Synergistic Computing for Sustainable Energy Systems: A Review of Genetic Algorithm-Enhanced Approaches in Hydrogen, Wind, Solar, and Bioenergy Applications

Abstract

The imperative for sustainable energy solutions has spurred extensive research into renewable resources such as hydrogen, wind, solar, and bioenergy. This paper presents a comprehensive review of recent advancements (2015–2024) in the application of Genetic Algorithms and associated computational technologies for the optimisation and forecasting of these energy systems. This study synthesizes findings across diverse areas including hydrogen storage design, wind farm layout optimization, solar irradiance prediction, and bioenergy production and utilization. The review categorizes the literature based on renewable energy sources and their specific areas of application, such as system optimization, energy management, and forecasting. Furthermore, it examines the role of sensitivity analysis and decision-making frameworks enhanced by Genetic Algorithm-based approaches across these domains. By highlighting the synergistic potential of computational intelligence in addressing the complexities of renewable energy deployment, this review provides valuable insights for researchers and practitioners seeking to accelerate the transition towards a more sustainable energy future.

1. Introduction

With the dawn of electrification, the conversion and distribution of electricity became increasingly important. Initially, simple systems could be controlled manually, but as they developed, they required the introduction of further automation. Currently, the power sector faces new challenges. The increasing share of green energy in the overall energy mix poses challenges in properly balancing energy in the distribution grid. It is no longer sufficient to simply react by reducing or increasing the energy supply according to demand. It is essential to be able to predict future production and demand.

This article analyses current trends in renewable energy. Beginning with an overview of various solutions proposed over the past 10 years, it will describe the key issues that are most frequently raised. However, it will also consider new threats and trends, the emergence of which may already be noted and which could develop into major issues in the future.

Photovoltaic panels and wind turbines are among the most widely used renewable energy sources. They are characterised by low energy production costs. However, their disadvantage is the inability to control the amount of energy produced, as it depends on weather conditions, which cannot be controlled. Energy banks, not only battery-based but especially hydrogen-based ones, can be helpful here. Furthermore, power compensation can be achieved using biomass energy. Despite its limitations, it allows for energy generation when needed.

Recently, these problems have been extensively analysed, and many solutions have been proposed in this area. Section 2 describes the method for selecting the articles and provides an overview. Section 3 presents a statistical analysis, identifying the main problems and their solutions, as well as the countries conducting research in this area. Section 4 discusses the collected data and presents new potential directions for development. The whole review is concluded in Section 5.

2. Literature Review

The initial component of this review involves selecting and categorising the relevant source material. Subsequently, the next section considers the organisation of the chosen materials.

2.1. Article Selection and Research Category Definition

To identify trends and development directions in ongoing research, the Scopus database was selected as a reliable source of information on publications from various publishers. A summary of the search and categorisation is presented in Figure 1.

Figure 2 presents a PRISMA-style flow diagram that describes the identification and screening process. It shows the number of records retrieved from Scopus, those excluded after title and abstract screening, duplicates removed, and the final set of 90 articles analysed. This visualisation supports reproduction and complements the automated classification method.

In the first step, the following limitations were applied, all of which needed to be fulfilled:

• A combined field that searches abstracts by Hydrogen Storage, Wind Farm Layout, Solar Irradiance, and Biofuel.
• Years: Between 2015 and 2024.
• Limited by language: English. This limitation is due to maintaining terminological consistency and supporting automated analysis. While this ensures uniform data quality, it may introduce a language bias by excluding relevant regional studies published in other languages.
• Limited by publication stage: Final.
• Limited by subject areas: Energy, Engineering, and Computer Science.

Scopus Query:

TITLE-ABS-KEY(‘‘Hydrogen Storage’’ OR ‘‘Wind Farm Layout’’

OR ‘‘Solar Irradiance’’ OR ‘‘Biofuel’’)

AND PUBYEAR > 2014 AND PUBYEAR < 2025 AND (

EXCLUDE (SUBJAREA,‘‘AGRI’’) OR EXCLUDE (SUBJAREA,‘‘BIOC’’) OR

EXCLUDE (SUBJAREA,‘‘ECON’’) OR EXCLUDE (SUBJAREA,‘‘NEUR’’) OR

EXCLUDE (SUBJAREA,‘‘CHEM’’) OR EXCLUDE (SUBJAREA,‘‘MULT’’) OR

EXCLUDE (SUBJAREA,‘‘SOCI’’) OR EXCLUDE (SUBJAREA,‘‘BUSI’’) OR

EXCLUDE (SUBJAREA,‘‘MEDI’’) OR EXCLUDE (SUBJAREA,‘‘CENG’’) OR

EXCLUDE (SUBJAREA,‘‘EART’’) OR EXCLUDE (SUBJAREA,‘‘DECI’’) OR

EXCLUDE (SUBJAREA,‘‘MATE’’) OR EXCLUDE (SUBJAREA,‘‘PHYS’’) OR

EXCLUDE (SUBJAREA,‘‘ENVI’’) OR EXCLUDE (SUBJAREA,‘‘MATH’’) )

AND (EXCLUDE (DOCTYPE,‘‘no’’ )) AND (LIMIT-TO (LANGUAGE,‘‘English’’ ))

AND (LIMIT-TO (EXACTKEYWORD,‘‘Genetic Algorithms’’) )

This resulted in 194 publications. These articles were limited to those that met the following keywords:

• Sensitivity Analysis;
• Decision Making;
• Optimisations;
• Forecasting;
• Energy Management;
• Costs;
• Economic And Social Effects.

Scopus Query:

LIMIT-TO (EXACTKEYWORD,‘‘Sensitivity Analysis’’) OR

LIMIT-TO (EXACTKEYWORD,‘‘Decision Making’’) OR

LIMIT-TO (EXACTKEYWORD,‘‘Optimisations’’) OR

LIMIT-TO (EXACTKEYWORD,‘‘Forecasting’’) OR

LIMIT-TO (EXACTKEYWORD,‘‘Energy Management’’) OR

LIMIT-TO (EXACTKEYWORD,‘‘Costs’’) OR

LIMIT-TO (EXACTKEYWORD,‘‘Economic And Social Effects’’)

Applying this filter, the number of results decreased to 93 documents. Thus, the next filter was applied, which limited the results to the following keywords:

• Hydrogen Storage;
• Hydrogen Storage System;
• Wind Power;
• Wind Turbines;
• Wind;
• Offshore Wind Farms;
• Solar Power Generation;
• Solar Energy;
• Solar Radiation;
• Solar Irradiances;
• Biomass;
• Biofuels;
• Biofuel;
• Biodiesel;
• Fuels; Optimisations;
• Forecasting;
• Economic And Social Effects;
• Costs;
• Energy Management;
• Decision Making;
• Sensitivity Analysis.

