Green Energy Management in Manufacturing Based on Demand Prediction by Artificial Intelligence—A Review

Abstract

Energy efficiency in production systems and processes is a key global research topic, especially in light of the Green Deal, Industry 4.0/5.0 paradigms, and rising energy prices. Research on improving the energy efficiency of production based on artificial intelligence (AI) analysis brings promising solutions, and the digital transformation of industry towards green energy is slowly becoming a reality. New production planning rules, the optimization of the use of the Industrial Internet of Things (IIoT), industrial cyber-physical systems (ICPSs), and the effective use of production data and their optimization with AI bring further opportunities for sustainable, energy-efficient production. The aim of this study is to systematically evaluate and quantify the research results, trends, and research impact on energy management in production based on AI-based demand forecasting. The value of the research includes the broader use of AI which will reduce the impact of the observed environmental and economic problems in the areas of reducing energy consumption, forecasting accuracy, and production efficiency. In addition, the demand for Green AI technologies in creating sustainable solutions, reducing the impact of AI on the environment, and improving the accuracy of forecasts, including in the area of optimization of electricity storage, will increase. A key emerging research trend in green energy management in manufacturing is the use of AI-based demand forecasting to optimize energy consumption, reduce waste, and increase sustainability. An innovative perspective that leverages AI’s ability to accurately forecast energy demand allows manufacturers to align energy consumption with production schedules, minimizing excess energy consumption and emissions. Advanced machine learning (ML) algorithms can integrate real-time data from various sources, such as weather patterns and market demand, to improve forecast accuracy. This supports both sustainability and economic efficiency. In addition, AI-based demand forecasting can enable more dynamic and responsive energy management systems, paving the way for smarter, more resilient manufacturing processes. The paper’s contribution goes beyond mere description, making analyses, comparisons, and generalizations based on the leading current literature, logical conclusions from the state-of-the-art, and the authors’ knowledge and experience in renewable energy, AI, and mechatronics.

1. Introduction

Energy efficiency in production systems and processes is currently a key research problem, especially considering the wide implementation of the Green Deal and Industry 4.0/5.0 paradigms in the conditions of rising energy prices. Previous research on improving the energy efficiency of production based on consumption analysis and balancing demand and supply using artificial intelligence (AI) has yielded promising solutions, and the digital transformation of the industry towards green energy is slowly becoming a reality. New rules for production planning using digital twins (DTs), the optimization of the use of the Industrial Internet of Things (IIoT), industrial cyber-physical systems (ICPSs), the ongoing implementation of the sixth generation (6G) networks, and more effective use of production data and their optimization using machine learning (ML) bring further opportunities for sustainable, energy-efficient production. The synergy of the above solutions will reduce the severity of the observed environmental and economic issues regarding the need to improve energy consumption, the accuracy of energy forecasts, and production efficiency. There is also a growing demand for Green AI technologies used to create sustainable solutions, reduce the impact of AI on the environment, and improve forecasting accuracy, including in the area of the optimization of electricity storage. Production control policies/strategies are increasingly based on dynamic models (also: digital twins (DTs)) of entire production lines and their individual devices in order to optimize/reduce the energy consumption of production systems without reducing their performance. Various solutions are already being applied today, e.g., to reduce global energy consumption or to avoid power peaks during the operation of production systems by optimizing the use of the equipment at the periphery of production lines. This takes into account both AI/ML-based energy consumption models, operational constraints, and the actual processes performed by peripheral equipment (including future processes reflected by their DTs) [1,2,3]. So far, this approach has yielded a reduction in energy consumption of around 7% [1], which, with current high energy prices, is a significant reduction for which the effort is worthwhile even in a small company. Public awareness of saving energy and minimizing carbon dioxide (CO₂) emissions is also growing, and the governments of many countries have introduced regulations regarding the manufacturing industry in these areas [4].

Demographic factors translate into customer opportunities and needs, shaping the market to a large extent. Poland’s senior citizens’ policy, for example, can initially be assessed positively, although the scale and pace of change are too small in relation to needs, and some key changes are impossible (e.g., raising and equalizing the retirement age for men and women). On the other hand, solutions that are almost impossible to implement have a high probability of coming into force with a guaranteed pension.

Matching energy consumption to needs while depleting some resources are major challenges, and saving energy resources and improving the efficiency of their use are important goals for modern production and improving living standards. Furthermore, the implementation of modern industrial technologies is often equated with their energy efficiency, and inefficient energy use is associated with a decrease in income [5,6,7].

The aim of this study is to systematically evaluate and quantify the research results, trends, and impact of research dedicated to energy management in manufacturing based on AI-based demand prediction.

The article is a review, but its contribution is significantly extended as it goes beyond mere description, making analyses, comparisons, and generalizations based on the leading current literature on the subject, logical conclusions from the state-of-the-art, and the authors’ knowledge and experience in the areas of renewable energy sources, artificial intelligence, and mechatronics.

