Publications

Artificial intelligence (AI) is driving unprecedented growth in data center (DC) scale and power demand. AI workloads impose highly dynamic, difficult-to-forecast power profiles on the utility grid, creating reliability and stability challenges that conventional DC architectures are not designed to address. This paper provides a critical review of energy storage systems (ESSs) as the key enabling technology for reliable grid integration of AI DCs. We organize the review around a four-layer hierarchical taxonomy, namely chip-level buffering, rack/server-level ESSs, facility-level uninterruptible power supply (UPS) systems, and grid-scale battery energy storage systems (BESSs), supplemented by non-battery technologies including fuel cells (FCs) and thermal energy storage (TES). Each layer is analyzed with respect to response timescale, power and energy ratings, operational role, integration challenges, and coordination requirements. Grid-interactive UPS (GiUPS) systems are examined in depth as a critical evolution of passive backup infrastructure into active grid-support assets capable of frequency regulation, fast frequency response, and voltage ride-through. Grid-scale BESSs are reviewed for their roles in peak shaving, renewable integration, and ancillary service provision. Second-life battery energy storage systems (SLBESSs) are evaluated as a cost-effective alternative for large-scale deployment. Key findings include: (i) AI DC load profiles differ fundamentally from traditional loads in their sub-second variability, making conventional ESS dispatch strategies insufficient; (ii) hierarchical, coordinated ESS deployment across all layers is necessary for effective load smoothing and grid support; and (iii) significant gaps remain in simulation tools, degradation modeling, load forecasting, and optimal multi-layer sizing. This review identifies open research challenges and future directions at the intersection of AI computing infrastructure and power system integration.

Artificial intelligence (AI) data centers are driving rapid electricity load growth across all U.S. ISO/RTO regions, raising both system costs and carbon exposure. This study develops a three-level Stackelberg--Bayesian game in which a regulator (leader) sets carbon penalties and subsidies, a single ISO capacity market clears against an energy balance modeled as a classical generation-expansion problem, and technology-specific investors (followers) decide capacity and operation under incomplete information, yielding a Bayesian Nash equilibrium. The AI impact is captured parsimoniously as an additional load-growth factor on a greenfield-incremental expansion, isolating how much new capacity the growth pulls in and which technology fills it. Within this framework, we consider second-life battery (SLB) storage competing against new/first-life storage for capacity-market revenue. We quantify how a carbon tax, a renewable subsidy, and an SLB subsidy reshape the equilibrium investment mix, carbon emissions, and profit. Different scenarios are compared at the end based on cost-effectiveness and reduced carbon emissions.

Buildings account for a substantial share of global energy consumption and associated carbon emissions, which has increased interest in net-zero energy buildings (NZEBs) and their efficient operation. In NZEB energy management, operational scheduling must coordinate distributed generation (DG), energy storage systems (ESSs), utility-grid power exchange, and thermal devices while minimizing operating cost under variable renewable generation and demand. Although mixed-integer linear programming (MILP) can provide optimal schedules, repeated online optimization may impose a considerable computational burden in real-time applications. This paper proposes a fine-tuned large language model (LLM) framework as a surrogate for MILP-based NZEB scheduling, with the aim of providing computationally efficient decision support and a flexible architecture for future smart-building energy management. The MILP model is first used to generate optimal set-points for DG dispatch, ESS charging/discharging power, grid buy/sell power, and heating/cooling device operation. These optimization-derived input-output pairs are then used to fine-tune the LLM so that it can directly predict near-optimal operational decisions without repeatedly solving the MILP problem. The proposed framework is evaluated against benchmark learning models using the same prediction-error metrics. The results show that the fine-tuned LLM achieves the best overall performance, with a test MAE of 0.0853 and RMSE of 0.286, outperforming both ANN and XGBoost under the same evaluation setting. Compared with ANN, the fine-tuned LLM reduces MAE and RMSE by 88.2% and 33.3%, respectively, while also improving over XGBoost by 79.7% in MAE and 24.7% in RMSE. In addition, the predicted schedules are assessed in terms of operational feasibility and system-level performance. These findings indicate that fine-tuned LLMs can serve as computationally efficient surrogates for near-optimal NZEB control.

This paper investigates the capability of off-the-shelf large language models (LLMs) to solve the economic dispatch (ED) problem. ED is a hard-constrained optimization problem solved on a day-ahead timescale by grid operators to minimize electricity generation costs while accounting for physical and engineering constraints. Numerous approaches have been proposed, but these typically require either mathematical formulations, face convergence issues, or depend on extensive labeled data and training time. This work implements LLMs enhanced with reasoning capabilities to address the classic lossless ED problem. The proposed approach avoids the need for explicit mathematical formulations, does not suffer from convergence challenges, and requires neither labeled data nor extensive training. A few-shot learning technique is utilized in two different prompting contexts. The IEEE 118-bus system with 19 generation units serves as the evaluation benchmark. Results demonstrate that various prompting strategies enable LLMs to effectively solve the ED problem, offering a convenient and efficient alternative. Consequently, this approach presents a promising future solution for ED tasks, particularly when foundational power system models are available.