Scopus Query:

LIMIT-TO (EXACTKEYWORD,‘‘Hydrogen Storage’’) OR

LIMIT-TO (EXACTKEYWORD,‘‘Hydrogen Storage System’’) OR

LIMIT-TO (EXACTKEYWORD,‘‘Wind Power’’) OR

LIMIT-TO (EXACTKEYWORD,‘‘Wind Turbines’’) OR

LIMIT-TO (EXACTKEYWORD,‘‘Wind’’) OR

LIMIT-TO (EXACTKEYWORD,‘‘Offshore Wind Farms’’) OR

LIMIT-TO (EXACTKEYWORD,‘‘Biomass’’) OR

LIMIT-TO (EXACTKEYWORD,‘‘Biofuels’’) OR

LIMIT-TO (EXACTKEYWORD,‘‘Biofuel’’) OR

LIMIT-TO (EXACTKEYWORD,‘‘Biodiesel’’) OR

LIMIT-TO (EXACTKEYWORD,‘‘Fuels’’) OR

LIMIT-TO (EXACTKEYWORD,‘‘Solar Power Generation’’) OR

LIMIT-TO (EXACTKEYWORD,‘‘Solar Energy’’) OR

LIMIT-TO (EXACTKEYWORD,‘‘Solar Radiation’’) OR

LIMIT-TO (EXACTKEYWORD,‘‘Solar Irradiances’’)

In the next step, the resulting list of articles was manually checked to ensure that the Scopus-assigned category was correctly assigned. This resulted in a list of 90 articles that were further analysed.

Information about selected publications was exported to a CSV (Comma-Separated Values) file containing data such as title, authors, year of publication, etc., which was then imported to a PostgreSQL 16.2 database. This enabled later analysis through SQL (Structured Query Language) queries for data mining and aggregation. This process was automated. A programme written in Python 3.12.2 was used for this purpose and to prepare tables and graphs. These data summaries allowed researchers to perform further analysis. Statistical information is presented in Section 3.

All relevant replication materials, including the raw scopus.csv export (Table S1) and the thesaurus_mapping.csv file (Table S2), are provided in the Supplementary Materials to enable full replication of the analysis.

2.2. State of Art

During the extraction of categories, four of them were formulated. Thus, the review of the state of the art is divided into four parts.

2.2.1. Hydrogen Storage

In early experiments, hydrogen was identified as a flammable gas and later produced via decomposition of methane—a process that generates carbon dioxide as a byproduct. The first electrochemical production of hydrogen occurred when scientists used Volta’s battery (the voltaic pile) to drive water decomposition in cells, producing hydrogen and oxygen. Subsequently, it was discovered that by electrolysis, that is, by applying an electric current to water, hydrogen could be generated more cleanly, without relying on fossil fuels. This electrolysis route makes it possible to produce hydrogen without emitting ${\text{CO}}_2$ or other pollutants, as long as the electricity comes from non-polluting sources. Thus, over time, the shift moved from hydrogen production tied to methane reforming (with ${\text{CO}}_2$ emissions) to green hydrogen made by pure electrolysis [1].

Due to the consumption of conventional energy sources and the need to protect the climate, renewable energy sources are becoming increasingly important. The most commonly used are photovoltaic panels [2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36], wind power [2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37] and hydrokinetic [15,28]. Their undoubted advantage is their low operating costs. However, they are characterised by poor controllability because their ability to generate energy depends on atmospheric conditions. This prevents increasing energy supply during periods of increased demand. One of the possible solutions to this problem is to store the energy generated in hydrogen storage tanks. Generally, this solution involves hydrolysing water during excess energy production, storing hydrogen in tanks, and then burning it in fuel cells when the energy demand exceeds the current production capabilities [38]. It should be noted that hydrogen combustion produces water, and thus it remains a green energy source.

Both battery and hydrogen energy storage systems have their own unique characteristics that make each solution better in a specific situation. Therefore, hydrogen storage systems can be found in various configurations. The first is for use in powering a low-emission house. In this case, it is possible to use both a battery and a hydrogen storage system. As presented in [5], the combination of both energy storage systems provides the best solution. This is related to their efficiency and estimated usage time, which influence changes in their parameters. Due to the system’s sensitivity to weather changes, control of such a system can be implemented using Neural Networks. The method of selecting parameters using a Genetic Algorithm is presented in [4]. The method for selecting the parameters of the entire system is described in [27].

Stored energy is most often used to balance the energy supply during periods of increased demand and production shortfall. Basic systems focus on optimising the parameters for zero-emission and low-emission buildings. However, larger systems can serve as a standalone power source or be part of a transmission grid. Therefore, proper management of these resources is essential. Excess battery storage can increase costs while preventing full utilisation of this technology. On the other hand, hydrogen fuels are much more durable and have a longer lifespan [2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,37]. Particular locations where such power balancing can be important are data centres [37] and machine parks [2,13].

Transport is a separate category in power balancing. Starting with infrastructure, we can consider Chinese ports. They have access to photovoltaic panels, wind turbines, and hydrogen storage facilities. Properly selected parameters for their use allow for both cost reduction and a reduction in ${\text{CO}}_2$ (carbon dioxide) emissions [14]. However, the energy stored in hydrogen tanks does not necessarily have to be immediately converted to electricity. An alternative is to connect them to charging stations for hydrogen-powered vehicles [39]. Using an Elite Genetic Algorithm allows vehicle refuelling planning, which reduces the uncertainty of renewable energy supplies. Vehicle engines and other drive components can also be optimised for hydrogen use, reducing the cost of their operation [11]. Hydrogen fuel cells also have applications in aviation. In aircrafts, it is possible to supplement fuel engines with hydrogen fuel cells to reduce exhaust emissions [26]. In drones, the use of a combination of batteries and hydrogen cells allows a reduction in cost and an increase in range [7].

In the real world, every energy conversion condition involves losses. Most often, its dissipation results in heat emission. An interesting solution was proposed in [16]. A standard energy storage system in hydrogen tanks generates heat during the recovery of electrical energy. Therefore, the entire system was optimised to simultaneously maximise the capacity to store energy and provide heat, for example, as hot water.

The recurring motivation for research in the analysed works is very often the reduction in cost [2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37]. The differences here lie mainly in the method of obtaining them, beginning with determining the factors influencing the cost of the system [36]. Another improvement is the selection of the cheapest system that still meets the assumed criteria [35]. However, construction costs are only one component of the total cost, which also includes operating costs. For this reason, some studies also take this aspect into account when optimising the cost [17,18,27,30]. This can be influenced, for example, by battery ageing [31]. However, cost minimisation is not always the only goal. Equally important is the stabilisation of energy prices, even if they are slightly higher than assumed [34]. The benefit here is the security of supply [28]. This can be achieved by burning biomass [25], selecting parameters for an Artificial Neural Network that optimises the operation of the power plant, and using LNG (liquefied natural gas) to liquefy hydrogen [23]. Optimising strategies for managing energy from multiple sources is therefore crucial. The methods mentioned above for optimising hydrogen processing are helpful in this regard. The containers in which hydrogen is processed also impact economic viability, from the selection of the appropriate size for demand to their overall design in zinc–cobalt deposits. Yet another type of improvement is the use of a Tesla valve [3]. Optimisation of its parameters allows controlled decompression of stored hydrogen.

