Smart charging goes beyond plugging an electric vehicle into a conventional charger. It uses digital communication among vehicles, charging equipment, and electricity networks so charging can be scheduled or adjusted in response to grid conditions and electricity prices. The technology can direct charging toward off-peak periods and periods of renewable generation, potentially easing pressure on the grid and reducing charging costs. Some smart-charging arrangements can also support vehicle-to-grid operation, in which vehicle batteries supply electricity back to the network. These functions make consumer acceptance important because use may involve an application, automated equipment, smart meters, tariff information, and decisions about when a vehicle should charge.

The study focused on people who already owned and used electric vehicles, rather than prospective buyers. Participants were at least 18 years old and were recruited in Chennai, Bengaluru, Mumbai, Hyderabad, Delhi, and Kolkata between March and June 2026. Researchers distributed an anonymous questionnaire online through social media, electric-vehicle owner communities, professional networks, and vehicle forums, and approached owners offline at charging stations, public places, dealerships, service centers, and workplaces. A pilot study with 50 respondents was used to check clarity and reliability before minor questionnaire revisions were made. The final analysis included 535 complete responses, with every item assessed on a five-point scale from strongly disagree to strongly agree.

Partial least squares structural equation modeling was used to evaluate a model containing six predictors and one outcome: intention to use smart electric-vehicle charging when available. The predictors were technology readiness, performance expectancy, effort expectancy, facilitating conditions, price value, and environmental concern. Technology readiness captured comfort, willingness, confidence, and preparedness for new technologies. The acceptance factors represented perceived usefulness, ease of learning and use, access to supporting resources, and the balance between benefits and cost. Environmental concern was included as the ecological dimension of the framework. Statistical significance was tested with 5,000 bootstrap resamples, while reliability, convergent validity, discriminant validity, model fit, explained variance, predictive relevance, and effect sizes were assessed with additional measures.

All six proposed relationships were supported in the statistical model. Technology readiness was associated with intention at β = 0.241, t = 5.863, p < 0.001, with a 95% bootstrap confidence interval from 0.160 to 0.324. Performance expectancy had an identical coefficient of 0.207 to effort expectancy, with t values of 5.651 and 5.694, respectively; both had p values below 0.001. Price value was associated with intention at β = 0.164, environmental concern at β = 0.151, and facilitating conditions at β = 0.140, also with p values below 0.001. The measurement model performed adequately: Cronbach’s alpha values ranged from 0.861 to 0.886, average variance extracted values ranged from 0.705 to 0.739, and all Heterotrait-Monotrait ratios were below 0.85.

The numerical ranking did not translate into large independent effects. Technology readiness had the largest effect size, but its f² value was only 0.075, which the researchers classified as small. Effort expectancy and performance expectancy each had f² = 0.055, followed by price value at 0.035, environmental concern at 0.029, and facilitating conditions at 0.025. The model’s R² value of 0.227 therefore indicates a partial explanation of intention, leaving most of its variation associated with factors not included in the model. The researchers did not include possible influences such as government support, perceived risk, social influence or trust. The standardized root mean square residual was 0.040, below the cited recommendation of 0.08, while the predictive Q² value was 0.202.

The findings indicate that convenience with digital systems may be especially relevant to willingness to try smart charging, but they do not show that comfort with technology alone determines adoption. The strongest association is consistent with a system that asks users to manage automated processes, applications, charging schedules, and sometimes dynamic price signals. Performance expectancy covers the perceived benefits of automatic charging, convenience, improved energy management, and potentially lower electricity costs. Effort expectancy reflects whether users believe they can easily learn and operate the system. Price value asks whether the benefits appear commensurate with the cost, while facilitating conditions capture perceptions of equipment, reliable electricity, internet access, charging facilities, and technical support.

Environmental concern was also associated with intention, fitting with smart charging’s potential to support more efficient electricity use and renewable-energy integration. However, the study did not measure whether participants actually shifted charging to lower-demand or renewable-electricity periods, and it did not demonstrate reductions in emissions, grid stress, or household expenditure. Likewise, the positive facilitating-conditions result does not show that infrastructure improvements caused stronger intentions. The outcome was self-reported intention among existing owners, not observed use. The model’s predictive comparison found lower PLSpredict loss for all four intention indicators than for a linear-model benchmark, with average loss differences from −0.043 to −0.052 and p values no higher than 0.004, but the overall root mean square error was 0.898 and mean absolute error was 0.712.

Several features limit how broadly the results can be generalized. The design was cross-sectional, so it did not track changes in attitudes or behavior over time and cannot establish temporal ordering. The sample was drawn through non-probability methods and may not represent all electric-vehicle owners in India. Respondents were predominantly male, accounting for 67.3% of the sample, and 65.4% reported charging at home, making the results potentially less applicable to owners without private parking or private charging. The largest age group was 25–35 years, at 35.5%, and 43.0% reported monthly income between ₹50,001 and ₹100,000. The study also took place in six cities at a single point in time, when charging infrastructure, consumer awareness, and smart-grid development may differ across regions.

For policymakers, electricity providers, electric-vehicle manufacturers, and charging-service companies, the results point to several possible areas for attention without guaranteeing increased use. Consumer familiarization could address uncertainty surrounding automated charging and digital interfaces, while reliable electricity, suitable equipment, internet access, and technical support could strengthen perceived facilitating conditions. User-friendly applications and clearer pricing could improve expected usefulness, ease of use, and perceived value. Communicating potential environmental and energy-management benefits may also resonate with environmentally concerned owners. Yet the modest effect sizes, limited explained variance, and broad geographic and demographic gaps mean that the results are best treated as an association map rather than a prescription. Longitudinal research that observes actual smart-charging use, alongside models incorporating trust, risk, social influence, and policy support, would be needed to determine what helps owners move from intention to sustained behavior.

Subject of Research: Consumer intention to use smart electric-vehicle charging among electric-vehicle owners in India

Article Title: Determinants of consumers’ intention to adopt smart electric vehicle charging in India

Article References: Shunmathy, V., & Selvam, V. (2026). Determinants of consumers’ intention to adopt smart electric vehicle charging in India. Frontiers in Sustainable Cities, 8, 1963503. https://doi.org/10.3389/frsc.2026.1963503

Image Credits: AI Generated

DOI: 10.3389/frsc.2026.1963503

Keywords: smart EV charging, electric vehicles, India, consumer intention, technology readiness, UTAUT2, environmental concern, PLS-SEM, charging infrastructure, sustainable mobility, Determinants, consumers

The researchers modeled a network of three charging stations, each with two physical ports rated at 150 kW. Station A represented a high-demand urban location, Station B a medium-density residential area and Station C a corridor serving long-distance commuters. The simulated fleet combined commercially available vehicles capable of direct-current fast charging, with battery capacities from 40 to 100 kWh and maximum onboard charging powers from 26 to 147 kW. Arrival state of charge ranged from about 16% to 83%. Analysis of individual sessions used vehicle lookup records to retrieve battery capacity, energy consumption and charging limits, while a continuous 24-hour scheduling horizon and a congested 12-hour operating period were used to examine allocation performance.

Before presenting drivers with charging options, the algorithm performs a physical feasibility filter. For an incoming vehicle, it calculates the maximum distance that can be traveled from its current battery level, then compares that range with GPS-derived travel distances to every active station. Stations beyond the reachable range are removed rather than being offered as cheaper or more convenient choices. Energy demand is calculated from the difference between arrival and requested departure state of charge, multiplied by battery capacity. Expected charging duration then follows from that energy demand divided by the lower of the port’s output and the vehicle’s maximum acceptance rate. In the illustrated fleet, every displayed session was limited by the vehicle rather than the 150 kW station supply.

This filtering changes both safety and traffic distribution. In three vehicle-level cases, a car with a calculated maximum range of 60 km could see all three stations, a vehicle with 30 km of range was denied access to a station 45 km away, and a vehicle with 15 km of range was offered only a station 10 km away. A separate stress test selected the 20 vehicles with the lowest recorded arrival states of charge. The naive approach routed all 20 toward the geographically central station and classified every one as stranded. The proposed system rerouted 11 vehicles to a reachable peripheral station and issued a system rejection to nine others whose ranges could not cover even the nearest operational port. No on-road stranding event occurred under the proposed model, although vehicles that could not reach any station were still unable to charge.