The research gaps identified so far in the area of green energy managed by artificial intelligence, which our article seeks to fill, cover several key areas. Like most revolutionary changes, they are struggling with resistance due to a lack of public awareness and local interests (e.g., farmers’ resistance to the Green Deal in the EU), the problems of high upfront costs, and the need to attract a sufficiently large number of specialists and users. Few advanced artificial intelligence algorithms have been identified that can optimally and cost-effectively manage distributed renewable energy sources, such as solar panels and wind turbines. There is a need for more accurate methods of forecasting energy production from renewable energy sources that are useful in the constrained budgets of households and energy cooperatives, taking into account dynamically changing weather conditions. There is also a need to better understand the interactions between different types of renewable energy sources and power grids to ensure system stability. Another challenge is the integration of energy storage systems with RES artificial intelligence management, which requires innovative solutions for demand forecasting and the efficient storage of resources. There is also a need to develop communication standards and protocols that will enable the smooth operation of the various systems managed by artificial intelligence. Another challenge is cyber security, as the growing number of devices connected to the network increases the risk of attacks. Understanding the impact of regulatory policies on the cost-effectiveness of implementing artificial intelligence in green energy management remains under-researched. More research is needed on the social and economic aspects of implementing artificial intelligence in green energy to ensure acceptance and social equity. It should be noted that all of the above-mentioned gaps require intensive interdisciplinary research to fully meet the challenges of the future. Looking at the problems in the implementation of electric cars, we can conclude that technological excellence or social acceptance alone is not sufficient for their widespread adoption.

The overall structure of the manuscript is as follows: Section 2 recalls and summarizes the state-of-the-art in sustainable manufacturing system design, then in Section 3 we move on to pose the research questions, define the dataset, and review the research methods. The results of the review are summarized in Section 4, and Section 5 is a discussion of these with reference to specific examples, limitations of the research to date, and proposed key directions for further research. The article concludes with conclusions relating to the extent to which the objective set in the Introduction Section has been achieved.

2. Sustainable Manufacturing System Design

The research, planning, and development of sustainable production systems are considered one of the most effective solutions to achieve sustainable development, including the following:

• Improving system efficiency and productivity as well as cost balance;
• Minimizing the impact on the environment, also indirectly by reducing energy consumption and CO₂ emissions;
• Reducing production losses and the amount of waste [4].

To prepare the above-mentioned plan, the following is necessary:

• The measurement of energy consumption, CO₂ emissions, and other factors using various energy sources (both for powering machines, lighting, heating, water heating, cooling, etc.: crude oil, oil, solar energy, etc.);
• A mathematical model taking into account real-world constraints, including technological, economic, and ecological ones, with the objective function (with weights specific to a given enterprise/industry) minimizing costs, energy consumption, and CO₂ emissions within the production system itself and the entire enterprise;
• Determining the uncertainty (also: randomness characteristics, if necessary) of input parameters using, e.g., a fuzzy model (including directed fuzzy numbers);
• Analyses and decisions regarding the number of machines/devices for each stage of the production process in connection with the amount of energy and material flow and their DT;
• The construction of an optimizing AI/ML model;
• The validation of the developed system model based on real data, e.g., case study;
• The research and verification of applicability [4,8,9,10,11,12,13,14].

Previous research and the literature have observed the need to develop a systematic approach to designing a sustainable production system that would meet the following requirements:

• Inclusion at the earliest possible stage of planning and design;
• Allows to fully explore the inclusion of ecological (e.g., energy consumed) and economic aspects;—It involves the operations of the production system;
• Uses computer simulation tools;
• It would provide a framework for a hybrid approach to optimization;
• It would allow for uncertainties, multi-goals, and discrete event simulations while ensuring the required accuracy and assessing the performance of the production system [4,8,9,10,11,12].

Sustainable production covers three areas that must be taken into account during analyses: economic, environmental, and social responsibility. Not only customer satisfaction must be important and measurable, but also environmental protection issues and the awareness of the “environmental friendliness” of the products themselves and their production technologies and their shortages on the market. In previous studies, the MOGWO algorithm performed better than the NSGA-II algorithm and the exact solution method (GAMS) [15]. Of course, the modeling can be extended to the optimization of green supply chain management (GSCM).A review of 159 GSCM optimization articles (published 2017–2022) showed that the main solutions included mathematical modeling, multi-body optimization, and modeling/solver tools. The main area of optimization was the minimization of greenhouse gas emissions (27.67% of the publications), and the most popular choices were mixed integer linear programming (32.08% of the publications) and game theory modeling (18.87% of the publications). In multi-objective optimization, the ε constraint method was most often used, and in metaheuristic optimization, evolutionary algorithms. The most popular solver was CPLEX, and the most popular modeling/solver combination was GAMS/CPLEX [16].

