In this era of fast globalization, industrial growth, population expansion, and technological advancements, the global energy demand has reached an all-time high. An internalizing cycle of global economics emerges when countries become mutually dependent on one another for energy supplies, driven by an ever-increasing demand for governments to satisfy their populations' lifestyle expectations. The global instability in particularly fossil fuel markets among nations, leads to an energy crisis concerning developing countries of the world like Bangladesh[1].

At the same time, one of the most vulnerable countries to global warming, Bangladesh has rapidly widening problems[2]like rising sea levels, floods, and cyclones exacerbating energy crisis & thereby economic peril. Using renewable energy is seen to be explored by utilizing the possible source of renewable energy generation domestically. The optimization algorithms will become more crucial to control renewable energy systems efficiently due to high dimension problems even existing methods cannot attend those. Although exact optimization methods are highly effective in some cases, they turn out to be intractable for large non-linear problems due to the search space increasing exponentially. Such cases cannot just be solved in any traditional way like the greedy algorithms[3]. Metaheuristic algorithms [4], [5] are more flexible and better at working with large-dimensional tasks like global searches.

Solar and wind microgrid systems, as part of renewable energy help to tackle the growth in power demand attributed to rapid industrialization and urbanization[6]. The decrease in the use of fossil fuels and their damaging greenhouse gas emissions provides support to this view. Switch to sustainable energy sources for accommodation. The serious scenario regarding the depletion of fossil reserves and challenges related to s emissions underline the need for a transition away from the use of carbon compounds. This is especially true for microgrids, where hybrid systems might be more cost-effective than expanding traditional power networks to rural area[7].

The Whale Optimization Algorithm (WOA), a meta-heuristic optimization algorithm based on humpback whale social behavior and their favorite hunting strategy—bubble-net hunting—is presented in this work in [8], [9]. To replicate the steps involved in hunting ring creation, bubble-net feeding, and prey search, WOA uses three different types of operators. Conversely, WOA is used for six optimization issues in structural engineering (tension/compression spring design and pressure vessel design), demonstrating its superiority and effectiveness over conventional optimization methods. The method's utility for solving various optimization problems is further expanded by current research on binary and multi-objective (MO) WOAs[9].

In the Particle Swarm Optimization (PSO)[10]method, Every particle in the search space represents a possible solution and updates its position in accordance with both its own and the swarm's best-known roles. PSO has evolved into various variants that focus on specific enhancements such as parameter adjustment, multi-objective optimization, and hybrid techniques. Because of its adaptability and ease of coding, PSO has several practical uses (customized), ranging from engineering to artificial intelligence, as well as simple/complex searches[11].

Mirjalili et al. introduced a Grey Wolf Optimizer(GWO)[12]. This is similar to their leadership structure, which includes alpha, beta, delta, and omega wolves with distinct positions represented in the algorithm. This implementation has been compared with a few well-established optimization techniques such as PSO and Differential Evolution (DE), demonstrating competitive outcomes on both standard test functions as well as real-world engineering problems[12]. T. Agajie et al. explored optimal sizing of Renewable energy sources that are hybrid for dependable, hygienic, and cost-effective power production systems, highlighting essential elements, variables, procedures, and information[13].

In[14]researchers reviewed Microgrid power-quality problems, optimization methods, and management schemes for hybrid microgrids, emphasizing their increased system reliability. According to [15], An electrical energy control technique reduces the proposed energy expenditures and greenhouse gas emissions significantly for hybrid solar-powered electric vehicle charging stations and biogas by 74.67%. A. Karmaker et al. proposed a hybrid PV-Wind-FC system with battery energy storage which reduces costs and improves energy management by eliminating converters and minimizing intermittency[16]. In[17] the proposed HFAPSO algorithm effectively optimizes an islanded green energy system, achieving 100% renewable energy and minimizing annual energy costs while meeting energy demands. For electric car hybrid energy storage systems, N.-D. Nguyen et al. suggested a predictive control-based energy management approach that boosts performance and efficiency without needing to know the speed of the vehicle or anticipated demands[18]. Another study examined a gradient-based energy management system with deep determinism method that outperforms a twice the width Q-learning-based strategy in hybrid electric tracked cars and increases fuel economy by 13.1%[19]. The proposed predictive control strategy in[20]effectively manages a wind and solar microgrid that is linked to the grid and has hybrid storage for energy, reducing costs and increasing battery life. The proposed energy management strategy by M. Behera and L. Saikia for plug-in hybrid electric buses improves fuel economy and avoids the curse of dimensionality, with a fuel consumption increase of 3.23% compared to the dynamic programming algorithm[21]. The proposed For dual-DC-port dc-port dc-ac converter-connected PV-battery hybrid systems, a zero-vector-regulation-based closed-loop power distribution approach provides flexible power distribution and grid-side current quality without complex Synthesis for the voltage vectors[22].

