Integrating renewable generation into modern electricity networks requires reconciling the sharp fluctuations of weather-dependent resources with the shifting power requirements of daily human activity. Solar photovoltaic systems generate electricity strictly during daylight hours, often producing power curves that mismatch residential and commercial consumption patterns. In distribution networks where renewable penetration continues to expand, managing the imbalance between peak power demand and local solar generation represents an ongoing operational hurdle. When demand surges during evening hours after sunset, grid infrastructure must absorb steep load spikes that threaten system reliability, inflate generation costs, and strain distribution feeders. Electrochemical battery energy storage systems offer an adaptable technological remedy by absorbing surplus clean electricity during midday generation peaks and dispatching stored energy when network demand escalates. However, realizing the full potential of these decentralized assets depends heavily on coordinated charge and discharge scheduling strategies that navigate complex physical limits and variable power profiles across multi-hour operating horizons.
To address this coordination problem, researchers Mohamed Sayed Ibrahim, Sawsan Morkos Gharghory, and Hanan Ahmed Kamal developed a novel computational framework that schedules battery storage operation within a grid-connected photovoltaic microgrid. Reported in the Journal of Electrical Systems and Information Technology, their study introduces a hybrid metaheuristic approach designated as the HGPO algorithm. The technique blends the exploratory strengths of the Grey Wolf Optimizer with the localized exploitation capabilities of Particle Swarm Optimization to govern battery operation across a twenty-four-hour scheduling period. Validating the algorithm against historical demand and solar generation records collected from a distribution substation in the southern United Kingdom, the research team demonstrated that coordinated scheduling achieves daily peak load reductions between 21.67 percent and 28.06 percent. Compared to established single and hybrid heuristic methods documented in engineering literature, the proposed formulation improved optimization performance metrics by 1 to 3 percent while achieving faster numerical convergence and lower overall objective cost values.
The physical setting investigated in the study reflects an increasingly common configuration in modern electrical distribution systems: a hybrid microgrid where a five-megawatt solar photovoltaic farm operates alongside a stationary battery storage unit and a dedicated residential feeder. Both the generation asset and the local demand center connect to a common distribution substation, which maintains an active tie to the broader municipal utility network. This grid interconnection serves as an essential backup path, ensuring continuous power delivery during periods when combined solar and battery output cannot satisfy total community consumption. Power consumption across the residential feeder follows a classic daily duck curve, characterized by moderate demand during morning working hours, an afternoon lull, and an intense evening surge that coincides with the complete cessation of solar photovoltaic production. Conversely, photovoltaic generation exhibits non-uniform, weather-dependent daytime output. This operational dichotomy creates two distinct analytical windows across each twenty-four-hour day: a daytime interval dedicated to localized consumption and battery charging, and an evening window requiring managed storage discharge.
Formulating the battery scheduling challenge required establishing a rigorous mathematical optimization model evaluated across a twenty-four-hour horizon divided into forty-eight discrete thirty-minute intervals. The primary objective function seeks to minimize the maximum net power drawn from the utility grid by optimizing the active power exchanged with the battery system alongside available solar output and consumer demand. Mathematically, the decision variables represent the battery active power output at each half-hour step, designated as negative when energy flows into the storage cells during charging and positive when power discharges into the local network. The objective function identifies and suppresses the highest grid demand point across all forty-eight time intervals while strictly enforcing physical boundary conditions. These governing constraints dictate that charging power combined with directly delivered solar energy must never exceed total instantaneous photovoltaic generation, and that discharging power plus direct solar feed cannot exceed instantaneous consumer demand. Furthermore, the model tracks cumulative battery state of charge across successive intervals, enforcing strict upper and lower capacity boundaries set between zero and five megawatts.
Solving this non-linear, multi-interval scheduling problem effectively demands robust optimization algorithms capable of navigating multidimensional search spaces without becoming trapped in suboptimal local minima. Metaheuristic optimization approaches draw structural inspiration from natural biological behaviors to explore candidate solutions across complex engineering surfaces. Among these techniques, the Grey Wolf Optimizer simulates the social leadership hierarchy and collaborative hunting mechanisms observed in natural wolf packs. The population divides into four tiers: alpha wolves representing the premier candidate solutions that steer decision timing and direction, subordinate beta wolves providing advisory oversight, tertiary gamma wolves, and trailing omega individuals that follow the dominant pack members. The mathematical formulation tracks encircling and hunting dynamics through directional vectors and distance calculations, using parameter variations to alternate between broad global exploration and focused local exploitation. In parallel, Particle Swarm Optimization emulates the coordinated motion of flocking birds, evaluating swarms of particles where individual velocities and trajectory updates integrate both personal best historical positions and global swarm achievements.