However, most of the time, the parameters of the hydrogen storage facilities and the entire installation are selected using the Genetic Algorithm [2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,39]. Optimisation includes both the selection of the capacity and power of the basic elements of the system [21] and more advanced installations with components such as storage tanks using high-temperature molten salt and the organic Rankine cycle [29]. Modifications of this method can also be found in the literature. For example, in the work [25], the Hybrid Firefly Genetic Algorithm was proposed for a power plant consisting of photovoltaic cells, wind turbines, a hydrogen storage system, and biomass combustion, which in this case proved to be more effective than other compared methods. Another example of a modification is the use of the Swarm Algorithm to optimise the parameters of the Genetic Algorithm [21]. However, the Swarm Algorithm itself also has very good optimisation properties, as presented in [10]. Other methods used for optimisation are Monte Carlo [36], the Cuckoo Algorithm [35], and Herd of Horse Optimization [15]. A comparison of the methods is presented in [19], where the Modified Multiverse Optimiser turned out to be the best for the case considered.

The distribution of energy in the power grid is a complex issue. The first problem is the ability to supply energy only “now”. Energy cannot be fed into the grid “in advance”. With unstable energy sources, this creates additional complications, hindering compensation for the energy that needs to be supplied to the grid. Therefore, already at the stage of designing the power supply system, the Genetic Algorithm can be used to optimise the entire investment and Artificial Neural Networks can be used to optimise parameters during the installation operation [24]. Other methods used for this purpose are Sunflower Optimization [22] and the Levy Flight Algorithm [33]. In the next case, four strategies for managing energy in the grid and its storage facilities were proposed. Each is adapted to a different season and the availability of renewable resources. However, the use of Artificial Neural Networks to manage energy compensation during generator operation has been reported in the literature. They can be used in everything from low-emission home power supply design [4,27], through energy flow analysis [28], to network parameter optimisation [13].

In a distributed energy generation system, it is also necessary to coordinate the amount of energy entering the system. This is to ensure that the amount is adequate for current demand. However, transmitting information on potential energy demand and supply is vulnerable to cyberattacks. If incorrect data are entered, the transmission system may not consume excess energy if deliveries exceed declared amounts. Shutdowns could also be possible if declared amounts of energy are not delivered. Blockchain technology [32] could provide a solution to this problem. The information about the deliveries stored in it is easier to verify. It is also more difficult to modify such data later, for example, through unauthorised access to the database. Information about energy overproduction can be used to place it in a hydrogen storage facility [6] and determine its profitability [9]. A measure of optimisation efficiency and uncertainty assessment can be helpful for this purpose [36]. Optimisation is not always cost-effective. In the case of subsidised electricity supplies, the use of hydrogen storage may be an uneconomical solution [20].

2.2.2. Wind Energy

One of the most common sources of renewable energy is wind turbines. They can be used on a small scale, such as home energy installations [4]. They are often designed to complement other forms of renewable energy generation to increase the stability of supply. Predicting energy demand is also an important element here. The paper [40] considers an installation that produces both electricity and thermal energy. Depending on the season, a different algorithm was proposed to optimally divide the generated energy to secure the heat supply during periods of increased demand. Forecasting demand is also important in the design stage of such a system [41]. This allows for optimal sizing of system components. For the same reasons, predicting electricity demand is also important in large-scale installations. This allows for the selection of the optimal parameters of the designed wind farms [42]. However, this is not the only aspect that research is focused on. The trained Neural Network can be used to manage the power generation capacity in installations combining energy from conventional and renewable sources [43]. This allows for improved grid stability. Moreover, knowing the expected network load at a given time and the production capacity, it is possible to reconfigure it if necessary [44]. Sometimes, load control is also possible, as in the case of data centres [37], which also allows the planning of energy storage or resale. However, these are isolated cases in the literature.

When planning a wind farm, its impact on the well-being of people and the surrounding environment cannot be ignored. The noise generated by rotating turbines is a significant problem. In [45], the focus was on optimising the turbine layout to minimise noise while maximising electricity production. This research was further developed in [46], extending it to include wake analysis and noise level calculations compliant with ISO-9613-2 standard [47].

The installation location of turbines on a wind farm is not a simple matter. The first problem is the aerodynamic wake of the turbines, which may affect the operation of subsequent ones [46]. Therefore, not every location is optimal. In addition, there are legal restrictions that must be taken into account during design [48,49,50]. These include, for example, minimum distances between turbines for safety reasons. Different distances also result in varying cabling costs. The solution to this problem could be the use of the GeoSteiner algorithm to determine the position of the cable [51], which allowed for an increase in efficiency 27% and a reduction in cabling costs. Due to the spaces between turbines, they can also be used for other purposes, such as a photovoltaic installation. The design of the turbine layout on the farm is then multicriteria, as it takes into account both the relative position of the turbines and panels, as well as the wind force, its most frequent direction, sunlight, and temperature [52]. The variability of these parameters is also taken into account.

The location of turbines on a wind farm is discrete [53] due to the available locations where they can be placed. However, the power they generate is a real number. Using the Jensen turbine wake model and the sum-of-squares model, they were used to calculate the available power for each turbine. However, with a high mesh density, good results are easier to obtain [54]. Other restrictions include wind speed, terrain, and land availability. In [55], 156 cases were considered. However, the constraint most frequently analysed in the literature is the wind force and direction, for both onshore [56] and offshore wind farms [57]. An interesting concept is the idea of reusing the foundations of decommissioned offshore wind turbines [58]. These constitute a constraint on the preferred locations for the installation of new turbines. To address this, a conceptual two-dimensional (2D) wake model was proposed, which was adopted to calculate wind losses caused by wake effects. This solution also allows the selection of optimal parameters for new turbines. For example, in [59], different mast heights are considered in offshore wind farms, which reduces the impact of individual turbines on each other. In the case of onshore installations [60], selecting different heights and power outputs can also improve the profitability of a given investment.

Analysing the examples described above, it can be seen that their goal is to optimise investment for maximum profitability [18,19,28,31,35,40,41,45,46,52,53,54,55,56,57,58,59,61]. Not only is the construction cost crucial, but also the potential for energy harvesting. To enhance this feature, it is also beneficial to forecast the energy produced. This allows for optimal installation size selection [52,62]. In the case of overproduction, it is also possible to sell the excess energy or store it [37,41,42]. Therefore, a mathematical model was developed to predict overproduction [44].

For natural reasons, wind energy production can only be predicted. However, it cannot be controlled. Therefore, a way to stabilise the grid is to store energy during periods of overproduction. The most common methods include hydrogen storage [4,37], which are described in Section 2.2.1, and battery storage [41,42]. An advantage of the battery is the ability to supply energy immediately on demand. However, a disadvantage is the relatively rapid ageing of batteries. Hydrogen storage lacks this disadvantage, but the use of this energy is not immediate. Therefore, it is necessary to manage the installation to anticipate the growing demand and activate this energy source at the appropriate time. Such solutions are presented in [8]. Taking into account the specific characteristics of both storage systems, they are combined into a single system to leverage the advantages of both solutions [18,35]. The selection of the appropriate system size is crucial in such installations [62]. It is particularly important to determine the ratio of the power of the fuel cell to the power of the electrolyser to achieve minimal dependence of the system on the municipal power grid, while simultaneously minimising the cost of the electrolyser and fuel cells [13]. The operation of the electrolyser itself is also analysed in the context of wind farms and photovoltaic panels. To achieve optimal water electrolysis parameters, voltage-maximising converters are used to achieve constant electrolysis parameters [61]. Other energy storage methods include heat [40] and pumped storage systems [28].