Machine learning as well as artificial intelligence can effectively manage energy in hydrogen fuel cell vehicles, reducing greenhouse gas emissions and improving vehicle-to-everything connectivity[23]. The Artificial Gorilla Troops Optimization (GTO) algorithm effectively solves The optimal power flow problem for hybrid renewable energy systems using probabilistic methods, reducing total system cost[24]. Reinforcement learning (RL) can optimize energy use in smart buildings, hybrid vehicles, and cybersecurity, contributing to a sustainable environment and reducing carbon emissions[25]. DeepEE, a deep reinforcement learning framework, can optimize data center energy consumption by up to 15% and 10%, achieving more stable performance gains[26]. Hybrid policy-based learning through reinforcement (HPRL) energy management effectively optimizes island group energy systems in scenarios where energy transmission is constrained, ensuring energy supply and satisfying particular demands[27]. Learning through reinforcement can effectively manage energy in fuel cell hybrid systems, but requires careful training environments and reward function settings[28].

The hybrid energy storage system (HESS) with hydrogen/bromine redox flow battery and supercapacitor effectively smoothens power and accommodates pulse power loads in grid-integrated solar PV systems, improving overall system dynamics[29]. Compared to alternative approaches, the AMPC-based on an energy handling plan for electric vehicles with fuel cell hybrids reduces system deterioration and increases fuel efficiency[30]. The proposed energy-consumption management system effectively lowers peak-to-average ratios and power bills. by 28% and 49.32%, while maintaining user comfort[31]. The proposed predictive control strategy in[32]effectively allocates load current in Energy storage using a battery-supercapacitor hybrid systems, reducing energy losses and battery degradation in electric vehicle applications.

This paper focuses on the application, modification, and hybridization of WOA in various fields of engineering [8]; and also identifies its pros and areas that still need investigation. WOA techniques are applied in five different sectors with 61% focusing on modifications, 27% on hybridizations, and the remaining 12% involving multi-objective variants. It points to the growing relevance of WOA approaches in engineering[8].

This study's primary goal is to develop an effective optimization algorithm for the energy management of hybrid energy systems while taking into account renewable sources' maximum potential to lower operating costs. The DPO algorithm has been proposed that combines the GWO and WOA and addresses the exploration versus exploitation trade-off in dynamic and highly constrained energy management problems. The benefits of DPO are that it helps keep the cost of electricity (COE) down while maintaining reliable service via a reduction in the loss of power supply probability (LPSP). Additionally, this optimization can significantly enhance the renewable factor, reduce load shedding, and maximize the production of hydrogen which in turn can be utilized for powering hydrogen vehicles. On top of that, the flexibility of the algorithm allows it to make more cost-effective use of renewable energy sources and hence cuts carbon emissions while boosting environmental sustainability. By intelligently distributing and storing resources, DPO provides a way to transition to cleaner energy infrastructures with significant resilience which will be critical for deploying sustainable future supplies.

Additionally, this paper is structured as follows: The approach is described in Section 2. Techniques for microgrid energy planning and control are covered in Section 3. The findings and discussion are briefly presented in Section 4. The conclusion and future work direction are the main topics of Section 5.

This section with suitable subsections discusses the methodological approach of this research work. In this section, the methodology is well discussed with appropriate illustrations.

System configuration for the hybrid system

This section, with suitable subsections illustrates the hybrid system configuration on which the proposed optimization algorithm has been applied. The hybrid system has been shown in Figure 1.

Figure 1.

Proposed hybrid microgrid system’s schematic diagram

Wind Turbine.

