMC-HVDC) system due to the low overcurrent capability of power electronic devices. To investigate this issue, this paper develops a complete harmonic state space (HSS) model of the WF-MMC-HVDC system containing saturable transformers. The severity of the inrush current is investigated under different transformer configurations and the result is compared with EMTP simulations. More importantly, key factors that influence inrush current characteristics in a WF-MMC-HVDC system are studied using the single-input single-output impedance model derived from the linearized HSS model. The results indicate that wind farms have a minor impact on the inrush current characteristics, whereas V/F controlled modular multilevel converter (MMC) reduces its output voltage during transformer energization, thereby mitigating the severity of the inrush current. The severity of the inrush current largely depends on the resonance point determined by the transmission line. In the case of offshore WF-MMC-HVDC system, long submarine cables may cause severe harmonic amplifications and even do not attenuate for a long time.

Abstract:Modular multilevel converters (MMCs) have emerged as a promising solution for integrating renewables. In case of photovoltaic (PV) systems, PV arrays can be integrated at the submodule (SM) level, and the distributed maximum power point tracking (DMPPT) can be achieved through AC and DC circulating current control and perturb and observe (P&O) methods. However, this implementation is hindered by the need for numerous measurements, since the voltage and current of all PV arrays in each SM must be known. To address this issue, we propose a three-phase reduced-sensor MMC with distributed MPPT for PV integration based on an extended Kalman filter (EKF). For each MMC arm, the EKF estimates the voltage and irradiance of each SM by exploiting their gate signals and duty cycles as well as the arm current and voltage. This solution is compatible with uniform and non-uniform irradiance conditions both under the steady-state and transient conditions and uses significantly fewer sensors than other strategies employed in similar-purpose MMCs, while achieving comparable efficiency. Moreover, by exploiting the PV array characteristics, it allows performing DMPPT more directly, without using P&O methods. These features are confirmed by simulations of an MMC-based PV system with 12 SMs per arm.