Due to the combination of multiple energy sources and their storage, controlling them requires appropriate software. The simplest solutions involve only the management of the installation components [8,35] or an entire microgrid [19]. In more complex cases, it is possible to analyse and reconfigure the grid [44] or an entire Stand-Alone Multi-Source Grid Wind Turbine/PV (photovoltaics)/BESS (battery energy storage system)/HESS (hydrogen energy storage system)/Gas Turbine/Electric Vehicle system [31]. Due to the need to analyse numerous parameters and make quick decisions, such software often relies on Artificial Neural Networks [28]. In the cited example, this reduced power fluctuations in the grid by 77.6%.

A frequently discussed topic in the literature on wind farm design is the chosen design optimisation method. Many solutions are identified as the most advantageous only in individual articles. In [62], the Grasshopper Algorithm was used to select the optimal size of the wind turbine and photovoltaic panels to minimise the total cost of construction and operation. Furthermore, the installation site can be selected using a Swarm Algorithm [56]. However, the most commonly used tool to optimise the layout of wind turbines on a farm is the Genetic Algorithm [4,8,13,18,19,28,31,35,37,40,41,42,43,44,45,46,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63] in various variants. Therefore, depending on the case, a different variant may prove to be the most appropriate. In [63], a hyperheuristic was proposed to select the most efficient variant. Other proposals include the Grey Wolf Algorithm and the Whale Algorithm, which were compared with the Genetic and Swarm Algorithms in [61] and the Extended Pattern Search in [59].

The generation of hot water is also the first step in the organic Rankine cycle. This is where the idea of cogeneration of electricity through biomass combustion originated. This is a way to obtain additional energy in small-scale installations. For the economic profitability of such a solution [64], it is crucial to select the appropriate heat exchanger and the correct proportions between the thermal and electrical energy generated. A Genetic Algorithm was used to optimally select these parameters. In larger-scale installations, biomass combustion is also used. In the case of hybrid power plants, it acts as a component that stabilises production during reduced production from wind turbines and photovoltaic panels. However, it is necessary to select the parameters of such an installation, for example, using the Hybrid Firefly Genetic Algorithm [25]. However, a more comprehensive study [65] that takes into account the goal of minimising total cost proposes the Dandelion optimiser, which turned out to be better than other compared solutions.

2.2.3. Solar Energy

Due to the size of photovoltaic panels and possible installation locations, they are a widespread source of renewable energy. Unlike the wind turbines discussed in Section 2.2.2, they do not emit noise; thus, they can be installed even on residential buildings. However, the disadvantage is the instability of the energy supply. These can be forecasted [66], but a more common solution in low-emission buildings is to combine the panels with an energy storage system. In addition to the hydrogen storage systems discussed in Section 2.2.1, battery systems [27,41] are also used. A non-standard solution was presented in [67], where excess energy was used to produce ice. The work takes into account the reduction in ${\text{CO}}_2$ and tests two algorithms (Latin Hypercube Sampling and the K-Means++), which uncovered the demand to be 72.37% and 61.27%. Due to the specificity of each building, the selection of installation parameters and its control should be selected individually. However, models for automatic selection of such parameters have already been developed [4].

Although, in the case of residential installations, their location is strictly limited to the investor’s property, in the case of large power plants composed of photovoltaic panels, the future investment site must be chosen much more carefully. A comparative analysis of the results of optimising the location of photovoltaic panels in South Africa [68] showed that the latitude of the location of the power plant plays a significant role in its performance throughout the season. Optimising along longitude, on the other hand, ensures higher energy production during peak demand hours during the day. Therefore, the installation location must take into account the projected load profile. The correlation of the power plant with other energy sources, such as wind turbines, which can also be installed at the same location, is also crucial. In such a case, it is necessary to select the parameters of all components of the system to complement each other.

Due to the unstable nature of the supplies, various methods of storing energy during overproduction are being considered. The most common methods in the literature include hydrogen storage [29,61], battery storage [41,69], and their combination [27,35]. However, other cases are also being considered, such as high-temperature molten salt [29] or ice production [67]. Another idea is to use excess energy to produce hot water for heating and hot water for domestic use [16].

Regardless of whether photovoltaic installations operate independently or on the grid with other power sources and energy storage systems, the selection of the system parameters is crucial. Using components with insufficient power will result in insufficient efficiency. Oversizing, on the other hand, will cause an increase in unused potential and increased installation and operating costs. Therefore, the parameters of such a system must be individually selected based on the planned components, their layout, and the specific environment in which they will be located.

A key issue for improving the energy utilisation of photovoltaic panels is the ability to predict their energy production. This allows for power balance in the system through the appropriate control of other energy sources. The operation of panels is strongly correlated with solar radiation. However, analysing it at only one measurement point is not always effective. One way to address this problem is to install additional measurement stations located at a distance from the analysed power plant. In [70], it was demonstrated that five additional measurement points were shown to allow an accurate prediction of energy production over a period of two hours. Another parameter affecting panel efficiency is temperature; thus, analysing and predicting its value are also helpful in forecasting their power. An additional approach is to improve weather forecasting models. Particular emphasis is placed on short-term solar radiation forecasting. Using Neural Networks and data such as wind, temperature, pressure, and solar radiation, forecasting can be improved by 80% over the persistence model [71]. In cases of limited data availability, two models have also been developed, powered solely by meteorological data and values calculated based on solar radiation, which are readily available [72]. These models enable solar radiation forecasting within an hourly time frame.

The statistical studies analysed often use modern machine learning methods to improve weather forecasting during the period considered. The first of these, a hidden Markov model, allows for high accuracy within a 5–30 min timeframe. Furthermore, using a Genetic Algorithm to optimise it and introducing an additional factor for cloudless and overcast days can further improve the forecast. Other methods used are the Dragonfly Algorithm [73], Neural Networks, GAN [43,73,74], Particle Swarm Optimization [61,69,74,75], Gray Wolf Optimization [61,74], Fuzzy Logic [74], Latin Hypercube Sampling and the K-Means++ Algorithm [67], Modified Multiverse Optimizer [19], support vector regression and Random Forest [72], The Whale Optimization Algorithm [61], Cuckoo Search [35], Dandelion optimizer, Slime Mold Algorithm, and Real Coded Genetic Algorithm [65]. In addition, Artificial Neural Networks are presented as the best for forecasting short-term photovoltaic power forecasts, and Online Sequential Extreme Learning Machines are excellent for adaptive networks. However, the Bootstrap technique is optimal for estimating uncertainty. Additionally, Convolutional Neural Networks are found to excel in eliciting a model’s deep underlying nonlinear input–output relationships [75]. An unusual technique was used in [76]. There, it was noted that the implemented weather forecast models have a 10 km grid, resulting in a forecast resolution of one hour. To increase this to 5 min, it would be necessary to refine the grid to 500 metres, which would generate significant computational costs. Therefore, an analogue downscaling technique was used to create a variability forecast using a coarse 10 km numerical weather prediction forecast.