For electrical power, wind energy is a popular renewable energy source, especially in the country's southeast and south, which have a lot of potential for using this plentiful energy [33]. The velocity of the wind at the height of the anemometer needs to be translated to the appropriate hub heights because wind speed changes with altitude. For this conversion, the power law equation is applied, as this correlation illustrates [1]:

V 21 V 11 = ( h 21 h 11 ) α (1)

Hither, V₁₁ and f₂₁ stand for the speed (m/s) at the corresponding height (h₁₁) and hub height (h₂₁) in meters, respectively. The friction coefficient is α, which is also referred to as the power law exponent or Hellmann exponent. The friction coefficient (α) can be impacted by changes in height above the ground, temperature, season, wind speed, and the roughness of the terrain.

The following formula is used to approximate a wind turbine's power production [34]:

P wi−output = 0, P rat ( V 3 − V c−in 3 ) ( V 3 − V c−in 3 ) , P r , V< V c−in ,V> V cut−off V c−in ≤V< V cut−out V rat ≤V< V cut−off(2)

Hither, P_wi-output refers to the wind turbine's kilowatt-output power, Prat refers to the rated power in kW, moreover, V_cut-off, V_rat, and V_c-in refer to cut-in wind speed, nominal wind speed, and cut-out wind speed similarly. Small-scale wind turbines can operate well even in low wind conditions because of their limited V_cut-in.

Eqs. (1) and (2), combined with the wind speed information from the NASA POWER API with a granularity for the specific geographic coordinates of the study area[35]were used to calculate per-unit power generation. Figure 2 shows the outcomes.

Figure 2.

Average monthly wind speed for a specific year

Solar.

Bangladesh's geographic location offers a plethora of opportunities for solar power harvesting [36]. Because Bangladesh is situated in an area with high levels of solar irradiation, it has year-round access to plenty of sunlight. The following formula can be used to determine the power generated by the panels as a function of solar radiation [6]:

Psolarout= PN−solar × G G reference ×[1+ K tem ( T am +(0.0256×G))− T reference (3)

Hither, P_solarout refers to the output power (kW) of solar, P_N-solar refers to rated power under reference conditions, G refers to solar radiation in Wm⁻², G_reference equal to 1000 Wm⁻², T_reference equal to 25 °C, K_tem equal to −3.7×10⁻³ (1/°C), Tam refers to the ambient temperature. Eq. (3) and data were extracted from the NASA POWER API with a granularity for the specific

geographic coordinates of the study area[35]can be used to compute the production of solar energy, as illustrated in Figure 3.

Figure 3.

Average monthly Solar radiation for a specific year

Diesel Generator.

The diesel generator lessens load shedding and enhances power quality by acting as a backup source when the battery runs low during periods of high demand. When designing a hybrid system, the efficiency and hourly fuel consumption of a diesel generator can be expressed using the following formula [37]:

Q(t)=a×Pi(t)+b× P rated (4)

Hither, P_rated is the rated power (kW), Q(t) is the fuel consumption (L/h), Pi(t) is the generated power (kW), and percentage of fuel consumption is denoted by the parameters (L/kW), a and b are the constant parameters (L/kW). The efficiency of a diesel generator is established by [38]:

n overall = n b × n g (5)

Hither, n_overall refers to the overall efficiency and n_b refers to the brake thermal efficiency of the diesel generator. n_g refers to the generator's efficiency.

Electrolyzer.

Electrolyzing water to create hydrogen is an intriguing method for producing hydrogen for a Proton Exchange Membrane (PEM) fuel cell. With the help of an electrolyzer made up of numerous cells, each cell having a cathode and anode submerged in electrically conductive water molecules are decomposed into their elements hydrogen and oxygen. The electric power used by the electrolyzer can be defined as the following [39]:

H 2 f ={ K i H 2 P input if P in ≤ P rated K i H 2 P r otherwise (6) O 2 f = H 2 f 2 (7)

Hither, O 2 f ( Nm 3 h −1 ) refers to the per-hour oxygen production and H 2 f ( Nm 3 h −1 ) refers to the per-hour Hydrogen production. K i H 2 ( Nm 3 /h/KW) refers to the flow rate of hydrogen per kW. P_input (kW) refers to the electrolyzer power used.

Battery.