While both standalone algorithms exhibit practical utility across diverse engineering applications, individual metaheuristics frequently encounter computational bottlenecks. Particle Swarm Optimization offers flexible exploitation but remains vulnerable to premature convergence in highly constrained problem landscapes, whereas standard Grey Wolf Optimization provides reliable exploration but can struggle with slower convergence trajectories near final solutions. Earlier hybrid approaches in literature attempted to merge these frameworks by replacing the historical cognitive and social reference points of particle swarms directly with the coordinates of the three leading wolves. The research team identified structural limitations in this static substitution strategy, noting that simply transplanting position coordinates does not fully exploit the complementary dynamic search strategies of the underlying algorithms. In response, their proposed HGPO formulation introduces a dynamic switching architecture that actively monitors population progress and selectively invokes the particle swarm mechanism only when the primary grey wolf search reaches an exploratory impasse.
Under the HGPO execution sequence, the optimization process begins by generating random initial spatial positions for an entire pack of candidate wolves, subsequently evaluating their performance against the objective function to designate alpha, beta, and gamma leaders. The algorithm then tracks the position coordinates of the alpha wolf across successive iterations. If the alpha solution exhibits no measurable improvement over a predetermined number of consecutive computational cycles, the algorithm recognizes an exploratory stall and temporarily transfers the population into the Particle Swarm Optimization framework. In this secondary phase, the swarm adopts the current wolf positions as its particle coordinates while establishing the static alpha solution as its global target. The swarm executes a bounded series of iterations, leveraging particle velocity equations and self-confidence memory parameters to escape the local plateau. Once these iterations conclude, control returns to the grey wolf mechanism, which assesses whether the swarm discovered improved coordinates to replace the prevailing beta and gamma positions before continuing the standard encircling and parameter update equations.
The research team evaluated the computational performance of this alternating architecture using empirical demand and generation records gathered from the southern United Kingdom substation between November 3, 2017, and December 17, 2019. The evaluation examined two contrasting seasonal operating days to assess performance under divergent meteorological conditions. The first scenario represented a low-generation day where aggregate solar production remained insufficient to satisfy basic residential demand during daylight hours. Under this constrained operating profile, the scheduling algorithm sought to harvest as much solar energy as feasible into the battery storage unit without increasing midday dependence on the external grid, reserving those stored reserves to counteract high evening consumption. The second scenario evaluated a high-generation day where daylight photovoltaic output substantially exceeded local consumer demand, challenging the optimization algorithm to achieve a full battery state of charge while simultaneously meeting total community load requirements directly from clean on-site generation.
Simulation assessments confirmed that the HGPO algorithm reliably achieved superior peak load shaving compared to standalone Grey Wolf Optimization, standard Particle Swarm Optimization, and the conventional hybrid GWO/PSO implementation. Across the two seasonal test profiles, the proposed algorithm reduced peak power demand by 21.67 percent on the resource-constrained day and 28.06 percent on the surplus generation day. Standalone Grey Wolf and Particle Swarm optimizers yielded stable but lower performance, securing peak reductions ranging from approximately 21 percent to 27 percent across both scenarios. The older literature hybrid technique, designated as HYB1, demonstrated substantial sensitivity to generation conditions, delivering a peak reduction of only 5.6 percent on the low-generation day before improving to 23.5 percent under favorable daylight conditions. Furthermore, the proposed method achieved the lowest overall cost function metrics across all trials, yielding final values between 0.0123 and 0.73 while outperforming existing benchmark approaches by 1 to 3 percent across standard optimization metrics.