The rapid rise in electric vehicle (EV) adoption presents significant challenges in managing the vast number of retired EV batteries. Research indicates that second-life batteries (SLBs) from EVs typically retain considerable residual capacity, offering extended utility. These batteries can be effectively repurposed for use in EV charging stations (EVCSs), providing a cost-effective alternative to new batteries and reducing overall planning costs. Integrating battery energy storage systems (BESS) with SLBs into EVCSs is a promising strategy to alleviate system overload. However, efficient operation of EVCSs with integrated BESS is hindered by uncertainties such as fluctuating EV arrival and departure times and variable power prices from the grid. This paper presents a deep reinforcement learning (DRL)-based planning framework for EV charging stations with BESS, leveraging SLBs. We employ the advanced soft actor-critic (SAC) approach, training the model on a year's worth of data to account for seasonal variations, including weekdays and holidays. A tailored reward function enables effective offline training, allowing real-time optimization of EVCS operations under uncertainty.

Decarbonization is driving power systems toward more decentralized, self-governing models. While these technologies improve efficiency, planning, operations, and reduce the carbon footprint, they also introduce new challenges. In modern grids, particularly with the integration of power electronic devices and high penetration of Renewable Energy Sources (RES) and Inverter-Based Resources (IBRs), traditional reliability concepts may no longer ensure adequate performance due to systemic restructuring. This shift necessitates new or significantly modified reliability indices to capture the characteristics of the evolving power system. Ensuring converter reliability is essential for effective planning, which requires precise, component-to-system-level modeling, as different converters impact system performance indicators. However, the existing literature in this field faces a significant limitation, as most studies focus on a singular perspective. Some examine reliability at the device-level, others at the component-level, while broader reviews in power systems often emphasize system-level analysis. In this paper, we aim to bridge these gaps by comprehensively reviewing the interconnections between these levels and analyzing the mutual influence of power converter and system reliability. A key point to highlight is that, with the rapid evolution of modern power grids, decision-makers must adopt a multi-level approach that incorporates insights from all levels to enable more accurate and realistic planning and operational strategies. Our ultimate goal is to provide an in-depth investigation of studies addressing the unique challenges posed by modern power grids. Finally, we will highlight the gaps in the literature and suggest directions for future research.

Estimation of phasors holds significant importance within power system protection applications. Due to the inherent inductive characteristics of electric power systems, fault currents often exhibit transient oscillations, such as decaying DC offset (DDC), higher-order harmonics, and noise. Fourier analysis stands out as a widely adopted technique for phasor estimation calculations. Nevertheless, when dealing with decaying DC currents, which may persist for up to 5 cycles (in the case of a 50/60 Hz system), the accuracy of Fourier analysis diminishes significantly. An inaccurate phasor estimation could lead to malfunction or misoperation of protective devices, most notably distance protection relays, an undesirable outcome. Therefore, it becomes imperative to address the issue of DC offset signal mitigation. While various analytical methods have been proposed in recent years, there exists a pressing need to explore less complicated data-driven approaches. In this paper, we aim to employ two neural network methods to estimate the current in less than one cycle, taking into account second-order DDC and system frequency deviations. Compared with other presented data-driven methods, the proposed scheme only uses a half cycle of the fault current signal with a smaller number of trained parameters and without applying encoder/decoder methods.

Natural disasters like earthquakes, storms, and floods have an adverse impact on the critical infrastructures operation. The severity of the occurred damages on the infrastructure can be assessed by failure probabilities using fragility curves. In this paper, a Stochastic Infrastructure Damage Evaluation (SIDE) method is proposed. SIDE is a partition-based event tree approach. It estimates cumulative damage from disasters on distribution systems. The system is broken down based on equipment damage levels. SIDE evaluates the probability of damage in each partition. It uses reliability block diagram and the fragility curves of its components. The method improves event tree analysis by using partition-based modeling instead of traditional series equipment sets. Additionally, it enhances analytical efficiency by ignoring unlikely or non-expecting scenarios from the partition-based event trees (PBET). The method was applied to a power distribution system (RBTS4) as a critical infrastructure under an earthquake scenario, with results compared to Monte Carlo simulations (MCS). Finally, the merit of the proposed method is illustrated using sensitivity analysis, considering changing disaster severity and budget. The paper highlights a key observation, where in most budget allocation scenarios, the location of switches often overlaps. This suggests that beyond a certain budget, the SIDE method stops adding more switches.