To summarise the methods used to predict generated power, three of them are mentioned most frequently. The first is Artificial Neural Networks [4,16,19,24,25,27,29,41,43,52,61,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83]. Their main purpose is to predict environmental conditions and respond to changes in the system parameters, such as generated power, consumed power, and stored power, during its operation. In addition to training the network, its parameters must also be selected. Two methods are used to do that: Particle Swarm Optimization [16,19,24,25,27,29,41,43,52,61,66,67,68,69,72,74,75,76,79,80,81,82,83] and Genetic Algorithms [4,16,19,24,25,27,29,35,41,43,52,61,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83]. They are used in the stage of system design and parameter selection. Therefore, many works describe the use of both groups to utilise its potential more effectively. Forecasting ability also facilitates power grid management, as it improves its stabilization [69,75,82,83].

2.2.4. Bioenergy

Unlike the examples described in Section 2.2.1–Section 2.2.3, bioenergy does not have to focus on electricity production. In addition, this issue also includes the supply of heat and biofuels. The latter are increasingly added as a fuel component for internal combustion engines. However, this changes the way the engine operates, which can alter the amount of gases and particulate matter it emits. Considering the diesel engine [84], this problem was solved using a pre-emission model. This improved the accuracy in predicting the emitted pollutants. Compared to the classic Genetic Algorithm, the proposed models, Non-dominated Sorting Genetic Algorithm II and Multi-Objective Particle Swarm Optimisation, improved the prediction of particulate matter emissions by 23.5% and 18.6%, respectively.

In addition to engine performance, the quality of the biofuel it runs on is also important. Its first parameter is the cetane number, which determines the autoignition temperature. It can be estimated on the basis of the ester content. This was presented in [85] using the least squares method with a support vector machine, which was combined with a Genetic Algorithm, Particle Swarm Optimisation, and a hybrid of Genetic Algorithms and Particle Swarm Optimisation. The next parameter is the viscosity of the biofuel. This viscosity is responsible for combustion, power output, and pollutant production. To evaluate the viscosity, the following was proposed by [86]: a support vector machine, an Adaptive Neuro Fuzzy Inference System, and a Feedforward Neural Network model trained by the Genetic Algorithm, Simulated Annealing, and Levenberg–Marquardt. Another way to use the described algorithms is to control the production process. Biofuels do not always have to be made from freshly pressed plants. Frying oil can also be used for this purpose. Using Genetic Algorithms and Artificial Neural Networks, the biodiesel production process can be monitored, achieving an optimal efficiency of 94.21%.

Planning the production and distribution of biofuel is equally important. Not every location is suitable for cultivation, and significant distances between harvesting, processing, storage, and distribution sites can unnecessarily increase the cost of this solution. Therefore, in [87], an optimisation model was proposed for the location–storage–transport problem to design a network for the production and distribution of algae fuel. This is an NP-hard problem and is solved with the modified versions of three algorithms: Simulated Annealing, Genetic Algorithm, and Firefly Algorithm. The performance of these methods was measured by test problems in various sizes. The results show that Simulated Annealing outperforms both the Genetic Algorithm and the Firefly Algorithm, and Genetic Algorithm outperforms the Firefly Algorithm in terms of the quality of the solutions. To facilitate decision making in the biomass-to-bioenergy supply chain, multicriteria decision-making methods have been developed [88]. Accounting for supply uncertainty, product quality, and availability in the model can help select high-quality biomass harvesting locations and reduce harvesting costs. A Genetic Algorithm was used for optimisation. It was also used in [89,90]. The authors focused on the selection of energy crop cultivation sites. Particular attention was paid to the impact of crop change on the natural environment, including soil, water (impact on soil erosion at the water catchment), and the atmosphere (greenhouse gas production). These studies used the Soil and Water Assessment Tool (SWAT) to estimate the effects on soil and water quality, and the Greenhouse Gases, Regulated Emissions and Energy Use in Transportation (GREET) model to evaluate Greenhouse Gas Emissions.

An example of biofuel production is the production of bioethanol in the process of hydrolysis of cassava starch [91]. This process is complex and depends on several factors. The goal of this study was to find the optimal parameters to maximise production. Tabu Search, Simulated Annealing, the Genetic Algorithm, and Hybrid Metaheuristic Algorithms were used to optimise the process. The algorithms were calibrated and parametrised to generate the values of the variables within their respective domains as follows: concentration (6.590 to 23.375 g/L), temperature (28.18 to 61.82 °C), and time (9 to 111 min). The results presented the best conditions for starch hydrolysis at 23.375 g/L, 61.9 °C, and 111 min, where a yield of 84.04% was achieved.

Efficient biofuel combustion depends not only on its quality but also on the design of the device it powers. For example, in the case of a pellet stove that supplies heat to a house [92], the key parameters influencing the combustion process are the airflow distances above the base of the burner pot, the airflow strength, and the power of the furnace. By changing these parameters, it is possible to maximise efficiency and minimise harmful gas emissions ($CO$ (carbon monoxide), $N{O}_X$ (nitric oxide), ${\text{CO}}_2$). The hyperparameters of the designed models are tuned using several optimisation approaches: Grid Search, Bayesian Optimisation, Particle Swarm, and Genetic Algorithms.

3. Results

We start our analysis of trends in the study area by dividing the time period into the last two periods: 2015–2019 and 2020–2024. The year 2025 is not analysed, because it was not yet completed at the time of writing this article, and the data of this year are incomplete, which could distort the statistical data.

Table 1. Publications by year in all categories.

Name	2015–2019	2020–2024	All Years	Share [%]
Total	37	53	90	100.0
Document Type
Conference Paper	20	6	26	28.89
Journal Article	16	46	62	68.89
Other	1	1	2	2.22
Renewable Energy
Hydrogen Storage	6	31	37	41.11
Wind Energy	18	14	32	35.56
Solar Energy	13	18	31	34.44
Bioenergy	5	8	13	14.44
Areas of Application
Optimisations	10	19	29	32.22
Forecasting	11	16	27	30.0
Costs and Effects	16	8	24	26.67
Energy Management	4	10	14	15.56
Decision Making	3	7	10	11.11
Sensitivity Analysis	0	10	10	11.11
Research Methodology
Experiment	28	31	59	65.56
Literature Analysis	7	8	15	16.67
Case Study	6	10	16	17.78
Conceptual	27	45	72	80.0

presents the changes in the number of publications in both periods. The analyses are performed in four categories, for which a Chi-square test (${\chi }^2$) is performed, the results of which are presented in

Table 2. Chi-square of Table 1.