The growing world population and the resulting demand for energy, products, food, and transport have led to not only an increase in waste generation and serious challenges in its disposal, but also to efforts to reduce its use and thus lower production costs. Moreover, traditional disposal methods (incineration and land filling) harm the environment and threaten human health and safety. This results in an urgent need to balance the management of energy, raw materials, and waste, in particular, the reuse of waste as a valuable resource (also as an element of the 6R strategy).The need for alternative approaches is driving a paradigm shift towards accepting waste as a valuable component of economic prosperity and environmental sustainability. Continued research and technological advances will help unlock its full potential in supporting a circular economy and driving sustainable development around the world [11,12,13,14,15,16,17].

3. Materials and Methods

3.1. Dataset

In this bibliometric analysis, we focus on examining the research picture in the area of energy-saving strategies and innovations regarding green energy management in manufacturing based on AI-based demand prediction. For this purpose, we use bibliometric methods as analytical tools in scientific publications. To better focus our research, we formulated a number of research questions to help discover key aspects of the study area, including the following:

• Research Question 1 (RQ1): the evolution of research topics/issues over time;
• RQ2: the geographical distribution of research/publications;
• RQ3: the authors and publications with the greatest influence;
• RQ4: the identification of cooperation networks between researchers and institutions;
• RQ5: the topics that may shape future research agendas.

The study was conducted in accordance with some points of the PRISMA guidelines for creating a review with bibliographic analysis, namely: item #3 Rationale, item #4 Objectives, item #5 Eligibility criteria, item #6 Information sources, item #7 Search strategy, item #8 Selection process, item #9 Data collection process, item #13a Synthesis methods, item #20b Results of syntheses and item #23a Discussion. In response, it seems crucial to have the most comprehensive understanding of the current state of research, industry practices, efforts, and future research directions in the pursuit of sustainable energy solutions. The analysis and interpretation of bibliometric data can make a significant contribution to the ongoing discussion on this topic and build a more solid foundation for further analyses and research (Figure 1).

3.2. Methods

We used the Biblioshiny component from the Bibliometrics Rv.4.1.3 package as a tool for conducting bibliometric analyses. The rationale for this tool is that it is tailored specifically to bibliometric and scientometric research, offering categorization into conceptual structures, and providing insight into authors, documents, and sources. A diverse range of results is presented using charts and informative tables with analysis functions.

In future articles, we plan to extend the tools to the use of Dimensions.ai (https://www.dimensions.ai/) and Open Alex (https://openalex.org/) or Google Scholar (https://scholar.google.com/) (as there is no knowledge of how many articles are omitted from the current reviews), and then use a quality assessment tool as a part of the eligibility criteria or SLR protocol with these criteria (such as the Rationale–Cogency–Extent criterion or GRADE for systematic reviews, etc.).

4. Results

We selected research articles from two major databases: Web of Science (WOS) and Scopus due to their broad coverage of research findings. Both of the aforementioned databases contain detailed citation data, which is particularly beneficial when conducting extensive bibliometric analyses and assessing the impact of research findings. In creating a sophisticated search query tailored to our research objectives, we applied filters to ensure the selection of the relevant literature. To facilitate bibliometric analysis, we limited our search to original articles in English. We then manually reviewed the articles again, excluding articles to align with our research objectives, resulting in a final sample size (Figure 2). WoS search was carried out using Topic (searches: title, abstract, and keyword plus), and Scopus search was performed using article title, abstract, and keywords. Searches using the keywords ‘green energy’ ’manufacturing’ ‘demand prediction’ ‘artificial intelligence’ and related did not yield retrieved papers in either the WoS or Scopus databases. Further reductions in the number of keywords yielded positive results for ‘green energy’ manufacturing ‘artificial intelligence’ and related (17 in the WoS database and 9 in the Scopus database).

We began the bibliometric analysis by conducting descriptive statistics to understand the characteristics of the scientific publication dataset, including prominent authors, research communities, thematic clusters, and emerging trends in the subject area under study. This allows us to identify the evolving vocabulary and research highlights. Examining temporal trends, in turn, allows us to see the changes in the focus of research over time and the type and dynamics of an area, including the categorization of publications into thematic clusters, including a picture of the interconnectedness of research topics. This will allow easier identification of key themes and subdomains. The years of publication, number of citations, and authorship patterns will provide a quantitative overview of the selected literature, and citation analysis will provide an assessment of the impact and influence of the publications, demonstrating their scientific relevance.

In the WoS database, 17 publications from 1990 to 2024 were observed (Figure 3), including 9originals (52.94%), 5reviews (29.41%), and 3proceedings (17.65%). A total of 25 authors with 60 affiliations wrote them (none of the authors and affiliations were predominant). All the publications were in English. To date, the publications have accumulated 375 citations, i.e., an average of 22.06 citations per article, which is quite high, so the articles are widely read and frequently cited (Figure 4).