For vehicles with viable options, the system estimates queue delay at each station by comparing the arrival time with the earliest estimated finish among active sessions. It then combines expected waiting time, driving time and financial cost into a weighted score and recommends the feasible station with the lowest composite value. Weights can reflect a driver’s priorities, with an urgent customer able to place more emphasis on waiting time. This mechanism differs from sending every driver to the geographically closest station. Under that nearest-feasible-station baseline, about 290 of 300 vehicles converged on Station A, leaving the other stations underused and creating a local bottleneck. The proposed method distributed approximately 100 vehicles to Station A while steering the remainder toward less congested options.

Congestion is communicated through a step-based electricity-price multiplier. The multiplier is 1.0 for an estimated wait of 10 minutes or less, 1.2 for waits between 10 and 30 minutes, and 1.5 when the wait exceeds 30 minutes. The displayed price also includes a priority fee for urgent users. The aim is not simply to maximize revenue but to shift price-sensitive demand toward stations with shorter queues, thereby spreading charging loads across the network. In the 12-hour Station A analysis, the nearest-station baseline produced a peak queue of 284 electric vehicles and a sustained average of 206. With the menu-based method, the maximum queue fell to 94 and the time-averaged queue was approximately 66. Across the full simulation, mean system waiting time fell from 2,129.5 minutes under the baseline to 631.7 minutes with the proposed algorithm.

The scheduler separates drivers into normal, urgent and emergency tiers. Normal users accept ordinary queue delays and congestion pricing. Urgent users can pay a priority fee to receive a better position in the queue, but they cannot interrupt a charging session already in progress. Emergency vehicles form the highest tier: if their selected station is occupied, the system identifies eligible active sessions, suspends one and starts the emergency vehicle at its physical arrival time. The interrupted driver is placed at the head of the queue for resumption, the congestion multiplier is waived and a 50% discount is applied to the base electricity rate for the remaining energy. In the reported fleet, average waits were 18.5 minutes for normal users, 4.2 minutes for urgent users and zero minutes for emergency vehicles.

Victim selection is designed to limit disruption and preserve the hierarchy. The algorithm first considers only active drivers at the lowest priority level present at the station. Among those candidates, it selects the session with the latest estimated finish time, allowing the emergency vehicle to take over a port that would remain occupied longest. A higher-priority urgent session is protected when lower-priority normal sessions are available. If every occupied port belongs to an emergency vehicle, however, no eligible lower-priority victim exists, and the protocol does not evict any session. In a separate case with four normal and two urgent sessions, the system chose a normal driver with a 45-minute remaining finish even though an urgent driver had a 60-minute finish. A demonstration scenario showed a normal session beginning at time zero, an emergency arrival displacing it at 15 minutes, and the original session returning at 27 minutes with its remaining 15 minutes of charging still required.

The results also indicate that physical capacity eventually constrains the emergency guarantee. Stress tests covering fleets from 50 to 3,000 electric vehicles maintained zero waiting time for emergency arrivals until the number of concurrent emergency requests approached the physical number of charging ports. Beyond that limit, emergency waits rose sharply. The authors describe this as graceful degradation rather than an algorithmic failure: the protocol cannot preempt an equal-priority emergency session, so additional emergency demand must wait when no legal target exists. Port utilization reached 100% during peak hours and 98.2% across the 12-hour study period, with the schedule preventing simultaneous assignments to the same port. The work was also limited by fixed fleet assumptions, trust in user-declared priority, the absence of real-time grid-condition constraints and limited assessment of how displaced drivers experience interruption. The authors propose authenticated emergency credentials, behavioral checks for questionable urgent claims, demand forecasting and more developed compensation as future work.

Subject of Research: Real-time electric-vehicle charging scheduling with range filtering, dynamic pricing and emergency preemption.

Article Title: An online menu based real time preemptive scheduling for electric vehicle charging with hierarchical priority management

Article References: Khan, S., Masud, F., & Aziz, T. (2026). An online menu based real time preemptive scheduling for electric vehicle charging with hierarchical priority management. Results in Engineering, 32, Article 113591. https://doi.org/10.1016/j.rineng.2026.113591

Image Credits: AI Generated

DOI: 10.1016/j.rineng.2026.113591

Keywords: electric vehicles, charging stations, charging scheduling, state of charge, congestion pricing, emergency preemption, queue management, simulation, charging infrastructure, route feasibility, online, menu

The proposed pipeline targets two coupled controls in a battery thermal management system: coolant flow rate and coolant inlet temperature. Higher flow generally improves heat removal, while a colder inlet creates a larger temperature difference between the coolant and cells. The neural network predicts peak battery temperature, temperature non-uniformity and a dimensionless thermal-runaway risk index in a single forward pass. Differential Evolution then searches for coolant settings that balance peak temperature against spatial uniformity, which matters because hot cells in a module can age faster than cooler cells even when the absolute temperature remains within a broader acceptable range.

To build the training data, the researchers used a reduced heat-generation formulation inspired by the Bernardi model and a calibrated cooling-conductance expression based on the Dittus–Boelter correlation. They generated 3,000 samples for each of two cell formats, Tesla 21,700 and Tesla 4680, for a total of 6,000 synthetic cases. Each case varied charge or discharge rate from 0.5C to 5C, coolant flow from 0.01 to 0.15 litres per minute and inlet temperature from 20 to 35 degrees Celsius. Gaussian noise was added to emulate measurement or discretization uncertainty, and a fixed random seed was used for reproducibility.

The thermal-runaway risk measure is not a probability of a battery fire. It is a relative screening index that becomes positive only when both peak temperature exceeds 45 degrees Celsius and temperature non-uniformity exceeds 3 degrees Celsius. The index then grows with both exceedances. Its double-threshold construction reflects the model’s assumption that a high absolute temperature combined with localized thermal gradients is more concerning than either condition alone. However, the index has not been calibrated against observed thermal-runaway events, and a percentage reduction describes the model-predicted joint threshold exceedance rather than a measured reduction in real-world failure probability.

The neural network used three inputs, three outputs and hidden layers containing 128, 64 and 32 neurons, with hyperbolic-tangent activations, L2 regularization and early stopping. On a held-out set of 1,200 samples, it achieved an R-squared value of 0.9068 for peak temperature, with a root-mean-square error of 3.87 degrees Celsius, and 0.9972 for temperature non-uniformity, with an error of 0.0966 degrees Celsius. Its risk-index R-squared was 0.8854. A five-fold cross-validation of peak-temperature prediction produced an R-squared of 0.9014 plus or minus 0.0029, broadly consistent with the single test result. The model also substantially outperformed a linear-regression baseline for the risk index, whose R-squared was 0.4128.

Those accuracy figures describe how closely the network reproduces the equations that generated its training data. They do not establish how accurately those equations represent real batteries. The temperature calculation is a calibrated, lumped mapping whose numerical coefficients combine several physical effects without strict dimensional consistency, while the cooling relation uses an effective square-root dependence on flow in a regime where fully turbulent Dittus–Boelter scaling may not strictly apply. The dataset was neither produced by computational fluid dynamics nor measured experimentally. Consequently, systematic errors in the underlying reduced-order model would be inherited by the surrogate and its optimization results.

Across 20 charging conditions, the uncontrolled baseline used constant flow of 0.05 litres per minute and a 30-degree inlet temperature. It produced an average peak temperature of 45.23 degrees Celsius, average non-uniformity of 3.35 degrees Celsius and an average screening index of 0.1690. A threshold controller switched to 0.10 litres per minute and a 20-degree inlet when temperature exceeded a preset limit, improving the three averages to 30.61 degrees Celsius, 2.33 degrees Celsius and 0.0227. The optimized strategy, based on a weighted objective assigning 0.6 to peak temperature and 0.4 to non-uniformity, achieved averages of 28.46 degrees Celsius, 1.89 degrees Celsius and 0.0087. These are model-based controller comparisons, and the benchmark did not account for all practical costs of varying coolant settings.

A weight sweep at a charging rate of 3C revealed that the competing thermal objectives were only weakly conflicting within the assumed hardware limits. Across nine weight combinations, predicted peak temperatures ranged from 25.21 to 25.54 degrees Celsius, while non-uniformity ranged from 1.24 to 1.68 degrees Celsius. Extreme weighting of temperature uniformity selected maximum flow and the minimum inlet temperature, whereas strong weighting of peak temperature chose about 0.115 litres per minute at the minimum inlet because additional flow produced little further reduction. Retraining five networks with different initialization seeds produced optimized flow rates between 0.137 and 0.150 litres per minute, indicating that the flat temperature response leaves the precise flow optimum weakly determined.