Detailed examination of the resulting dispatch trajectories revealed critical operational behaviors that govern successful peak demand shaving. The scheduled charging profiles demonstrated an intentional operational delay between the initial dawn onset of solar energy generation and the physical commencement of battery charging. This delay occurs because the algorithm prioritizes direct load feeding during early morning hours when solar production remains modest, ensuring that scarce renewable generation offsets immediate residential consumption rather than incurring storage conversion losses. Only when solar irradiance intensifies does the system divert surplus energy into the storage cells, driving the battery smoothly toward its maximum five-megawatt capacity without triggering compensatory demand surges from the external utility grid. Although the simulation plots revealed a minor uptick in grid consumption during specific midday charging intervals, grid draw remained substantially below the unmanaged nocturnal peak that would otherwise occur without storage intervention.
Discharging schedules generated across both scenario days exhibited similarly disciplined dispatch dynamics. As daylight waned and residential power demand climbed toward its evening crest, the battery system initiated discharge immediately, dispatching stored reserves into the feeder precisely when clean generation dropped to zero. Rather than exhausting available capacity in an uncontrolled burst, the HGPO algorithm shaped the discharge profile over consecutive half-hour intervals to flatten the evening peak into an extended, manageable demand curve. The state-of-charge curves verified that storage cells reached full capacity during the planned daytime windows and discharged down to their zero-megawatt lower operational boundary by the conclusion of the evening peak. Crucially, the algorithm satisfied all physical voltage, power rating, and state-of-charge constraints without violating network boundaries, confirming that the dynamic metaheuristic produces physically viable schedules suitable for automated distribution controllers.
Beyond absolute load shaving magnitude, the convergence characteristics demonstrated by HGPO highlight its suitability for practical distribution operations. Comparative convergence plots across five hundred iterations indicated that the alternating hybrid technique achieved optimal fitness values in significantly fewer computational cycles than competing methods during both charging and discharging optimization runs. While standalone Grey Wolf Optimization eventually converged toward competitive load reduction values, its convergence trajectory proceeded substantially slower, requiring elevated computational time to refine leading candidate vectors. In microgrid management environments operating on dynamic thirty-minute dispatch cycles, computational efficiency and rapid convergence represent vital operational attributes. By combining wolf pack exploratory diversity with swarm velocity acceleration, the proposed methodology resolves complex scheduling surfaces within tight operational windows, offering grid operators a verified computational tool to integrate expanding solar arrays, mitigate distribution strain, and bolster localized power stability.
Subject of Research: Microgrid energy management and optimal photovoltaic-battery storage scheduling using a hybrid metaheuristic algorithm
Article Title: Optimal charge-discharge scheduling of PV-battery storage system microgrid using a hybrid optimization algorithm
Article References: Ibrahim, M. S., Gharghory, S. M., & Kamal, H. A. (2026). Optimal charge-discharge scheduling of PV-battery storage system microgrid using a hybrid optimization algorithm. Journal of Electrical Systems and Information Technology, 13(1), Article 80. https://doi.org/10.1186/s43067-026-00383-5
Image Credits: AI Generated
DOI: 10.1186/s43067-026-00383-5
Keywords: microgrid energy management, battery energy storage system, photovoltaic generation, peak load shaving, hybrid optimization algorithm, grey wolf optimizer, particle swarm optimization, smart grid technology, solar power scheduling, state of charge, distribution network, load demand profile, power grid reliability, energy storage dispatch, substation power data, metaheuristic optimization
Cite Scienmag News
Faith Mcneil. (November 21, 2026). Hybrid Optimization Algorithm Cuts Microgrid Peak Power Demand by Twenty-Eight Percent. Scienmag. https://scienmag.com/hybrid-optimization-algorithm-cuts-microgrid-peak-power-demand-by-twenty-eight-percent/
Faith Mcneil. "Hybrid Optimization Algorithm Cuts Microgrid Peak Power Demand by Twenty-Eight Percent." Scienmag, 21 November 2026, https://scienmag.com/hybrid-optimization-algorithm-cuts-microgrid-peak-power-demand-by-twenty-eight-percent/. Accessed 22 November 2026.
Faith Mcneil. "Hybrid Optimization Algorithm Cuts Microgrid Peak Power Demand by Twenty-Eight Percent." Scienmag. November 21, 2026. https://scienmag.com/hybrid-optimization-algorithm-cuts-microgrid-peak-power-demand-by-twenty-eight-percent/








