Time-series forecasting in the energy sector is a labor-intensive process requiring expertise in multiple areas, such as data preprocessing, feature engineering, and neural network optimization. Although large language models offer automation potential, existing large-language-model-based multi-agent systems lack specialized structures for forecasting workflows, balancing computational cost and tool capability remains challenging. To address these issues, we propose an Energy-domain End-to-end Automated Forecasting Multi-Agent System, an on-premises end-to-end forecasting framework built upon a novel Department-Collaborative Multi-agent System Structure, specifically fine-tuned for energy forecasting applications. Within the Energy-domain End-to-end Automated Forecasting Multi-Agent System dominated by Department-Collaborative Multi-agent System Structure, two core and innovative modules are introduced: (i) Structured Behavioral Knowledge Distillation, which enables small-parameter large language models to operate feature engineering tools, facilitating lightweight and efficient on-premises deployment; and (ii) the Alignment-to-Selection Framework, which automates neural network selection by integrating architectural knowledge with historical performance data. Extensive experimental results demonstrate that the novel Energy-domain End-to-end Automated Forecasting Multi-Agent System significantly reduces manual effort, while maintaining high forecasting accuracy across diverse datasets.

The gap between the industry expectations and power engineering education is becoming one of the significant barriers in workforce development in the energy sector. For the purpose of closing the gap, the curriculum in power engineering is expected to follow the up-to-date industry demand and workflow so that the students can be job-ready and competent in the energy job markets before graduation. However, the renewal of the deliverables of the courses is time-consuming and resource intensive. This paper presents an approach of data collection and analysis for curriculum development, which considers all the related stakeholders in curriculum renewal. The preliminary results of the pilot data collection are presented, demonstrating the effectiveness of the data collection method and the initiative of updating the curriculum framework. The contribution will benefit power engineering educators in course renewal for further iterations considering timely industry perspectives and students’ feedback.

In the past, many studies were applied various statistical analysis methods to evaluate students’ learning achievement and satisfaction for improving the effectiveness of online teaching. However, most of these decided to rely on relatively fixed fundamental quantitative methodologies to determine essential results. Few studies have adequately classified statistical methods in engineering education to critically consider correlational trends or causal mechanisms in the field and make research results more explanatory and inclusive. Therefore, our main challenge is appropriately selecting quantitative or qualitative statistical methods used in online engineering education to make the research results more convincing. To fill this ‘gap,’ this article re-examines previous papers to summarize a statistical method in the online engineering discipline from diverse perspectives and construct a new mechanism of evaluating statistical methods for effective research in this field. Our goal is to provide an unexplored review of statistical methods of the online teaching and learning process considering the engineering educational perspective.

Modern electrical power systems have an increasing penetration of distributed generation (DG). This increase yields the requirement for integration of renewable energy and distributed energy generation into the educational curriculum of power engineering. However, the integration of associated knowledge in the curriculum framework is challenging because power system industry is experiencing a fast and continuing transition and associated teaching and learning materials are becoming quickly outdated in higher education institutions. This paper demonstrates the use of the backward design model in the power engineering discipline, where a Year 3 course is redesigned. The course is entirely implemented in an online mode, and the feedback and perception from the students are presented and analyzed. The results show the necessity of the curriculum framework development for power engineering education in a qualitative manner and provide guidance for course instructors to integrate distributed energy generation in undergraduate coursework programs.

Loss of synchronization (LOS) in case of synchronous reference frame based phase-locked loop (SRFPLL) affects the low voltage ride-through performance of grid-connected power converters. To avoid this issue, this paper proposes a hybrid grid synchronization technique that includes two frequency and two phase-angle estimators. It uses the frequency and phase-angle estimation by the SRFPLL during normal grid operation and switches to the arc tangent based phase-angle and corresponding frequency estimation in the αβ-frame during grid faults. To avoid sudden mode transfer between these estimators a common transition algorithm is proposed as well, which is controlled by the phase-angle difference between them. The proposed hybrid grid synchronization transition is implemented in the frequency dependent proportional and resonant current controller of the converter. The controller is run with the low voltage ride-through strategy during the faults. It is observed that the converter having the proposed grid synchronization technique provides robust current controller performance as the fast tracking of grid current on the fault inception and recovery.

In this article, the impact of prediction errors on the performance of a domestic power demand management is thoroughly investigated. Initially, real-time peak power demand management system using battery energy storage systems (BESSs), electric vehicles (EVs), and photovoltaics (PV) systems is designed and modeled. The model uses real-time load demand of consumers and their roof-top PV power generation capability, and the charging-discharging constraints of BESSs and EVs to provide a coordinated response for peak power demand management. Afterward, this real-time power demand management system is modeled using autoregressive moving average and artificial neural networks-based prediction techniques. The predicted values are used to provide a day-ahead peak power demand management decision. However, any significant error in the prediction process results in an incorrect energy sharing by the energy management system. In this research, two different customers connected to a real-power distribution network with realistic load pattern and uncertainty are used to investigate the impact of this prediction error on the efficacy of an energy management system. The study shows that in some cases the prediction error can be more than 300%. The average capacity of energy support due to this prediction error can go up to 0.9 kWh, which increases battery charging-discharging cycles, hence reducing battery life and increasing energy cost. It also investigates a possible relationship between environmental conditions (solar insolation, temperature, and humidity) and consumers' power demand. Considering the weather conditions, a day-ahead uncertainty detection technique is proposed for providing an improved power demand management.