The battery compensates for the discontinuous nature of renewable energy and for future applications renewable power is stored in a battery. Hither, the amount of energy retained (kWh) in the batteries during the time slot, t is denoted by E batery time , E b minimum and E b maximum refers to the lowest and highest battery capacity for storage (kWh), accordingly. Let P charging tb and P discharging tb refer to peak battery charge and discharge power (in kW),similarly. P c m and P d m refers to the maximum power of charging and discharging. The system's battery capacity (kW) is calculated based on demand and autonomous days, using the calculation below [40]:

{ E battery time = E battery time−1 + ( P charging tb , e charge − P discharging tb e discharge )Δt when0≤ P charging tb ≤ u b time , P charging tb = P c m when0≤ P d m ≤(1− u battery t ), P discharging tb = P discharging max E b minimun ≤ E battery time ≤ E b maximun E battery time ≤ E battery 1 (8)

The change in stored energy throughout the time period ∆t prior to and there after charging/discharging is shown in the first section of eq. (8). The charging and discharging power should not be greater than their maximum values, as indicated by the second and third claims. The binary parameter u b time makes sure the charging and discharge processes do not happen sequentially. The fourth element of eq. (8) constrains the energy stored, ensuring that it remains between its smallest and highest capacity. The last condition states that the energy that remains at the close of the transmit term is the same as its initial state.

Load.

The seasonal pattern of energy demand for a typical day in Rangpur has been illustrated by the monthly average load curve as shown in Figure 4 which reveals that peak loads occur during the summer months (higher cooling loads) and minimum during the winter months (higher heating loads).

Figure 4.

Average monthly load Curve for a specific year

The Dual Predator Optimization Algorithm (DPO) is a meta-heuristic algorithm that is proposed in the course of this research. The proposed DPO algorithm combines the WOA and the GWO characteristics to efficiently tune the trade-off between exploration and exploitation for hybrid energy systems optimization. This flexible structure allows DPO to deal with renewable energy resources, multiple contingency plans, batteries, and diesel engines among others as required always having energy available. The DPO comprises the following seven primary steps:

• Initialization

• Evaluation

• Selection

• Exploration

• Adjustment

• Reassessment

• Identification

I. Initialization

Initially, the search agents are dispersed around the search space at random. They all can start by exploring different solution spaces due to the fact that each agent is initialized in a different state. In the first part, since there is a lot of variability, it allows for the algorithm to within a few iterations quickly around where in the search space it is.

II. Evaluation

The fitness of each agent is calculated which helps to find out, how well the agents are capable of solving the problem. This feature identifies the agents who are productive and the ones that are a burden. The foundation for selecting the best agents is fitness assessment, which is also used to monitor an algorithm's development in terms of quality.

III. Selection

In fitness evaluations, the three highest-scoring agents, are chosen as leaders named Alpha, Beta, and Delta. These agents represent the optimal solutions at that iteration and are used as a reference for other agents to learn the way to reach a potentially optimum solution.

IV. Exploration

The loop continues up to a predefined maximum number of iterations. During each iteration, parameters are updated to define the movement strategy. If the parameters suggest insufficient progress, movement is directed toward Alpha (α), Beta (β), or Delta (δ). Otherwise, random exploration of the search space is performed to identify new potential solutions. Additionally, if no improvement is observed over several iterations, repositioning is applied using a chaotic map to escape local optima.

V. Adjustment

In the main loop, all agents are forced to move inside the search space boundary. This is done to ensure all the agents are within valid bounds which allows the search process to work well.

VI. Reassessment

After each iteration, the fitness of every agent is evaluated with their new position. This new fitness value gives you an idea of which agents are the top performers at this point. In case some agent has a better value than the current leaders, i.e. if one of the α, β, or δ values is worse than that shown by an individual of the population then based on this comparison will update their corresponding α, β, and δ to move towards ideal solutions.

VII. Identification

The loop ends after that number of iterations. The best agent is honored as Alpha, a by-product, or output of an optimal solution. This last stage finds the optimal solution and its exploration in the search space is exhaustively scrutinized.

The 'Result and Discussion' consists of two main headings (primary analysis and commercial analysis). The central analysis directs to the results relating to the critical findings extracted from experimental or survey data with an in-depth analysis of the results. In the commercial analysis looks at what the data reveals in the more practical terms of economics or business and in what ways it is reasonable to use these findings commercially.

Primary analysis

This section discusses the findings of the research work by dividing the contribution into several subsections in the following. It is divided into eight individual cases such as Cost of Electricity (COE) analysis using DPO, GWO, and WOA, Loss of Power Supply Probability (LPSP) evaluation, Renewable Energy Utilization Comparison, Load Shedding Analysis, Hydrogen Production Comparison, Carbon Emissions Reduction, Net Present Cost (NPC) Analysis. Each of these individual cases are discussed below.