This chapter reviews the genetic algorithm (GA), a population-based metaheuristic inspired by natural selection, and its application to two central power-system problems: generation expansion planning (GEP) and unit commitment (UC). After an overview of heuristic optimization in power systems, the chapter formulates GEP and UC, explains how each is encoded and solved with GA operators, and solves a sample GEP problem with the GA approach, discussing the results and the advantages of GA over gradient-based methods for integer-valued planning decisions.

This chapter presents the optimal power flow (OPF) problem and its solution with the genetic algorithm (GA). OPF is formulated as a constrained nonlinear optimization problem with equality and inequality constraints on bus voltages, generator outputs, and line flows. Evolutionary algorithms are compared with classical and mixed-integer approaches in terms of constraint handling, scalability, robustness, and flexibility. The application of GA to OPF is described step by step, and a sample OPF problem is solved with the GA approach, with the results presented and discussed.

In recent years, active distribution systems have been exposed to congestion more than in the past, and different congestion management mechanisms have been investigated in the literature. Adopting a sharing framework for energy storage systems (ESSs) to deal with the long pay-back period and high investment cost of distributed energy resources (DERs) can bring a promising solution for relieving congestion. In this paper, a framework for simultaneous energy cost optimization and congestion management by using community energy storage (CES) is proposed. As a case study, a CES within a distribution system, connected to four microgrids (MGs), is considered. The shared storage system enables the MGs to reduce their energy costs by optimizing the operation of the battery using a heuristic optimization algorithm, specifically the teaching-learning-based optimization (TLBO) algorithm. Simultaneously, the distribution system operator (DSO) leverages the shared storage to alleviate congestion by purchasing charged power from the CES manager. In the proposed approach, the DSO pays a premium price for the charged power from the CES, surpassing the prevailing electricity price during congested hours. Moreover, to manage uncertainties arising from load variations and intermittent renewable energy resources (RES), Monte Carlo simulation is employed. Through comprehensive simulations and analyses, the proposed approach demonstrates the potential of CES as an effective tool for congestion management and operational cost optimization in distribution systems, providing economic benefits to both the MGs and the DSO.

A shared pool of grid-scale storage resources called Cloud Energy Storage (CES) can bring substantial benefits to the economical and reliable operation of MGs. However, the investment cost of CES may not be affordable for a single microgrid (MG). As a solution, we propose an approach in which neighboring microgrids in a distribution network collaborate and form a multi-microgrid (multi-MG) to install a shared CES to increase their profit and improve their reliability. Different investment scenarios are evaluated by considering the yearly reward from TSO and DSO. For each of the investment scenarios, TSO and DSO give a yearly reward based on the contribution of CES in peak-shaving and distribution network operation yearly cost reduction. Afterward, a decision table is provided in which, for all investment scenarios, profit, reliability index based on expected energy not supplied (ENS), and TSO-DSO yearly reward are determined. Finally, the microgrids select one of the investment scenarios using a multi-attribute decision-making approach. Simulation results of a case study validate the effectiveness of the proposed collaborative decision-making framework in increasing the economic value of CES investment, reliability enhancement in multi-MG, and peak-shaving.

In the present climate, due to the cost of investments, pollutants of fossil fuel, and global warming, it seems rational to accept numerous potential benefits of optimal generation expansion planning. Generation expansion planning by regarding these goals and providing the best plan for the future of the power plants reinforces the idea that plants are capable of generating electricity in environmentally friendly circumstances, particularly by reducing greenhouse gas production. This paper has applied a teaching–learning-based optimization algorithm to provide an optimal strategy for power plants and the proposed algorithm has been compared with other optimization methods. Then the game theory approach is implemented to make a competitive situation among power plants. A combined algorithm has been developed to reach the Nash equilibrium point. Moreover, the government role has been considered in order to reduce carbon emission and achieve the green earth policies. Three scenarios have been regarded to evaluate the efficiency of the proposed method. Finally, sensitivity analysis has been applied, and then the simulation results have been discussed.

Global warming and climate change caused by human activities are some of the major environmental problems that the most important cause of this phenomenon is greenhouse gases, especially CO2. Due to the many consequences of this global phenomenon in the last two decades, the attention of many scientists and policymakers in the world has been drawn to this issue. Thus, they set various regulations to reduce greenhouse gases and carbon footprints. On the other hand, one of the major causes of increasing carbon emissions is electricity generation power plants; as a result, the most important solution is to reduce CO2 gas generation in these power plants. Nowadays, generation expansion planning (GEP) is an important and inevitable issue for power system designers due to energy consumption growth. Besides, finding a solution in order to satisfy the increasing electricity demand in the future years and considering carbon reduction as a new constraint to this problem is a challenging subject. This paper modelled the GEP problem with a game theory approach by considering government regulations, including carbon tax and subsidy. Therefore, four possible strategies have been assumed to achieve the goal of carbon reduction. We have considered Iran as our case study, and the results have been analyzed and discussed. Eventually, a comprehensive sensitivity analysis is carried out based on the carbon tax and government subsidy.