The clearest hardware limitation appeared during extreme charging. At 1C to 3C, optimized peak temperatures remained near 25 degrees Celsius and the screening index was close to zero. At 4C, the optimizer reached the assumed limits of 0.15 litres per minute and a 20-degree inlet while keeping peak temperature below the critical threshold. At 5C, however, predicted peak temperature rose to 46.45 degrees Celsius and non-uniformity to 4.49 degrees Celsius, with a screening index of 0.085. In the calibrated model, the 5C heat-generation term exceeded the maximum available cooling term for both cell formats. The authors therefore conclude that safe 5C operation would require more pump capacity, colder coolant or another cooling-system upgrade rather than better set-point tuning alone.

Several safety and deployment qualifications remain. In the safety-relevant 40-to-45-degree band, the network’s local root-mean-square error was about 4.89 degrees Celsius and it showed a 0.72-degree tendency to underpredict temperature. The reported 0.125-millisecond inference time covers only the neural-network forward pass on a general-purpose processor, excluding sensing, communications and actuator response. Before safety-critical use, the authors propose conservative margins, an independent hardware over-temperature cutoff, validation against representative computational fluid-dynamics cases and hardware-in-the-loop testing with an instrumented battery module. The study therefore offers a promising computationally efficient control strategy and a clear 5C hardware benchmark, but its quantitative safety implications remain model-based until those validation steps are completed.

Subject of Research: Neural-network surrogate modeling and evolutionary optimization for multi-objective electric-vehicle battery thermal management

Article Title: Artificial neural network surrogate modeling and evolutionary optimization for multi-objective control of electric vehicle battery Thermal Management Systems

Article References: Tabeshmoghaddam, M., Zadehbagheri, M., Nejatian, S., & Kiani, M. J. (2026). Artificial neural network surrogate modeling and evolutionary optimization for multi-objective control of electric vehicle battery Thermal Management Systems. Results in Engineering, 32, Article 113618. https://doi.org/10.1016/j.rineng.2026.113618

Image Credits: AI Generated

DOI: 10.1016/j.rineng.2026.113618

Keywords: electric vehicles, lithium-ion batteries, battery thermal management, artificial neural networks, surrogate modeling, differential evolution, fast charging, thermal runaway, multi-objective optimization, model validation, Artificial, neural

The research, led by Wei Xiao of Ningbo University of Finance and Economics, introduces a framework called PGMD-Twin, built to estimate the net battery energy consumed by an electric vehicle during each one-minute window of real driving. Rather than claiming a fully deployed digital twin, the author carefully frames the work as a digital-twin-oriented energy-state estimation layer, the virtual component of a larger cyber-physical architecture that would eventually include live data synchronization and closed-loop control. The distinction matters. In the strong sense, a digital twin pairs a persistent virtual model of a physical asset with continuous synchronization and decision interfaces. This study evaluates only the virtual estimation layer, and it says so explicitly, a rare act of restraint in a field crowded with inflated claims.

The data came from the extended Vehicle Energy Dataset, a naturalistic collection of real driving records. After applying quality and target-availability screens, the analysis retained 4,114 one-minute windows from 503 vehicle-specific trips, all recorded by three 2013 Nissan Leaf vehicles. The response variable was net battery energy per minute, with a mean of 0.0831 kilowatt-hours and a range stretching from negative 0.1454 to 0.5227 kilowatt-hours, where negative values mark intervals in which regenerative braking recovered more energy than the drivetrain consumed. Crucially, the study excluded any variable that could leak the answer, such as battery current, voltage, end-of-window state of charge, or measured contemporaneous energy. Several of those signals could reconstruct the target almost directly, and leaving them in would have inflated apparent accuracy without saying anything useful about real-world performance.

At the heart of the framework sits a physics-guided core. Instead of trying to identify drag coefficients, frontal areas, or rolling resistance as separate vehicle constants, which one-minute aggregates cannot uniquely pin down, the model uses a reduced longitudinal load basis with sign-constrained effective coefficients. Velocity, cubic velocity, positive inertial load, uphill load, downhill load, and idling terms form a grey-box prior whose coefficients are restricted to physically plausible directions: rolling-like, aerodynamic-like, inertial, uphill, and idle terms must be non-negative, while the downhill term must be non-positive. This prior is then linked to a compact latent state-space encoder that carries short-term history across the current window and the preceding four minutes of the same trip, with hidden states reset at every trip boundary. The result is a hybrid in which an explicitly structured energy component anchors the estimate and a flexible learned correction absorbs what the simplified equations cannot resolve.

What sets PGMD-Twin apart from ordinary hybrid models is its treatment of distribution shift. Training was organized into episodes that mimic an unseen operating domain: the data were clustered into four statistical operating archetypes, ranging from slow, stop-and-go traffic at a mean speed of about 20.7 kilometers per hour to faster travel averaging 55.0 kilometers per hour, and during each episode three archetypes supplied training data while the fourth was held back as a pseudo-unseen domain. A first-order meta-learning update rewarded parameter adjustments that preserved performance on the withheld domain. Three softly gated experts then shared the same latent state, with a gating function conditioning their contributions on the current operating representation, and a weak worst-domain penalty gave extra attention to poorly performing source domains. The author is candid that the meta-domain term did not improve every ablation, and that the experts developed soft, gradual specialization rather than clean categorical roles.

Reliability was handled through trip-grouped conformal calibration. Conformal prediction can convert point estimates into prediction intervals with distribution-free guarantees, but only under exchangeability, and neighboring one-minute windows from the same trip are plainly dependent. Calibrating at the window level would overstate the effective sample size, so the study preserved complete trips during splitting and calibration, computing one nonconformity score per trip before deriving the interval radius. Coverage and interval width were then reported empirically across every test protocol, without extending finite-sample guarantees to arbitrary temporal or covariate shift. This is the kind of statistical hygiene that separates a credible uncertainty estimate from a decorative one.

The validation design is arguably the paper’s most distinctive contribution. Five distinct split structures were predefined: five-fold trip-grouped validation for within-distribution performance, five route-family holdouts for unfamiliar routes, four leave-one-archetype-out folds for unseen operating regimes, three leave-one-vehicle-out folds for transfer between individual cars, and four compound vehicle-archetype holdouts testing novel pairings of familiar factors. In trip-grouped testing, PGMD-Twin reached a mean absolute error of 0.0196 kilowatt-hours with an R-squared of 0.825. Under route shift the error rose only slightly to 0.0208 kilowatt-hours, to 0.0211 when an entire operating archetype was withheld, and to 0.0205 across the compound holdouts. Empirical coverage of the nominal 90 percent intervals ranged from 91.5 to 97.9 percent, staying above the nominal line in every protocol.

Yet the results also contain an uncomfortable truth. Simpler baselines, a plain multilayer perceptron and a fixed physics-data hybrid, achieved lower mean absolute error, 0.0162 and 0.0161 kilowatt-hours respectively, in the internal tests. The proposed framework’s advantage emerged only under stress: its relative error degradation was smaller than the baselines’ under route, condition, and vehicle shifts, with generalization degradation ratios of roughly 6.1 percent under route shift and 7.7 percent under condition shift, compared with 12.3 and 15.4 percent for the perceptron and 9.9 and 16.1 percent for the hybrid. The clearest failure came at the vehicle level, where leave-one-vehicle-out error averaged 0.0255 kilowatt-hours and mean R-squared collapsed to 0.194, including a negative R-squared of minus 1.140 when one vehicle was held out. A descriptive audit showed that vehicle differed substantially in starting state of charge, mean speed, and idle ratio, and the study notes that vehicle support was severely imbalanced, with one car contributing only 43 eligible windows against 2,604 for another.

The author draws conclusions that are notably narrower than the framing might invite. PGMD-Twin should be read as an offline, digital-twin-oriented estimation framework, not as a validated online digital twin, a closed-loop controller, or a universally superior predictor. Its value lies in combining a physical prior, short-horizon state dynamics, structured domain-shift testing, and uncertainty auditing in a single workflow, and in demonstrating how such components can be audited together while exposing where the framework does not generalize. Because all labeled vehicles were the same model, cross-platform transfer remains unresolved, and the study is explicit that balanced multi-vehicle data, mechanistically constrained experts, adaptive uncertainty methods, and hardware- or vehicle-in-the-loop validation are required before broader deployment claims can be supported.