Category	Contingency Table (Observed Counts)	${\mathit{\chi }}^2$	df	p-Value	Cramér’s V	Interpretation
Document Type	Journal = 16→46; Conference = 20→6; Other = 1→1	19.84	2	<0.001	0.47	Significant; strong shift from conference to journal papers.
Renewable Energy	Hydrogen = 6→31; Wind = 18→14; Solar = 13→18; Bioenergy = 5→8	12.26	3	0.007	0.37	Significant; major growth in hydrogen research.
Areas of Application	Opt. = 10→19; Forecast. = 11→16; Costs = 16→8; Energy Mgmt = 4→10; Decision = 3→7; Sensitivity = 0→10	15.43	5	0.009	0.33	Significant; shift toward sensitivity analysis and energy management.
Research Methodology	Experiment = 28→31; Lit. Anal. = 7→8; Case Study = 6→10; Conceptual = 27→45	1.59	3	0.66	0.13	Not significant; distribution stable across periods.

. Two hypotheses are formulated for each category:

• Null hypothesis ($H_0$)—There have been no significant changes in the proportion of publications in the studied category over the past two periods.
• Alternative hypothesis ($H_1$)—There have been changes in the proportion of publications in the category studied over the last two periods.

Figure 3 presents changes in the type of publication. As can be seen, the number of conference papers has clearly decreased in favour of journal articles. This observation is confirmed by the test result, where ${\chi }^2=19.84$, and thus p-value = 0.0, and the null hypothesis ($H_0$) must be rejected in favour of the alternative hypothesis ($H_1$). In this case, this may indicate that the initial versions of the papers were discussed at conferences and are now presented as refined works in journals.

The next category examined is Renewable Energy. The distribution of the articles is presented in Figure 4. In this case, we also accept the alternative hypothesis ($H_1$). Although the differences in the number of publications in the subcategories Wind Energy, Solar Energy, and Bioenergy are not significant, a significant increase in publications is visible in the Hydrogen Storage subcategory. Based on the information presented in Section 2.2, it can be assumed that this increase is related to the need to store renewable energy. Furthermore, classic battery cells only perform well when storing small amounts of energy, such as in a home installation [27]. For storage of large amounts of energy for longer periods, hydrogen storage systems are more effective.

Significant changes can also be observed among the Areas of Application, confirmed by the acceptance of the alternative hypothesis ($H_1$). As shown in Figure 5, the importance of Cost and Effects has decreased. However, it would be incorrect to conclude that cost reduction is no longer important. This may be due to an increase in the number of publications in the other categories. Furthermore, in the last decade, a subcategory, Sensitivity Analysis, has been added. This may suggest that multicriteria optimisation has become increasingly important. Not only are installation and operating costs now important, but also supply security and management process automation.

Only in the case of Research Methodology were no significant changes in the number of articles observed. Since p-value = 0.66, there was no basis to reject the null hypothesis ($H_0$). As shown in Figure 6, the differences in almost every category are insignificant. Only in the case of the Conceptual subcategory is a significant increase in the number of publications observed. However, it is difficult to clearly determine whether this is due to chance or whether these are new ideas that are still in the very early stages of development and difficult to make a decision about.

The place of publication of an article can also provide information about who is interested in developing new energy sources. This information is summarised in

Table 3. Publications by year in countries.

Country	2015–2019	2020–2024	All Years	Share [%]
All countries	37	53	90	100.0
China	4	17	21	23.33
India	2	10	12	13.33
United States	7	4	11	12.22
Iran	3	7	10	11.11
Egypt	2	6	8	8.89
United Kingdom	5	1	6	6.67
Australia	2	3	5	5.56
Canada	2	2	4	4.44
Nigeria	1	3	4	4.44
France	0	3	3	3.33
Malaysia	0	3	3	3.33
Other	10	8	18	20.0

. For these data, we formulated the null hypothesis ($H_0$) in which the participation of individual countries in these studies has not changed. However, as the results of the Chi-square test (${\chi }^2$) presented in

Table 4. Chi-square of Table 3.

Group	Contingency Table (Observed Counts)	${\mathit{\chi }}^2$	df	p-Value	Cramér’s V	Interpretation
All Countries	China = 4→17; India = 2→10; USA = 7→4; Iran = 3→7; Egypt = 2→6; UK = 5→1; Australia = 2→3; Canada = 2→2; Nigeria = 1→3; France = 0→3; Malaysia = 0→3; Other = 10→8	21.52	11	0.03	0.35	Significant; strong regional growth in Asia (China, India, Iran).

Notes: Observed counts correspond to periods 2015–2019→2020–2024. All expected frequencies ≥ 5 for valid cells; few low-frequency cells (<10%) did not affect results. Cramér’s V indicates a moderate association between period and publication geography.

indicate, this null hypothesis should be rejected and the alternative hypothesis ($H_1$) accepted. Therefore, a relationship was demonstrated between the number of publications in a given country and the period.

To avoid the dominance of large publishing systems, publication trends were monitored separately for each country rather than aggregated globally. It is natural for countries to emphasise technologies consistent with their local resources, such as solar energy in high-irradiance regions or wind energy in coastal areas. No weighting by country size or research capacity was applied, as the purpose of this study was to identify which optimisation trends occur most frequently on the global scale. Thus, a higher representation of certain technologies in larger countries reflects their genuine prevalence in national research practice rather than methodological bias.

The number of publications in a given country is presented in Figure 7. In China, a four-times increase in the number of publications can be observed. India also saw a five-times increase in publications, although in absolute terms, this is lower than in China. Other countries that recorded an increase in publications in the area analysed include Iran, Egypt, Nigeria, France, and Malaysia. It is easy to see that these countries have abundant sunlight. Therefore, it can be concluded that they aim to obtain energy from these sources in the near-future.

The next section of the paper presents a summary of the share of publications on renewable energy sources in other categories. It should be noted that some publications may belong to more than one category.

Table 5. Publications by renewable energy in other categories.

Name	Hydrogen Storage	Wind Energy	Solar Energy	Bioenergy	Total
Total	37	32	31	13	90
Areas of Application
Optimisations	12	14	9	3	29
Forecasting	4	4	17	4	27
Costs and Effects	7	15	2	4	24
Energy Management	12	4	3	1	14
Decision Making	5	3	2	3	10
Sensitivity Analysis	10	2	4	0	10
Research Methodology
Experiment	22	18	21	9	59
Literature Analysis	5	7	4	2	15
Case Study	8	5	7	3	16
Conceptual	32	27	23	11	72

presents two divisions. Chi-square (${\chi }^2$) tests were performed for both, the results of which are summarised in

Table 6. Chi-square of Table 5.

Group	Contingency Table (Observed Counts)	${\mathit{\chi }}^2$	df	p-Value	Cramér’s V	Interpretation
Areas of Application	Hydrogen = 37; Wind = 32; Solar = 31; Bioenergy = 13 across 6 application areas (see Table 5)	45.58	15	<0.001	0.46	Significant; clear relationship between energy type and dominant research application.
Research Methodology	Hydrogen = 37; Wind = 32; Solar = 31; Bioenergy = 13 across 4 method categories (Experiment, Literature Analysis, Case Study, Conceptual)	2.26	9	0.99	0.10	Not significant; methodological distribution similar across energy types.