Case 1: A yearly profit comparison for the average electricity cost (relative to DPO) over the course of a full year is displayed in Figure 5. DPO typically produces the lowest COE, saving between 13% and 27%, and roughly 10% to 15% when compared to GWO. As an example, in January, DPO's COE was 4.02 Taka, which is 14.1% less than WOA's 4.68 Taka and 27.04% less than GWO's 5.51 Taka [1 USD = 117.70 Taka as of 21^stDecember, 2024]. The DPO algorithm efficiently lowers the COE by optimizing the combination of backup resources, like diesel generators, and renewable resources, like wind and solar power. This includes everything from how the hybrid system components are designed, constructed, and packaged through to their installation in a vehicle as well as all other factors checked upfront cost, materials for making each component of the hybrid systems, and even maintenance needs throughout ownership. Simulations showed that the best resource allocation method of the DPO algorithm makes it a more reliable cost reduction performance than traditional energy management algorithms.

Figure 5.

Average energy cost for a specific year

Case 2: This case illustrates the effectiveness of the DPO algorithm in reducing the LPSP that is satisfactory with the security and dependability constraints under consideration. In the absence of adequate renewable and functional systems, the power supply indicated by the LPSP score has a higher chance of turning non-dispatchable. The DPO algorithm consistently produces lower LPSP values as indicated in Figure 6; in this case, it eliminates 7.2% improvements over GWO and can even surpass WOA by 7.7%. These gains resulted in April LPSPs for WOA and GWO of 10% and 5.8%, respectively, versus an increasingly high one for DPO of just 1.93%. The DPO algorithm was able to reduce the LPSP by approximately 60% due to the ability of a battery storage approach for time-varying response from sustainable resources as well as support these gaps. By giving priority to battery discharge (the first rule of reductions), the algorithm not only provides more reliable performance but also improves fuel efficiency and reduces greenhouse gases in the process. This is done regardless of whether there are batteries to discharge from, or the diesel generator must be started at the earliest sign that intermediate battery states of charge will predict being out-of-usefulness in some future distribution cycle. The above image shows the very low LPSP of 0 which shows how the developed DPO algorithm is able to deliver relatively constant power output under such highly variable renewable conditions.

Figure 6.

Average loss of power supply probability for a specific year

Case 3: Monthly analysis of renewable energy for a year is shown in Figure 7, where the comparison between DPO, WOA and GWO algorithm is discussed. For DPO algorithm the value of renewable energy utilization is almost 100% for the whole year. On the other hand, the values for WOA and GWO are on average 3–5% less than DPO for every months of the year. The utilization of DPO results very practical and reliable results which proves the potential of greater optimization for this algorithm.

>Figure 7.

Average renewable energy for a specific year

Case 4: The annual load shedding analysis for a year is illustrated in Figure 8 where the comparison of DPO, WOA and GWO is shown. In case of WOA and GWO the values of load shedding are on average more than 8 kW/month and 10 kW/month, respectively. Meanwhile the load shedding for DPO algorithm is 0 kW/month which proves the fact that DPO algorithm is the most promising algorithm. It is quite evident from the data that DPO is also superior to WOA, GWO in terms of steadfastness in the energy distribution free from any interruption. It’s zero load shedding throughout the months proves that the system has the maximum ability to have an optimal and stable system and is even preferable to be used for critical applications where the supply of power cannot be interrupted. While the relatively lower load shedding is with WOA, GWO these remain open to challenge under high-stress operational conditions, as experienced in the latter half of Enhancing Operational Conditions.

Figure 8.

Average load shedding for a specific year

Case 5: In Figure 9, the monthly production of hydrogen for a year is compared with DPO, WOA and GWO algorithm. From the figure it is evident that DPO algorithm produces the highest amount of hydrogen over the whole year which is around 1.75 × 10⁴ kg/month on average. The hydrogen production using WOA and GWO algorithm is significantly lower than DPO and the average amounts are around 1.5 × 10⁴ kg/month for both of them. Hence the DPO algorithm maximizes the hydrogen production and makes the algorithm efficient and reliable. Looking at the average monthly analysis, average DPO performance was outstanding all around; it created 17.60% more hydrogen than WOA, 21.10% better than GWO. This consistent increase demonstrates that DPO is the most effective method of all for getting H2 month by month.

Figure 9.