For a field racing to wire artificial intelligence into every mechatronic system, the study offers a template worth copying. It shows that a deployable estimator needs more than a low average error: it needs leakage-safe inputs, physics that constrains rather than decorates, tests that deliberately break the model along every axis of real-world variation, and uncertainty intervals that are honest about their own limits. It also shows the value of publishing negative and neutral comparisons, since the simpler models that beat the proposed framework on point accuracy are reported alongside it rather than buried. As electric vehicle fleets grow and digital-twin ambitions expand, this kind of diagnostic rigor, knowing precisely where a model bends and where it breaks, may prove more valuable than any single leaderboard number.

Subject of Research: Physics-guided digital-twin energy-state estimation and domain generalization for electric vehicle mechatronic systems

Article Title: Physics-guided digital-twin-oriented energy-state estimation with meta-domain generalization for electric vehicle mechatronic systems

Article References: Xiao, W. (2026). Physics-guided digital-twin-oriented energy-state estimation with meta-domain generalization for electric vehicle mechatronic systems. Frontiers in Mechanical Engineering, 12, Article 2004324. https://doi.org/10.3389/fmech.2026.2004324

Image Credits: AI Generated

DOI: 10.3389/fmech.2026.2004324

Keywords: electric vehicles, digital twin, energy-state estimation, physics-guided machine learning, meta-domain generalization, mixture of experts, conformal prediction, state-space model, distribution shift, mechatronic systems, battery energy, vehicle telemetry

The study, published in Frontiers in Future Transportation, links household expenditure data to Consumer Price Index inflation series and gridded climate data from the India Meteorological Department. The analytical dataset covers 2,560,351 household-month observations from 195,400 unique households across 28 states and union territories, 508 harmonized districts, and 24 monthly survey periods. Households were classified as male-majority, female-majority, or balanced based on roster counts, a transparent classification rule the authors stress is not a measure of intra-household bargaining power. The researchers then applied ordinary least squares fixed-effects regressions with district and survey-month fixed effects, survey weights, state-clustered standard errors, and Benjamini-Hochberg false-discovery-rate adjustment to guard against spurious findings across many tested outcomes.

The headline regression result is deliberately counterintuitive. In the preferred rich-control specification, the female-majority coefficient for total transport expenditure share is a negligible negative 0.024 percentage points, with a false-discovery-rate q-value of 0.364, meaning it is statistically indistinguishable from the male-majority reference group. The unadjusted descriptive statistics had told a different story: female-majority and balanced households spent 2.566 and 2.637 percent of total expenditure on transport respectively, compared with 2.209 percent among male-majority households. That apparent gap, the authors conclude, largely reflects differences in income, household structure, vehicle access, and local conditions rather than gender composition itself. Male-majority households in the sample also had higher adjusted total expenditure and total income than the other groups, underscoring how easily raw spending differences can masquerade as behavioral ones.

Where gender does matter is in the composition of transport spending. Female-majority households show a 0.355 percentage-point lower fuel share, and female-headed households a 0.584 percentage-point lower fuel share, both highly robust after multiple-testing adjustment. The authors interpret this cautiously: in a country where two-wheelers and cars provide access to work, education, healthcare, and care responsibilities, lower fuel expenditure may signal constrained access to private motorized mobility rather than environmental virtue or lower vulnerability. A household that spends little on fuel may simply have no vehicle to fuel. Supporting this reading, female adult share is strongly negatively associated with fuel share, with a coefficient of negative 2.142 percentage points, while female member share is positively associated, indicating that different gender-composition measures capture distinct dimensions of household structure.

Female headship is the one gender variable that does survive rigorous adjustment for total transport burden. Female-headed households are associated with a 0.126 percentage-point higher transport-services share, a statistically precise but substantively modest estimate. The female adult share coefficient is nominally positive at 0.184 but does not survive false-discovery-rate adjustment, and the authors treat it as suggestive rather than a principal result. This pattern aligns with gender-and-poverty scholarship cautioning against relying solely on female headship while also distinguishing household composition from headship. It suggests that gendered mobility disadvantage operates through specific structural channels rather than a blanket effect of being a female-majority household.

The welfare stakes become clearer when mobility costs are set against other essential spending. A one percentage-point higher fuel share is associated with a 0.388 percentage-point lower food share, a 0.075 percentage-point lower education share, and a 0.049 percentage-point lower health share, all significant after adjustment. Total transport share is negatively associated only with education spending, at negative 0.176 percentage points, and the authors note this association is more sensitive to functional form than the fuel results. Because all these shares share a common denominator, the researchers are careful to frame the findings as budget-composition associations rather than causal displacement. Still, the consistent negative relationship between fuel burden and essentials across both share-based and log-level specifications points to genuine affordability pressure among fuel-dependent households.

Transport poverty itself is concentrated at the bottom of the income distribution, and its incidence varies sharply by household type. In the lowest income quintile, female-majority households record the highest incidence of high transport burden at 23.365 percent, followed by balanced households at 22.618 percent and male-majority households at 12.634 percent, with identical orderings for the low-income high-cost transport poverty measure. Fuel burden tells the reverse story, rising toward the top of the income distribution, where high fuel burden reaches 22.398 percent among balanced households and 21.818 percent among male-majority households in the richest quintile. The authors argue this asymmetry vindicates using multiple indicators: low-income households with high transport shares face affordability stress, while high-income households with high fuel shares reflect vehicle dependence, and each demands different policy responses.

The most policy-relevant findings concern vehicle assets and acquisition channels, which the authors treat as proxies for a household’s starting position in the coming transition. Female-majority households have an 8.4 percentage-point lower probability of owning a two-wheeler and a 1.4 percentage-point lower probability of vehicle loan borrowing, both robust after false-discovery-rate adjustment. They also show a 1.0 percentage-point lower probability of car ownership and modestly lower purchase intentions for both vehicle types. Electric-vehicle outcomes, by contrast, are too sparse in the data to support any broad conclusion about gender differences, and the authors explicitly flag that finding as exploratory. Given that two-wheelers are both the backbone of Indian motorized mobility and a major target segment for electrification incentives, this access gap means purchase subsidies, formal credit, and ownership-based benefits may flow disproportionately to households that already have the assets and borrowing pathways to use them.

Exploratory models probing climate and price pressure yielded mostly weak and mixed associations. Transport-price inflation, fuel and light inflation, days above 40 degrees Celsius, and absolute rainfall deviation showed no robust main effects on transport, fuel, or bus and train expenditure shares. The most intriguing nominal pattern, a negative interaction between female-majority status and extreme heat for transport share and a positive one for fuel share, did not survive false-discovery-rate adjustment, with q-values of 0.493 and 0.281 respectively. The authors emphasize that households are not randomly assigned to climate exposure and that expenditure shares may miss mechanisms such as altered trip timing, mode shifts, or service disruptions, which operate without producing large average expenditure changes.

The study’s broader message is that a fair transport transition requires more than electrification. Because the analysis is observational, the authors interpret all estimates as conditional associations rather than causal effects, and they acknowledge that expenditure data cannot reveal who traveled, who paid, or who sacrificed mobility when budgets tightened. Even so, the combination of lower fuel burdens, weaker two-wheeler ownership, and reduced vehicle borrowing among female-majority households paints a consistent picture of unequal starting points. The researchers argue that India’s mobility policy should pair clean-vehicle incentives with affordable and safe public transport, better first- and last-mile connectivity, regulated intermediate services such as auto-rickshaws, heat-resilient waiting environments, and credit designs that do not exclude asset-poor households. Without such measures, the transition risks reinforcing the very mobility inequalities it has the potential to resolve.

Subject of Research: Gendered mobility vulnerability, vehicle access, and transport poverty in India's sustainable transport transition

Article Title: Gendered mobility vulnerability, vehicle access, and unequal starting points for sustainable transport transition in India

Article References: Zeeshan, Z., Khosla, S., & Bhutia, P. C. (2026). Gendered mobility vulnerability, vehicle access, and unequal starting points for sustainable transport transition in India. Frontiers in Future Transportation, 7, Article 1919945. https://doi.org/10.3389/ffutr.2026.1919945

Image Credits: AI Generated

DOI: 10.3389/ffutr.2026.1919945

Keywords: gendered mobility, India, transport poverty, household expenditure, fuel burden, two-wheeler ownership, sustainable transport, electric vehicles, vehicle access, fixed-effects regression, Consumer Pyramids Household Survey, transport equity

The physics behind this achievement rests on a principle that has frustrated acoustic engineers for over a century. The mass law, a foundational rule of sound insulation, states that the transmission loss of a barrier increases by roughly six decibels for each doubling of its mass or frequency. This means that blocking low-frequency sound effectively demands either enormous weight or impractically thick walls. Conventional automotive acoustic packages, typically built from layers of dense insulation material and porous absorptive foam, obey this law faithfully. For combustion engines, which historically dominated the vehicle noise spectrum, such packages were adequate. Electric motors, however, concentrate their acoustic energy at frequencies where mass law solutions become prohibitively heavy, creating the design impasse that motivated the Shanghai team’s work.

