A peer-reviewed study has found that hybrid vehicle-to-grid (V2G) scheduling algorithms offer no advantage over simpler Adaptive Particle Swarm Optimization (APSO), challenging vendor claims in the smart charging market.
Researchers from the University of Engineering & Technology Lahore, the University of Management and Technology Lahore, the University of Johannesburg, and the University of Botswana tested seven optimization algorithms for scheduling electric vehicle (EV) charging and discharging in V2G systems. Their results showed that combining two algorithms into a hybrid does not automatically outperform the stronger individual algorithm.
The study, published in the journal Processes, evaluated the algorithms using a simulated smart parking facility. EVs arrived and departed within a scheduling window, each with unique battery capacities and driver-preferred charge levels, under a time-of-use electricity pricing scheme with peak, mid-peak, and off-peak periods.
Particle Swarm Optimization (PSO) and the Whale Optimization Algorithm (WOA) are metaheuristic algorithms that search for near-optimal solutions by mimicking natural processes. PSO models a swarm of candidate solutions influenced by their own and the group's best-known positions, while WOA is inspired by the bubble-net hunting behavior of humpback whales, alternating between encircling and spiraling movements.
APSO, a refined version of PSO, continuously adjusts its internal search behavior, tightening or loosening its search pattern based on how clustered or spread out its candidate solutions are at each step, rather than following a fixed pattern.
APSO outperformed the other algorithms, improving on standard PSO by 8% and on the PSO-WOA hybrid by 9.7%. It also produced a flatter peak load demand profile across the simulated charging facility. The PSO-WOA hybrid did not perform better than plain PSO, and an enhanced version of WOA showed no significant improvement over the standard WOA.
Of the seven algorithms tested, only APSO's advantage remained consistent after rigorous statistical testing across 100 simulation trials for each algorithm. The researchers used parametric tests such as t-tests and ANOVA, as well as non-parametric tests like Mann-Whitney U, Wilcoxon Signed-Rank, and Friedman tests, along with post-hoc corrections including Holm's Step-Down, Bonferroni-Dunn, and Nemenyi analyses to ensure genuine performance improvements.
The paper notes that this level of statistical validation is uncommon in existing V2G scheduling literature, where most studies report improvements without confirming their significance through statistical testing.
The researchers conclude that APSO's advantage comes from its adaptive control of internal search parameters, which adjust based on the distribution of the search population at each iteration, rather than from hybridizing two algorithms. This distinction is important for vendors marketing hybrid or multi-algorithm optimization as inherently superior to a single, well-tuned approach.
The findings come as V2G and smart charging platforms increasingly promote algorithmic and AI-driven capabilities as commercial differentiators. For example, Geely claims a 20% improvement in battery lifecycle with its pulse-restoration technology, but these figures are based on internal testing rather than independent statistical validation like that used in the Processes study.
The commercial importance of effective scheduling is growing as more automakers commit to V2G globally. In August 2026, Hyundai Motor Group and energy intelligence platform Kaluza launched a partnership to integrate smart charging into Kia and Hyundai apps in the UK and Australia, with V2G services planned from 2027. Kaluza CEO Stephen Fitzpatrick described the partnership as foundational for broader EV integration with the energy system.
The real-world value of such claims depends on how well the scheduling software performs against driver preferences and grid price signals under live conditions, precisely the kind of question the Processes paper argues should be tested independently rather than relying on vendor-reported results.
One of the most rigorously validated examples in the market comes from a randomized controlled trial by the Centre for Net Zero, a research institute founded by Octopus Energy. Over 12 months, more than 13,000 UK households participated, and AI-managed charging reduced peak household electricity use by 42% and annual bills by £650 (US$887), with over half of households never manually overriding the automated system.
This trial, which included a control group and four treatment groups tracked over a year using high-frequency smart meter data, provides a model for the kind of evidence base the Processes researchers argue is largely missing from the algorithmic V2G literature.
Not all claims in the smart charging market have undergone such scrutiny. The Processes study itself has limitations, as it excludes battery degradation costs and charging-efficiency losses, and treats the parking facility as a single grid connection rather than modeling feeder- or transformer-level constraints. The researchers acknowledge that including these factors would likely yield more conservative cost estimates and suggest this as a direction for future research.
Ultimately, the paper's main contribution is methodological, demonstrating how to rigorously test competing algorithms rather than providing a ready-made blueprint for production scheduling software. It shows the importance of thorough validation over simply assuming that hybrid or complex algorithms are superior.
.svg)