Notes: Counts derived from Table 5. Expected frequencies all >5. Cramér’s V computed for effect size; indicates a strong association for application domains and negligible for methodology. Analyses exploratory—no multiple-comparison correction applied.

.

Because a single publication may belong to multiple application areas, these data represent co-occurrence frequencies rather than mutually exclusive categories. Therefore, the ${\chi }^2$ test and Cramér’s V were used as descriptive indicators of association strength, not as strict independence tests.

Figure 8 presents a heatmap demonstrating the most frequently analysed Areas of Application with respect to various energy conversion mechanisms. The null hypothesis ($H_0$) for this comparison is rejected due to p-value = 0.0. Therefore, the alternative hypothesis ($H_1$) should be accepted, which states that there is a relationship between Areas of Application and the energy conversion method. Analysing the heatmap, it can be seen that the topics most frequently studied are

• Forecasting in the context of Solar Energy;
• Costs and Effects in the context of Wind Energy;
• Optimizations in the context of Wind Energy;
• Optimizations in the context of Hydrogen Storage;
• Energy Management in the context of Hydrogen Storage;
• Sensitivity Analysis in the context of Hydrogen Storage.

It is clear that the goal of most research is to counteract natural-occurring problems in energy production. Therefore, it is not surprising that energy storage, which can be crucial in balancing the power grid, is considered so frequently.

The final analysis presented concerns the adopted Research Methodology approach to energy conversion. However, since p-value = 0.99, the null hypothesis ($H_0$) should be accepted, according to which no relationship between these categories was identified. However, the heatmap presented in Figure 9 allows us to observe that

• The largest number of publications focusses on Hydrogen Storage.
• The next most commonly used methods are Wind Energy and Solar Energy.
• The most widely used methods are Experiment and Conceptual.

4. Discussion

Undoubtedly, the construction and operation of sustainable energy systems is becoming increasingly important. Due to ongoing climate change and the depletion of natural resources, it is necessary to look for new directions of development.

4.1. Network Visualization

Already in the stage of reviewing the current state of knowledge (Section 2.2), it was clear that many issues were interconnected. For example, Genetic Algorithms were the most frequently used optimisation tool in all areas. They also served as a reference tool for validating the proposed solutions. Therefore, a network of connections was developed between various issues, presented in Figure 10. Genetic Algorithms are the central and largest element here. They are densely interconnected with other elements of varying sizes. Other important terms are Hydrogen Storage, Forecasting, and Optimisations. These terms already indicate the first clear trend in sustainable energy systems: the need for system optimisation using multicriteria assessment. Not only is cost important, but also the efficiency of such a system. This can be achieved by weather forecasting, which determines generation ability, and energy storage, which allows the storage of excess energy and its distribution during shortages. Therefore, other important terms are Energy Management and Decision Making. Both are related to the topic of Automated Management of such infrastructures, which aims to increase the efficiency of its components. These could include Photovoltaic Panels, Wind Turbines, Energy Storage, and Biofuels. Deeper in the network, subsequent issues focus on the location of these installations to reduce construction and operating costs, maximise efficiency, and minimise nuisance.

To summarise the trends observed in Figure 10, it is expected that such systems will be diversified to consist of many different elements, include an energy storage component, and be optimised using multiple criteria. Optimisation goals will include management systems, construction and operating costs, and system stability.

During Keyword Network Construction, a keyword co-occurrence analysis was based on data exported from Scopus, including both Author Keywords and Index Keywords. The combined data set was processed using VOSviewer version 1.6.20 to generate the network visualisation (Figure 10). Only terms that occurred in at least five publications were included in the analysis, while those with lower frequencies were excluded. VOSviewer automatically handled tokenization and text normalisation, and links between nodes represent co-occurrence relationships weighted by frequency. The resulting map illustrates the strength and density of associations between research topics in renewable energy domains.

4.2. Current Trends in Synergistic Computing for Sustainable Energy Systems

Sustainable energy systems are becoming increasingly common and their importance is steadily growing. However, because of their complexity, they require automation for more efficient and effective use. The selection of such a system’s parameters is also not straightforward, and using the right tool can help find the optimal solution. Current trends include

1.

Hydrogen storage—Demand for this type of energy storage is growing significantly. This is because of the ability to store large amounts of energy for long periods while simultaneously reducing the system’s wear rate compared to battery storage. However, they require significantly more advanced automation mechanisms to increase their efficiency.

2.

Energy system optimisation—The use of tools such as Genetic Algorithms, Neural Networks, and Particle Swarm Optimisation to design and manage the energy system, allowing for the forecasting of demand and production capacity.

3.

Combining different energy sources—Due to the varying constraints of individual energy sources, combining several types into a single hybrid power plant is beneficial. Most often, the constraints of one source do not overlap with those of another, increasing the likelihood of ensuring a continuous energy supply. However, managing such infrastructure requires the use of advanced control mechanisms. Neural Network-based solutions are often used here.

4.

Grid automation and management—Not only should the power plant itself be automated, but, due to the growing share of renewable sources in the overall energy mix, power grids should also be automated. However, it is crucial that they act proactively, anticipating changes in energy demand and supply, rather than reacting only to observed changes.

5.

The use of bioenergy is considerably less significant compared to wind turbines, photovoltaic panels, or energy storage. However, they are a crucial complement to the system. Biomass can be used on its own as a heat source or as a component of biofuels. Computational tools can help test the quality of the fuel or optimise its extraction. It can also be used as an electrical source, further stabilising the energy supply system.

4.3. Current Trends in Optimization Techniques

As mentioned numerous times in earlier sections of this article, effective use of infrastructure requires the use of advanced optimisation systems. The most frequently used in the articles analysed include

1.

Genetic Algorithms—Undoubtedly the largest group. They are widely used in the design phase of installations to correctly select the parameters of each component and their location. In some cases, they are used to optimise Neural Networks.

2.

Neural Networks—The second largest group of solutions after Genetic Algorithms. However, they do not compete with them, as they are most effective during the infrastructure operation phase. They are used to forecast energy production and make energy balance decisions.

3.

Particle Swarm Optimisation—Used to optimise energy system parameters, such as the placement of photovoltaic panels and wind turbines.

In addition, attempts can be made to find new solutions. The most important include

1.

Hybrid Solutions, such as the integration of Genetic Algorithms with Neural Networks, Particle Swarm Optimisation, or Firefly Algorithms, offer superior convergence speed and robustness against local minima compared to single-algorithm methods. Nevertheless, they demand greater computational resources and careful parameter tuning, which may constrain their real-time deployment. Thus, hybridisation should be guided by the scale of the problem, the available computational capacity, and the need for adaptive learning during operation, for example, the Hybrid Firefly Genetic Algorithm [25].

2.

Using new algorithms to search for completely new solutions, often inspired by biology, for example, Sunflower Optimization [22].

4.4. The Most Frequently Analysed Issues

In the area of Synergistic Computing for Sustainable Energy Systems, the most frequently analysed issues were as follows:

1.