Acoustic metamaterials offer a way around this constraint by exploiting local resonance rather than raw mass. When a structure contains embedded resonant elements tuned to specific frequencies, incoming sound waves couple to those resonators, and the coupled system exhibits effective properties that do not exist in ordinary materials, including regions of negative dynamic mass and negative stiffness. Within these regimes, a thin, lightweight panel can reflect or absorb sound at wavelengths far longer than its own thickness, something the mass law declares impossible. The concept traces back to thin-plate metamaterials, in which membranes or plates carrying small attached masses create subwavelength bandgaps that block sound transmission. Subsequent research has extended the idea through composite laminates, sandwich panels, perforated plate assemblies, and combinations of metamaterial plates with porous absorbers, each iteration seeking to widen the frequency range over which strong insulation can be maintained.

The package described in the new study, designated MRPAM-AP, embodies a further refinement: multi-frequency resonant coupling. Rather than embedding a single family of resonators tuned to one frequency, the design places three distinct resonator types, labeled A, B, and C, on a thin thermoplastic polyurethane plate. Each resonator type has its own radius, height, and characteristic length, giving it a different natural frequency. Because the three families are arranged in a rotating pattern across the plate, their individual resonance bands overlap and interact, producing a broadened attenuation region rather than a narrow notch. This coupling strategy addresses a well-known weakness of single-resonance metamaterials, which achieve dramatic transmission loss at their tuned frequency but leave neighboring frequencies largely untreated. By staggering multiple resonance frequencies within one structure, the designers created a package capable of broadband performance across the low-frequency range where electric motor noise concentrates.

The choice of materials reflects the practical demands of automotive manufacturing. Thermoplastic polyurethane serves as the structural medium for the resonant plate, chosen for its processability, damping characteristics, and suitability for mass production techniques such as injection molding. Polyurethane foam, whose porosity, flow resistivity, tortuosity, and characteristic lengths were carefully characterized, provides the absorptive layer that complements the plate’s insulating behavior. The full assembly, including a square cover and a fifteen-millimeter foam layer, remains compact enough to wrap around a motor housing, honoring the volume constraints that make this engineering problem so difficult in the first place.

To understand and predict the package’s behavior, the researchers constructed a dynamic model that departs from the treatment of ordinary homogeneous panels. A conventional barrier can be described as a single continuous system, but a plate carrying many discrete resonators behaves as a multi-degree-of-freedom system: the main structure carries one effective mass, while each attached resonator contributes its own mass, stiffness, and displacement response. The team formulated the coupled equations of motion in matrix form, capturing how an external acoustic force excites the plate and how the plate in turn drives each resonator. The equivalent stiffness of each resonator was derived from its geometry and the elastic properties of the material, while the plate itself was treated as a simply supported thin plate under Kirchhoff-Love plate theory, with a correction accounting for the contact boundaries where resonators attach. This modeling framework allowed the sound transmission loss of the assembly to be computed across the frequency spectrum using finite element simulations, and, crucially, allowed the influence of each geometric parameter to be examined individually.

That parameter sensitivity analysis revealed the central challenge of metamaterial design: performance depends on many interacting variables, and the relationships between them are strongly nonlinear. Resonator radii, heights, plate thickness, and cover dimensions each shift the resonance frequencies and coupling behavior in ways that cannot be captured by simple one-at-a-time tuning. Exhaustive numerical search across such a design space would require thousands of finite element evaluations, each computationally expensive. The researchers therefore turned to a surrogate-assisted optimization strategy combining a Kriging model with a genetic algorithm.

The Kriging component, sometimes called Gaussian process regression, builds a statistical interpolation of the simulation response from a limited set of sampled designs. It predicts transmission loss at untested parameter combinations while also quantifying the uncertainty of those predictions, which allows the algorithm to balance exploration of uncertain regions against exploitation of promising ones. The genetic algorithm then evolves a population of candidate designs, applying selection, crossover, and mutation operators inspired by biological evolution, with the Kriging surrogate standing in for the costly finite element model during most evaluations. This hybrid approach, widely used in aerospace and structural optimization, dramatically reduces the number of full simulations needed to converge on an optimal topology, making it feasible to tune the many geometric parameters of the resonant plate within a practical engineering timeline.

Validation proceeded from simulation to physical measurement. Prototype packages were fabricated and tested against an operating electric motor, with microphones capturing the radiated noise under controlled conditions. The measured spectra confirmed the predictions of the optimized design: pronounced attenuation across the targeted low-frequency band, with reductions reaching 11.2 decibels at the most affected frequencies. For context, a reduction of that magnitude is perceived as cutting perceived loudness by a substantial fraction, and it was achieved without the dense barrier layers that conventional packages would require for comparable low-frequency performance.

The broader significance of this work lies in its demonstration that acoustic metamaterials have matured from laboratory curiosities into candidates for industrial deployment. Thin-plate metamaterials had already shown promise in automotive settings, outperforming mass law predictions in prior studies, but their narrowband character limited real-world usefulness. The multi-frequency coupling strategy, combined with systematic algorithmic optimization, closes that gap by delivering broadband attenuation within a lightweight, manufacturable form. The same methodology, surrogate modeling paired with evolutionary optimization, is readily transferable to other noise control problems, from gearbox housings and inverter assemblies to cabin panels and appliance enclosures, wherever low-frequency noise must be suppressed under strict weight and space budgets.

Several avenues remain open for future development. The current design targets frequencies below 2000 Hz, matching the dominant motor noise range, but higher-order electromagnetic harmonics and mechanical vibration paths may require complementary treatments. Durability under automotive thermal cycling, vibration exposure, and long-term aging of the polyurethane components will need verification over vehicle lifetimes. Manufacturing tolerances matter as well, since metamaterial resonance frequencies are sensitive to small geometric deviations, and production-scale molding must hold dimensions tightly enough to preserve the tuned response. Nonetheless, the study establishes a practical workflow, from physical modeling through surrogate-based optimization to experimental confirmation, that others in the field can follow, and it signals that the quiet electric vehicle of the future may owe its silence not to heavier insulation but to cleverly engineered internal geometry.

Subject of Research: Optimization of multi-frequency resonant coupling acoustic package for electric motor based on a Kriging-genetic algorithm

Article Title: Optimization of multi-frequency resonant coupling acoustic package for electric motor based on a Kriging-genetic algorithm

Article References: Wang, J., Sun, B., Guo, H., Sun, P., Huang, S., Xie, X., Liu, N., & Yao, M. (2026). Optimization of multi-frequency resonant coupling acoustic package for electric motor based on a Kriging-genetic algorithm. Frontiers in Materials, 13, 1980178. https://doi.org/10.3389/fmats.2026.1980178

Image Credits: AI Generated

DOI: 10.3389/fmats.2026.1980178

Keywords: acoustic metamaterials, electric vehicles, noise control, Kriging-genetic algorithm, sound transmission loss, electric motors, Optimization, multi-frequency, resonant, coupling, acoustic, package

The core of the innovation lies in the hybridization of two distinct metaheuristic algorithms. Particle Swarm Optimization is known for its rapid convergence but is prone to getting trapped in local optima, while Grey Wolf Optimization offers robust global exploration but can be slower in refining solutions. By integrating the hierarchical leadership mechanism of GWO with the velocity-update mechanism of PSO, the researchers created a unified optimizer that balances global exploration with local exploitation. This hybrid approach allows the algorithm to navigate complex, nonlinear search spaces more effectively, identifying optimal controller gains that minimize frequency deviations and tie-line power oscillations in interconnected power areas.

The study models an unequal two-area power system where Area 1 consists of thermal, hydro, and electric vehicle units, while Area 2 integrates wind, photovoltaic, diesel, and hydrogen-based energy storage. This configuration reflects the complexity of modern grids, where diverse generation sources and storage technologies must work in concert to maintain balance. The researchers used MATLAB/Simulink to simulate the system, incorporating transfer function models for each component, including an aqua electrolyzer and fuel cell for hydrogen storage. The objective was to minimize the Integral of Time-weighted Absolute Error, a metric that penalizes prolonged frequency excursions, which are operationally damaging to grid stability.