Average hydrogen production for a specific year

Case 6: DPO algorithm can also optimize generation and demand while backing out carbon-intensive fossil fuels from the mix. Power Management Update, the DPO also establishes a carbon-smart approach by continuing to increase the use of solar and wind resources which emit less carbon compared to diesel generators in order for further reduction of emissions in Table 1, moving progress towards outcomes envisaged by the New Energy Industrial plan. Furthermore, it maximizes the use of stored energy that is generated from renewable energy sources and helps to store them effectively for periods when high demand occurs and possibly prevents the plant at some times from utilizing non-renewable sources. In addition to improving the efficiency and lifespans of the overall energy system, it conserves resources by using less fuel. Flow batteries that generate safe, zero-carbon energy storage solutions must be integrated with the work of upstream engineers in order to be delivered in a system design.

Table 1.

Emission of gases per year

Quantity	Value (kg/yr)
Carbon Dioxide	0
Carbon monoxide	0
Sulfur Dioxide	0
Nitrogen Oxides	0

Case 7: The Dual Predator Optimization algorithm, applied to an analysis of the Net Present Costs (NPC) of the different energy components of Rangpur in Figure 10, outlines the economic implications of adopting an integrated generation facility for the defined hybrid energy system, providing excellent detail on financial feasibility. At USD 20,649.88 for all components, this is a rather costly investment to perform this system in this area. The electrolyzer has the highest NPC (USD 7,859.85). The cost is very high, which suggests that the hydrogen production technology, an essential technology for energy storage and release, is very expensive, which may be due to its high technology and material needs. Next in line is the load-flowing element, a crucial component for the energy distribution throughout the system, costing USD 5,378.25, again highlighting the significant infrastructure investment associated with effective energy distribution. The photovoltaic systems that utilize energy from the sun make up another big piece of the investment, accounting for USD 3,936.89. Solar panels, while declining in price over the years, still take a large capital investment and likely represent the ballpark market rate for such technology. The investment can be lower in terms of dollars per kilowatt for wind turbines (USD 1,840.38) because of regional wind profiles and the technology maturing and driving down costs over time. Items like the System Converter and Hydrogen Tank come in at USD 422.35 and USD 279.28, respectively, so they shouldn't put too much of a dent in your wallet. While these are the basic components of the technology to convert energy produced into energy that can be readily used and that of hydrogen storage, they also suggest that the technology is approaching economic viability. At the low end, the battery and generator cost USD 476.09 and USD 456.79, respectively. This lower cost is indicative of their secondary system roles as energy storage and backup power, roles that are essential to maintaining system reliability but do not carry the same technology investment needs as generation components.

Figure 10.

The Total Net Present Cost Analysis of Various Energy Components in Rangpur

In this paper, a new metaheuristic optimization method called Dual Predator Optimization (DPO) has been proposed, which combines the strengths of Grey Wolf Optimization (GWO) and Whale Optimization Algorithm (WOA) and also, further addresses the shortcomings of the mentioned algorithms. The results of the DPO method are promising, showing a significant reduction in both the Cost of Electricity (COE) and the Loss of Power Supply Probability (LPSP). Specifically, the DPO method achieves a decrement of an average 15–20% in COE, making it highly efficient in reducing energy costs. Furthermore, the LPSP is reduced by an average of 7.5%, which is an also a significant improvement over the other mentioned established methods such as GAO (7.2%) and WOA (7.7%). With zero load shedding and 100% renewable energy utilization, DPO sets a novel milestone ensuring greater operational efficiency and improved sustainability. It’s capabilities to lead in renewable energy technologies and hydrogen production are demonstrated by its 21.10% efficiency advantage over GWO and its 17.60% hydrogen production advantage over WOA, which highlights its superior technology and energy management effectiveness. To conclude, the results are competitive enough to other established algorithms to provide sustainable optimization solutions to the hybrid energy systems. However, as future upgradation to this proposed algorithm, further investigation is needed to analyze the DPO algorithm’s scalability and how it may perform in larger and more complex energy systems. While the system can scale up and down to accommodate fluctuating energy needs, integrating machine learning algorithms into real-time prediction and balancing could further optimize it. A possibility would be to further develop the idea with other renewable energy sources such as biomass or hydropower included in a hybrid system, suggests DPO. Embracing flow batteries would bring the goals of a resource-efficient and CO2 neutral energy system one step nearer to fruition.