Powering small islands with reliable, affordable and green electricity is a big challenge due to their dispersed geographical location with limited number of consumers and the heavy dependence on fossil fuels. This paper aims to address this challenge of reducing dependency on fossil fuel generators by providing an easy and feasible solution using available and accessible energy resources. The proposed method utilizes the bidirectional energy transfer mechanism available in electric boats (EBs) to support the consumers' power demand. It proposes a new real-time load-support (RTLS) system with a coordinated control using EBs, community generators, and battery energy storage systems. It analyzes the management of the intermittent source-dependent small-scale grid in real time, under various weather, load, and battery state-of-charge conditions. The RTLS system coordinates the customers' load demand with the available EBs, photovoltaics, and battery storage to provide efficient load support and to regulate the bus voltage and frequency. The efficacy of the proposed system is validated both computationally in a real network and in a laboratory setup. It is found that this novel system can substantially reduce the grid load demand and maintain the power quality under various load/source uncertainties and fault conditions. The system robustness is also evaluated considering undesirable conditions, such as severe three-phase faults and sudden EB disconnections. The performance of the proposed method is compared with that of the day-ahead load management approach to validate its effectiveness under various scenarios.

This paper presents second-order cone programming (SOCP) and semidefinite programming (SDP) models to solve the multiperiod optimal control problem of unbalanced three-phase distribution grids with battery energy storage systems. The decision variables are the active and reactive power of the battery energy storage system. The objective is to minimize the power loss and energy purchase cost from the distribution substation. The optimal dispatch problem requires solution by volt/VAR optimization. The SOCP and SDP models ensure global optimum of original nonconvex nonlinear programming problem. Mupowerltiobjective volt/VAR optimization is performed and the benefits of reactive power support from battery energy storage systems are explored. A simulation-based heuristic to extract rank-one solution from the higher rank solution of SDP model is also proposed. Empirical results show the numerical stability and exactness of the proposed three-phase SOCP and SDP models.

Accurate active and reactive power sharing is of utmost importance in an inverter-interfaced AC microgrid. Power droop-control has been proven as a good design tradeoff and communication less technique to achieve load sharing. Implementation of power droop-control requires both the voltage and current tapping at the inverter output to calculate the power and achieve synchronization with rest of the system. On the other hand, recently proposed virtual oscillator controller (VOC) features inherent synchronization using only the current measurement. However, VOC being a non-linear oscillator introduces harmonic distortion in the output. This paper investigates the power sharing capability, dynamic response and harmonics profile for a system consisting of parallel-connected (i) droop-controlled inverter and (ii) virtual oscillator-controlled inverter. It has been demonstrated that power sharing between these two different control techniques can be achieved by properly designing both the droop-controller and VOC-parameters. The harmonic distortion present in the output of VOC results in the harmonic current to flow between the two inverters. The harmonic distortion can be reduced by designing a tightly regulated VOC inverter. However, this results in a slow dynamic response. Several scenarios including the plug-n-play capability, step change in load, and the design trade-off in between dynamic response and harmonic content present in the output is demonstrated through the simulation results.

Accurate and fast estimation of magnitude, frequency, and phase of the grid voltage is an essential part of the synchronization process and power-sharing control among parallel connected inverter-based AC microgrids, while running in islanded mode. Synchronization can be done using both open-loop and closed-loop estimation techniques. The conventional closed-loop synchronization technique used is Synchronous Reference Frame based Phase Locked Loop (SRF-PLL), estimates the phase angle of the grid voltage with a typical settling time of one or multiple fundamental periods. Moreover, the SRF-PLL estimates both, the frequency and phase angle of the voltage in one single loop, thus a jump in the phase angle affects the frequency estimation, which will have a negative impact on the power sharing control. To overcome these issues, this paper proposes an open loop synchronization technique based on Teager-Kaiser Energy Operator (TKEO)and Cross Energy Operator (TKCEO)to estimate the grid voltage frequency and phase angle, respectively. Presented results show that the power-sharing with energy operator based calculations gives a better response when compared to classical SRF-PLL.

Due to current limits in inverter dominated low voltage AC microgrids, there exists challenges in using the existing protection schemes available for conventional power systems. The design of a proper protection scheme demands faster detection and classification of faults. In this paper, a novel technique using Teager Energy Operator (TEO) for the derivative of current in each phase is proposed. It is observed that the proposed technique can detect faults within a quarter cycle of fundamental frequency from the time of occurrence of fault. The TEO of each phase currents' derivative is then used to classify the faults.