When tested under step load disturbances in Area 1, the hybrid GWO-PSO algorithm demonstrated remarkable performance. It achieved the lowest objective value with an ITAE of 0.29888, representing a 71% reduction compared to the PSO-only approach and a 28.1% reduction compared to the Slime Mould Algorithm. In terms of time-domain response, the hybrid controller reduced the frequency settling time to 4.6 seconds, outperforming both PSO and SMA. The frequency in Area 1 remained between 55.5 Hz and 62.5 Hz, while Area 2 exhibited the lowest peak frequency of 60.5 Hz among all tested optimizers, indicating superior damping of inter-area coupling effects. These results suggest that the hybrid algorithm can more effectively manage the dynamic interactions between different power sources and storage units.

The integration of hydrogen storage via an aqua electrolyzer and fuel cell played a significant role in the system’s stability. During grid imbalances, the electrolyzer acts as a controllable load, absorbing excess power to produce hydrogen, which can later be converted back to electricity by the fuel cell. Under the GWO-PSO tuning, the electrolyzer delivered a peak hydrogen production of 1.7 × 10⁻³ kg/s within two seconds of the disturbance, providing rapid grid support. This dual-phase behavior, characterized by a high-amplitude transient response followed by sustained steady-state production, effectively combines the fast initial response of PSO-optimized systems with the stable long-term output of SMA-optimized systems, enhancing the overall resilience of the power system.

Electric vehicle units were also included as a fast-acting resource in the control loop. The study found that EVs have a material impact on steady-state frequency regulation, helping to eradicate persistent frequency offsets that other storage technologies might not address. An ablation study revealed that disabling the EV units led to a non-decaying offset in Area 1 frequency, resulting in a significantly higher ITAE value. This highlights the importance of treating EV fleets as first-class resources in AGC design, rather than merely as secondary loads. The hybrid controller effectively coordinated the EVs and hydrogen storage to ensure that the system returned to nominal frequency quickly and accurately, even under significant disturbances.

Further analysis under simultaneous step load disturbances in both areas showed that the GWO-PSO controller maintained robust performance. The frequency settling time was reduced to 5.0 seconds, compared to 7.0 seconds for PSO and 6.0 seconds for SMA. The tie-line power settled within 3.8 seconds, demonstrating the controller’s ability to manage inter-area power flows efficiently. The hybrid algorithm consistently selected higher proportional and integral gains than PSO, indicating that it identifies more aggressive yet stable control actions. This allows the system to recover swiftly without sacrificing steady-state accuracy, which is crucial for maintaining reliability in renewable-dominated grids.

Sensitivity analyses confirmed the robustness of the proposed controller under varying conditions. When renewable penetration and system inertia were perturbed by ±20%, the controller remained stable, with performance declining gradually rather than suddenly. The system was found to be more sensitive to inertia variations than to load-damping uncertainties, but it handled both types of perturbations effectively. Additionally, the controller showed low sensitivity to load disturbance amplitudes, with ITAE values varying by less than 1% across different disturbance levels. This consistency suggests that the hybrid GWO-PSO approach is well-suited for real-world applications where grid conditions are constantly changing.

The computational complexity of the hybrid algorithm was also evaluated. A full optimization run took approximately 2.21 hours, dominated by the cost of simulation rather than algorithmic overhead. However, since this tuning is performed offline, it does not impact real-time AGC operations. The convergence characteristics of the hybrid algorithm were superior to its components, achieving the lowest final ITAE while maintaining a convergence speed comparable to GWO alone. This indicates that the hybridization provides genuine benefits in solution quality without compromising convergence speed, making it a practical tool for grid operators seeking to enhance frequency regulation in complex, multi-source power systems.

In conclusion, the study demonstrates that hybrid metaheuristic tuning offers a viable path toward more resilient frequency regulation in renewable-integrated power systems. By combining the strengths of GWO and PSO, the proposed algorithm effectively optimizes PID controller gains to handle the dynamic complexities of modern grids. The integration of diverse generation sources and storage technologies, along with the careful tuning of control parameters, ensures that the system can maintain stability and efficiency even under significant disturbances. This research provides valuable insights for grid operators and engineers working to integrate higher percentages of renewable energy into the power grid, offering a practical and deployable solution for maintaining grid stability in an increasingly complex energy landscape.

Subject of Research: Optimization of automatic generation control in renewable-integrated power systems using a hybrid GWO-PSO algorithm

Article Title: Optimization of automatic generation control in renewable-integrated power systems using a hybrid GWO–PSO algorithm

Article References: Uddin, M. I., Imam, A. A., Alabdullah, F. T., Messini, E. M. B., Alyamani, N., & Almohammedali, M. A. (2026). Optimization of automatic generation control in renewable-integrated power systems using a hybrid GWO–PSO algorithm. Results in Engineering, 32, Article 113439. https://doi.org/10.1016/j.rineng.2026.113439

Image Credits: AI Generated

DOI: 10.1016/j.rineng.2026.113439

Keywords: automatic generation control, renewable energy, power systems, hybrid optimization, PID controller, frequency stability, hydrogen storage, electric vehicles, metaheuristic algorithms, grid resilience, Optimization, automatic

The first major contribution of the research is a fuzzy-adaptive maximum power point tracking algorithm. Traditional methods often struggle to maintain optimal efficiency when solar irradiance changes rapidly due to cloud cover or other atmospheric factors. The new algorithm dynamically adjusts its perturbation step based on the local sensitivity of power to voltage. This adaptive approach allows the system to track the maximum power point with greater accuracy and speed. In tests across fifteen randomized irradiance scenarios, the algorithm achieved a tracking efficiency of 98.1 percent, with a margin of error of only 0.1 percent. This performance significantly outperforms fixed-step methods, which averaged 95.0 percent efficiency, and linear-adaptive laws that required re-tuning for each specific scenario.

Statistical analysis confirms the superiority of the fuzzy-adaptive method. A Welch’s t-test comparing the fuzzy-adaptive algorithm against a re-tuned linear-adaptive law yielded a t-value of 5.2 with a p-value less than 0.0001. This indicates that the improvement in tracking efficiency is not due to chance but is a statistically significant result of the adaptive control strategy. The ability to maintain high efficiency with a single fixed parameter set is crucial for practical deployment, as it eliminates the need for complex recalibration procedures in the field. This robustness makes the technology more viable for widespread adoption in commercial charging stations where maintenance and tuning resources may be limited.

The second contribution focuses on the accurate estimation of the battery state of charge, a critical parameter for managing energy storage in charging stations. The researchers employed an extended Kalman filter to fuse data from Coulomb counting with a second-order Thevenin battery model. This fusion technique leverages the strengths of both methods to provide a more reliable and precise estimate of the remaining energy in the battery. The observability of this estimation pair was verified using a Popov–Belevitch–Hautus test, ensuring that the system’s internal states can be accurately inferred from its outputs. Across twenty Monte Carlo trials, this method reduced the mean root mean square error of the state of charge estimation from 2.45 percent to just 0.118 percent.

Such a dramatic reduction in estimation error has profound implications for the longevity and safety of battery storage systems. Accurate state of charge data allows the energy management system to avoid deep discharges or overcharging, both of which can degrade battery health over time. By maintaining the battery within its optimal operating range, the system can extend the lifespan of the storage units, reducing long-term operational costs. Furthermore, precise state of charge information enables the charging station to better predict its available capacity, allowing for more reliable service to electric vehicle users. This level of precision is essential for integrating battery storage into the grid, where accurate forecasting of available energy is required for grid stability and planning.

The third contribution involves the implementation of a model predictive control energy management system. This system replaces static, rule-based dispatch strategies with a receding-horizon optimization approach. Model predictive control allows the system to anticipate future energy demands and supply conditions, optimizing the flow of energy between the solar array, battery storage, grid, and electric vehicle in real-time. Under the assumption of a perfect forecast, this approach lowered the operating cost of the charging station by 6.8 percent. However, the study also reveals an explicit trade-off between cost optimization and grid independence. While the system can minimize costs by leveraging grid power when it is cheap, it can also prioritize solar and battery usage to reduce reliance on the grid, depending on the specific operational goals of the station.