Optimal System Operation—Designing tools for automatic control of the energy system. This takes into account both current installation parameters and possible changes. When multiple energy sources are combined, it is also necessary to decide the degree of utilisation of each.

2.

Forecasting—This is particularly important for wind and solar energy, as these sources are highly sensitive to weather changes. Forecasting allows the grid to prepare for upcoming changes. The key here is not long-term forecasts, but detailed forecasts for the next few hours. This is related to the need to launch processes that require time to start (e.g., drawing energy from hydrogen storage).

3.

Cost and Efficiency—This is often subject to multicriteria evaluation. Cost includes everything from construction costs to maintenance and operation costs. The selection of the most cost-effective components, such as turbines, panels, and energy storage units, their parameters, layout, and the cost of cabling between them are also considered. The stability of the system may be an additional criterion.

4.

Energy management—A challenge with green energy generation is the inability to influence the system’s production capacity. It depends on the weather. Therefore, developing a management strategy is crucial: whether it should be fed into the power system at a given moment, stored in a battery system, or hydrogen storage, or whether the system’s power is declining and energy storage must be activated.

5.

Sensitivity analysis—Examining the impact of various parameters on the cost, efficiency, and stability of energy systems.

4.5. Noticeable New Trends

Despite intensive research in this area, gaps in knowledge remain. Not all of them seem equally important. Some, however, remain relatively unexplored but seem to hold great potential for the future:

1.

Cybersecurity—An extensive infrastructure requires the transmission of control information in a secure and reliable manner. Information can be centralised, but thanks to blockchain technology, it can be decentralised. This could improve the security of the entire system.

2.

Adaptive Algorithms and Machine Learning—The Neural Networks used are often trained only in the design phase. However, environmental conditions can change over time. Therefore, it would be useful to develop solutions that adapt to these changes.

3.

Bioenergy and Biofuels—Although the article addresses the issue of biomass and its use, this topic remains under-researched. However, it appears to be very promising, as it could bridge the gap between traditional fossil fuels and renewable energy. The development of new simulation and management tools can improve this technology.

4.

Energy storage systems—A literature review has shown that energy storage systems are an important component in stabilising the system and improving its efficiency. Hydrogen storage is being intensively tested, but it is not the only solution. The storage of molten salt heat is promising but is poorly proven. Further research could improve the efficiency and effectiveness of this solution.

5.

Optimisation of energy infrastructure—Many studies that focus on the location of components of the power plant consider only economic costs as a criterion. However, considering environmental and social costs and constraints could reduce the social costs associated with the investment.

6.

The lower number of bioenergy-related publications comes from a real research gap, as relatively few studies apply advanced optimisation algorithms to this domain. This suggests that bioenergy offers promising opportunities for future computational research rather than indicating a methodological limitation of the present review.

Because these research areas address complementary dimensions of system reliability—technical, operational, and security-related—they should be developed in parallel rather than hierarchically prioritised.

4.6. Comparative Synthesis of Optimization Methods

To complement frequency-based analysis,

Table 7. Comparative summary of optimization algorithms by application domain, objective functions, and reported performance outcomes.

Algorithm	Main Application Domain	Typical Objective Functions	Reported Metrics/Improvements	Remarks/Advantages
Genetic Algorithm (GA), e.g., [17,18,21,26,37]	Hydrogen storage, wind layout, hybrid system sizing	Cost minimization, system reliability, power balance	10–25% approx. LCOE reduction; 15–30% improvement in energy utilization efficiency	Versatile; effective in discrete + continuous problems; moderate computational cost.
Particle Swarm Optimization (PSO), e.g., [56,61]	Solar irradiance forecasting, PV–wind hybrid control	RMSE minimization, stability enhancement	RMSE ↓ 8–20%; grid fluctuation ↓ 15–25%	Fast convergence; sensitive to parameter tuning.
Neural Networks (NNs) with GA tuning, e.g., [61]	Forecasting, energy management	Forecast accuracy (RMSE/MAE)	RMSE ↓ up to 30% compared with classical models	Enables adaptive control; requires sufficient data.
Modified Multiverse Optimizer (MVO), e.g., [19,77]	Wind/solar hybrid optimization	Cost–${\text{CO}}_2$ trade-off, efficiency maximization	Cost ↓ 12–18%; ${\text{CO}}_2$ ↓ 10–15% vs. GA baseline	Superior in high-dimensional problems; higher computational demand.
Hybrid Firefly–GA [25]	Multi-source microgrid optimization	LCOE and fuel cost reduction	LCOE ↓ 9–14% compared with GA and PSO	Combines GA global search with Firefly local refinement.
Grey Wolf Optimiser (GWO), e.g., [61]	Wind farm layout and PV–battery configuration	Wake-loss minimization, cost reduction	Wake loss ↓ 10–12%; payback period ↓ 8%	Good balance between exploration and exploitation.
Cuckoo Search (CS), e.g., [25,35]	PV–biomass hybrid systems	Cost–${\text{CO}}_2$ Pareto optimization	Total cost ↓ 10–13%; ${\text{CO}}_2$ ↓ 9%	Simple implementation; slower convergence.

Notes: Reported improvements are aggregated from representative studies (2015–2024) within the analysed corpus. Metrics: LCOE = Levelised Cost of Energy, RMSE = Root-Mean-Square Error. Values indicate typical ranges rather than universal benchmarks.

presents a comparative overview of the optimisation algorithms that are most frequently used in renewable energy domains. The comparison includes their typical application tasks, objective functions, and reported performance indicators from representative studies. Although the specific quantitative gains vary by case, consistent patterns can be observed. Genetic Algorithms (GAs) and Particle Swarm Optimisation (PSO) remain the most widely applied according to their versatility and robustness, whereas more recent methods such as the Modified Multiverse Optimiser (MVO) and Hybrid Metaheuristics demonstrate incremental performance improvements in complex, nonlinear environments.

5. Conclusions

In summary of the period studied, in terms of Synergistic Computing for Sustainable Energy Systems, there is a clear need for systems of this type. This is dictated by climate change and the gradual depletion of natural resources. However, manual control of such systems is too complex, so there is a trend toward increasing automation and optimisation using advanced algorithms. The most commonly used are as follows:

1.

Genetic Algorithms;

2.

Neural Networks;

3.

Particle Swarm Optimization.

These are crucial for the development of renewable energy, including hydrogen storage, wind turbines, solar panels, and bioenergy.

Integrating renewable energy sources into the power grid presents a new challenge in the energy sector. Optimising their use requires the use of advanced automation tools. The energy storage and distribution strategies adopted must be further developed and adapted to climate change. Therefore, these will be topics for future research. Based on a review of the current literature, future research is expected to focus on the following.

1.

Cybersecurity;

2.

Adaptive algorithms and machine learning;

3.

Bioenergy and biofuels;

4.

Energy storage systems;

5.

Energy infrastructure optimisation.

In each of the above examples, modern algorithms will play a significant role.

The article also highlights the need for further research, as despite intensive research, there are still relatively little-explored areas that could hold significant potential in the future. These areas could include new algorithms currently in the testing phase.