The fourth contribution addresses the stability of the system’s power electronics. The researchers conducted a switched-system stability analysis, constructing mode-dependent Lyapunov functions to analyze the dynamics of the power converters. They derived an average-dwell-time bound of 2.31 milliseconds, which certifies the exponential stability of the linearized converter dynamics for any mode-switching sequence that respects this bound. This theoretical guarantee is crucial for ensuring that the system remains stable and does not oscillate or fail when switching between different operating modes, such as charging, discharging, or grid interaction. The use of Lyapunov stability theory provides a rigorous mathematical foundation for the control design, ensuring that the system behaves predictably and safely under all expected operating conditions.

When all components are integrated, the complete simulation of the charging station demonstrates a system efficiency of 93.6 percent. The DC bus voltage is regulated within a tight tolerance of plus or minus 5 percent, ensuring stable power delivery to the electric vehicle. The system achieves a grid dependency of 14.2 percent, indicating that the majority of the energy used for charging is sourced from the local solar array and battery storage. This high level of self-sufficiency reduces the load on the electrical grid and lowers the carbon footprint of the charging process. The ability to maintain high efficiency and stability while significantly reducing grid dependency makes this technology a promising candidate for sustainable urban energy infrastructure.

To validate the theoretical findings, the researchers implemented the fuzzy-adaptive MPPT and extended Kalman filter state of charge layers on an ESP8266 hardware prototype. The hardware-measured system efficiency was 91.2 percent, which agrees with the simulation result of 93.6 percent to within 2.4 percentage points. This close agreement between simulation and hardware testing provides strong evidence that the control algorithms are effective in a real-world environment. However, the study notes that the model predictive control dispatch results and the Lyapunov stability proof remain simulation-level findings. Hardware validation of these components is identified as a priority for future work, as it will be necessary to confirm that the theoretical stability guarantees hold in physical implementations with real-world noise and component variations.

This research represents a significant step forward in the development of smart, solar-integrated electric vehicle charging stations. By combining advanced control techniques with rigorous stability analysis, the authors have demonstrated a system that is both efficient and reliable. The ability to accurately track solar power, estimate battery state, and manage energy flow in real-time addresses key challenges in the integration of renewable energy with electric mobility. As the adoption of electric vehicles continues to grow, the need for efficient and sustainable charging infrastructure becomes increasingly important. This work provides a robust framework that can be adapted and scaled to meet the demands of a rapidly evolving energy landscape, contributing to the broader goal of decarbonizing the transportation sector.

Subject of Research: Control systems for solar-integrated electric vehicle charging stations

Article Title: Fuzzy-adaptive MPPT and Kalman-filter SOC estimation for Lyapunov-stable predictive energy management in solar-integrated EV charging stations

Article References: Santhoshkumar, R., Suresh, S., Peramanan, A., & Sathishkumar, P. (2026). Fuzzy-adaptive MPPT and Kalman-filter SOC estimation for Lyapunov-stable predictive energy management in solar-integrated EV charging stations. Scientific Reports. https://doi.org/10.1038/s41598-026-75396-1

Image Credits: AI Generated

DOI: 10.1038/s41598-026-75396-1

Keywords: solar energy, electric vehicles, energy management, Kalman filter, fuzzy logic, battery storage, power electronics, model predictive control, stability analysis, renewable energy, Fuzzy-adaptive, MPPT

The research, led by Yantao Lou and Jing He with colleagues Tonghui Wang and Ruiqian Liu, tackles the load frequency control (LFC) problem in a multi-microgrid system (MMS). In such systems, several microgrids are linked through tie lines, and power disturbances in one network propagate through the connections, superimposing on the random fluctuations already present in its neighbors. The frequency dynamics become strongly nonlinear and substantially harder to control. Electric vehicles, which can act as mobile energy storage through vehicle-to-grid (V2G) capability, add flexible regulation resources but also new uncertainties: their state of charge and the randomness of user behavior complicate the control problem further.

Traditional LFC approaches, from PID controllers to fuzzy logic and model predictive control (MPC), rely on explicit models of the system. They work well when the model matches reality, but under strong stochastic and intermittent disturbances they can suffer from model mismatch and limited adaptability. Machine learning methods, particularly deep reinforcement learning (RL), offer an alternative: agents learn control policies by interacting with the environment. Deep Q-Networks (DQN) have been applied to isolated microgrid frequency control with discrete action spaces, while Deep Deterministic Policy Gradient (DDPG) extends this to continuous control. But single-agent approaches rely on centralized information and decision-making, which is less suitable for a multi-microgrid system requiring local autonomy and privacy preservation, and may introduce communication bottlenecks and single-point failures.

Multi-agent reinforcement learning (MARL) addresses these limitations by giving each microgrid its own agent, achieving distributed coordinated control. Yet MARL brings its own problems: Q-value overestimation, inefficient early exploration, and slow convergence. The new study’s key insight is that the physical constraints and coordination information embedded in expert experience can be systematically exploited to improve training efficiency and policy stability. Rather than using model information only as auxiliary guidance, the researchers let MPC generate high-quality, constraint-satisfying expert trajectories through model-based rolling optimization, then use these demonstrations to guide the learning process.

The proposed framework combines a model of the multi-microgrid system—diesel generators and EV clusters as controllable frequency-regulation resources, photovoltaic generation and local loads as exogenous disturbances—with an MPC controller that solves a constrained optimization problem at each sampling instant. The MPC generates control sequences through receding-horizon optimization, explicitly considering constraints on control inputs, control increments, and system outputs. These state-action pairs are stored as an expert knowledge library, providing physically interpretable and constraint-satisfying prior knowledge for the subsequent MARL training.

The learning architecture is a Multi-Agent Twin Delayed Deep Deterministic Policy Gradient (MA-TD3), which mitigates Q-value overestimation through dual-critic networks, nonsmooth control actions through target policy smoothing, and policy oscillations through delayed policy updates. Each agent’s state vector includes the current and previous frequency deviations of all interconnected microgrids, the current and previous tie-line power variations, and the current and previous net source-load disturbances of its own microgrid. The action is a three-dimensional continuous vector corresponding to the regulation commands of the DG and two EV clusters in the microgrid. The reward function integrates local frequency regulation, inter-microgrid coordination, and penalties for excessive frequency deviations, with episode-dependent exploration and collaboration factors that gradually shift the agents from local stabilization to inter-microgrid coordination.

The training strategy is phased. In the imitation learning stage, agents learn state-action mappings from the MPC expert knowledge through behavior cloning, without environmental interaction. In the hybrid training stage, environmental interaction is introduced based on the policy obtained from imitation learning, with a proportion of expert experiences preloaded into the experience replay buffer. As environmental interaction data accumulate, their relative proportion increases while the influence of expert experiences gradually decreases. The agents transition smoothly from expert guidance to autonomous optimization, mitigating abrupt policy changes and reducing the risk of training instability and policy forgetting.

Simulation results on a two-microgrid validation platform, with parameters based on operational data from actual microgrids, demonstrate the effectiveness of the approach. Under a random load disturbance scenario, the proposed method achieves the lowest mean absolute frequency deviation, maximum absolute frequency deviation, and frequency-deviation standard deviation of 0.0019585 Hz, 0.0053834 Hz, and 0.0019452 Hz respectively, with a pass rate of 99.010 percent. Under a two-sided source-load random disturbance scenario, it again achieves the lowest values of the three deviation indices and a pass rate of 99.519 percent. In a DG fault exit scenario simulating the sudden loss of key frequency regulation resources, the proposed method limits the frequency deviation to approximately 0.005 Hz, achieving a pass rate of 99.704 percent. In scenarios where EV regulation capability is restricted to charging only or subject to time-varying limits, the proposed method consistently outperforms comparison methods including MA-TD3, MA-A2C, MPC, behavior cloning, and fixed-ratio expert replay.

A multi-seed statistical analysis across five independent training runs confirms the consistency of the proposed method. Across the five operating scenarios, it achieves mean absolute frequency deviations of 0.001667 to 0.002200 Hz, maximum absolute frequency deviations of 0.005793 to 0.007209 Hz, and pass rates of 97.2910 to 99.4330 percent, with generally lower across-seed variability than the comparison methods. A preliminary extension to a three-microgrid configuration, with each agent’s observation dimension increased from 8 to 10, maintains coordinated frequency regulation under disturbance, providing evidence that the framework can scale beyond the two-microgrid case.

The study acknowledges limitations: the simulations assume ideal and reliable communication links among microgrids, so communication delays, measurement noise, and packet loss are not explicitly considered. Future research will incorporate these uncertainties and investigate robustness under nonideal communication conditions. The researchers also note that because constraints are enforced during expert-data generation rather than directly on the learned policy during online execution, the guidance does not provide a formal safety guarantee during deployment; mechanisms such as distributional safety critics and risk-constrained policy optimization could be incorporated as complementary safety layers. Nevertheless, the framework demonstrates that systematically exploiting expert knowledge—rather than using it only as auxiliary guidance—can improve the efficiency and stability of multi-agent learning for coordinated frequency control in renewable-rich power systems.

Subject of Research: Reinforcement learning guided by model predictive control keeps multi-microgrid frequencies stable under renewable-rich conditions.

Article Title: Coordinated frequency control of multi-microgrids: A phased MARL guided by constraint-informed expert knowledge

Article References: Lou, Y., Wang, T., Liu, R., & He, J. (2026). Coordinated frequency control of multi-microgrids: A phased MARL guided by constraint-informed expert knowledge. Results in Engineering, 32, Article 113255. https://doi.org/10.1016/j.rineng.2026.113255

Image Credits: AI Generated

DOI: 10.1016/j.rineng.2026.113255

Keywords: microgrid, reinforcement learning, model predictive control, frequency stability, multi-agent systems, electric vehicles, renewable energy, load frequency control, Coordinated, frequency, control, multi-microgrids

The research emerges from a structural shift in power systems. International climate accords have driven national pledges to replace fossil-fuel infrastructure with renewables, and distributed energy architectures are replacing centralized utility models. Placing generation close to consumption reduces transmission losses and enhances grid resilience, while commercial entities and residential consumers become active market participants through rooftop photovoltaic installations, battery energy storage systems and small-scale wind turbines. Electric vehicle charging stations equipped with solar arrays can even bid strategically into electricity markets. In this landscape, peer-to-peer trading allows prosumers to exchange energy directly, bypassing conventional utility transactions—but it also raises questions about benefit sharing that economic efficiency alone cannot answer.

Those questions belong to the framework of energy justice, which analyzes fairness in the energy sector across distributive, recognition and procedural dimensions. As distributed generation penetrates deeper into grids, distributive justice—how benefits and burdens are shared among participants—becomes increasingly prominent. The authors note that unfavorable tariff structures and complicated interconnection requirements substantially reduce the economic feasibility of residential solar installations, and prosumers are often paid significantly less for grid exports through feed-in tariffs, creating a pricing disparity that deters investment in distributed energy resources. A well-designed decentralized market, they argue, can mitigate these barriers and foster a more equitable environment for small-scale renewable investment.

The study’s central contribution is the Adaptive Fairness-Based Allocation (AFBA) framework, which embeds fairness as a core objective rather than a post-hoc adjustment. The method works at the market layer of a five-node community microgrid, where a solar-equipped prosumer shares surplus energy with three buyers: an industrial consumer with a 300 kWh demand cap bidding ₹5.90/kWh, a commercial EV charging station capped at 250 kWh bidding ₹7.15/kWh, and a domestic consumer capped at 200 kWh bidding ₹5.58/kWh. Surplus energy is structured into six four-hour slots per day, and any energy left after all buyers are served is exported to the grid at a feed-in tariff of ₹4.5/kWh.

Technically, AFBA computes each buyer’s participation factor as the ratio of its weighted bid to the total weighted bids of all buyers, with weights updated adaptively from historical allocation outcomes. After every surplus slot, the framework stores one cumulative deviation per buyer—the gap between that buyer’s actual allocation and the equal-share level—using exponential memory decay with a factor of 0.95, so shortfalls in one slot can be corrected in subsequent slots. A bounded nonlinear mapping prevents stored deviations from producing oversized weight jumps, while a learning rate of 0.1, a compensation factor of 0.2 providing momentum, a nonlinearity factor of 1.2, and a smoothing factor of 0.5 stabilize the update. A minimum weight constraint of 0.1 ensures no buyer is ever completely excluded, and weights are normalized to sum to unity before each allocation.

The authors benchmarked AFBA against five alternatives: fixed contractual allocation, priority-first allocation with proportional redistribution, a price-weighted fairness method, linear programming (LP), and a genetic algorithm (GA). LP maximizes slot revenue subject to power balance and demand constraints and serves as the revenue benchmark; GA searches for near-optimal solutions stochastically, with results averaged over 20 independent runs. The evaluation used real-world data: solar generation from a private plant in Kuzhikattussery, Thrissur, Kerala, and EV charging demand from Perth & Kinross Council’s charging stations in Perth, UK, covering 2024 at 15-minute intervals, aggregated into 2,190 calendar slots of which 646 were surplus slots totaling 154,309.89 kWh of available energy.

The results reveal a stark trade-off. Under fixed pricing, LP achieved the highest annual revenue of ₹10.49 lakhs but the second-worst fairness at 0.4409, concentrating 72.74 percent of annual energy on the highest bidder while the domestic buyer received just 1.8 percent. The priority-first method fared even worse on fairness at 0.4162, starving the lowest-bid buyer to 3.98 percent of yearly allocation. AFBA, by contrast, delivered almost perfectly balanced yearly allocations of 33.33, 33.27 and 33.13 percent, a fairness index of 0.9995, value fairness of 0.9877, and revenue of ₹9.58 lakhs—only 9.0 percent below the LP benchmark. The price-weighted method offered the best revenue among fair-oriented approaches at ₹9.69 lakhs with fairness of 0.9882, while GA reached 0.9935 fairness but required roughly 28.6 milliseconds per slot on average, far too slow for time-critical clearing.

Computation time proved a decisive differentiator. AFBA completed each allocation in 0.0255 milliseconds under fixed pricing and 0.1467 milliseconds under time-of-day (ToD) pricing, while the price-weighted method ran in 0.0093 to 0.0183 milliseconds. LP required around 1.18 milliseconds and GA between 24.5 and 28.6 milliseconds. The authors also tested AFBA’s resilience under ToD tariffs derived from Kerala State Electricity Regulatory Commission regulations and Ministry of Power guidelines, applying multipliers that discount bids by 20 percent during solar hours and surcharge them by 10 percent at night. Fairness performance remained essentially unchanged under dynamic pricing, and the fairness progression analysis showed AFBA improving by roughly +0.4 percent over successive surplus slots, confirming that the system genuinely learns from experience—a property absent in static or optimization-only methods.

Scalability testing reinforced the practical promise. When the number of buyers was increased from 3 to 100, average fairness rose from 0.931 to 0.984 and stabilized, while per-slot computation time stayed between 0.087 and 0.192 milliseconds. On the same machine, LP took about 1.18 milliseconds and GA about 31 milliseconds per slot with three buyers. The authors acknowledge limitations: the allocation operates only at the market layer, ignoring voltage limits, line ratings and power-flow equations, and the generation and demand profiles come from two independent regional datasets rather than a physically co-located microgrid. Equal energy allocation, they note, is also inappropriate when buyers have contracts of different grades, non-deferrable loads, or significantly varying demand constraints—in such cases a fill-rate criterion or priority rule serves better.

The study’s implications reach beyond the test system. Existing research had achieved at most a Jain fairness index of 0.94, lacked frameworks to track cumulative fairness deviations with memory decay, and had not integrated ToD tariffs for distribution fairness under real regulatory frameworks. By raising the yearly index to 0.9995 at the cost of an 8.8 to 9.0 percent revenue gap relative to unconstrained revenue maximization, AFBA quantifies precisely what equity costs in a local energy market—and shows that the price is modest. For community microgrids deciding how to split their solar surplus, the message is clear: long-term equality among participants is achievable in microseconds, and markets that prioritize it may retain the trust and participation needed to sustain the decentralized energy transition.

Subject of Research: Energy

Article Title: Achieving equitable and efficient energy trading in peer-to-peer microgrids: An evaluation of allocation strategies

Article References: Sreekumar, A. V., & Lekshmi, R. R. (2026). Achieving equitable and efficient energy trading in peer-to-peer microgrids: An evaluation of allocation strategies. Results in Engineering, 32, Article 113537. https://doi.org/10.1016/j.rineng.2026.113537

Image Credits: AI Generated

DOI: 10.1016/j.rineng.2026.113537

Keywords: peer-to-peer energy trading, microgrids, energy justice, fairness algorithms, solar power, electric vehicles, distributed energy resources, Achieving, equitable, efficient, energy, trading