The three most popular application areas for AI in green technology management are energy fuels, green sustainable science technology, and electrical engineering (Figure 5).

The largest number of studies have been carried out in China and Taiwan (Figure 6).

Source impacts are presented in

Table 1. Journals and conference proceedings according to source impact (WoS and Scopus) (n = 26).

Journal	Number of Articles [-]
Sustainability [18]	1
Journal of Composites Science [19]	1
Applied Sciences [20]	1
Journal of Electrochemical Science and Engineering [21]	1
International Journal of Green Energy [22]	1
International Conference on Advances in Green Energy (ICAGE) [23,24]	2
Solar Energy [25]	1
4th International Conference on Artificial Intelligence and Evolutionary Computations in Engineering Systems (ICAIECES) [26]	1
Green Energy & Environment [27]	1
Biomass Conversion and Biorefinery [28]	1
ACS Applied Materials & Interfaces [29]	1
Futures [30]	1
IEEE Sensors Journal [31]	1
Sensors [32]	1
Journal of Experimental & Theoretical Artificial Intelligence [33]	1
Applied Energy [34,35,36]	3
Machine Learning and Computer Vision for Renewable Energy [37]	1
Virtual and Physical Prototyping [38]	1
Science of the Total Environment [39]	1
International Conference on Artificial Intelligence and Computer Vision (AICV2020) [40]	1
Engineering Applications of Artificial Intelligence [41]	1

.

None of the most proliferative institutions and research organizations were identified. Despite this, more and more scientists and research centers are engaging in activities related to sustainable development in this area. This is performed both in the circulation of scientific publications and through social networks to spread concepts and support motivation for sustainable development. Even research issues that have not yet been identified can become popular topics discussed on social networks (especially if they are controversial).Research so far has focused on detecting and tracking the above-mentioned areas based on historical data, which is not always effective—it is necessary to predict emerging topics supported by AI. This is because not all popular controversial issues are linked to historical data. This may constitute an important research gap; hence, AI plays an important role in detecting problems at very early stages of development [18]. The development of alternative propulsion systems, including the use of hydrogen to replace fossil fuels in transport applications, seems to be a priority now. Three leading on-board hydrogen storage technologies: (1) compressed gas storage, (2) cryogenic liquid storage, and (3) solid storage can be subject to performance evaluation using AI models [19]. Green hydrogen and ammonia have emerged as promising sources of clean, sustainable, and low-carbon energy in the green energy transition, but their production and storage challenges hamper their widespread use.AI and additive manufacturing (AM and 3D printing) can be used in this area, including AI-assisted catalyst design and 3D-printed reactors [20]. Non-lithium-ion batteries (Na, K, Mg, Zn, Al, ion batteries, etc.) are becoming an increasingly popular alternative thanks to the design of new materials with desirable properties to support a faster transition to a green energy ecosystem.AI-based screening has proven to be a tool for accelerating the discovery of active materials and can be useful to reduce the time and capital expenditure needed to discover advanced materials and develop a final product viable for commercial applications [21]. Computational tools significantly affect the accuracy of horizontal axis wind turbine (HAWT) and vertical axis wind turbine (VAWT) models; hence, the proper selection of the AI model and the number of its experimental validations are crucial [22]. For photovoltaic (PV) systems, the key problem solved by AI is the lack of forecasts for all the areas in terms of the power of the PV system connected to the grid—solving this problem makes it possible to use the full potential of solar energy. As predictive models, models based on an artificial neural network (ANN) trained on real data with regression mapping were proposed, which for next-day forecasting gives low mean square error (MSE) values ranging from 0.019 to 0.025 [23]. AI is also used in preventive or predictive diagnoses to avoid various types of damage, especially electric motors, 80% of which are used in industrial applications [24]. Moreover, the early detection of defects in the production of solar cells is important for stable energy production from PV. Automatic detection models (e.g., SeMaCNN) based on deep learning are used here for the classification and semantic segmentation of electroluminescent images for anomaly detection and solar cell quality assessment with an accuracy of 92.5% on a subset of 1049 manually annotated images [25]. AI-driven intelligent factory energy planning, including wireless power transfer from IoT devices, shows progress in energy supply and demand balance [26]. AI has impacted standards for the research and development of new materials for energy and the environment, including membrane design and discovery [27]. Even used cooking oil (WCO) can be recycled, e.g., to produce solid alcohol fuel (SAF) as a new fuel or green energy source, and their properties predicted by machine learning (ML) support vector regression (SVR) with an accuracy of up to 0.9214 [28]. AI applications in IoT will also accelerate the development of ultra-low power devices for future advanced technologies, e.g., a capacitive piezotronic sensor for touch sensing operating at very low voltage under near-zero readout bias conditions with a constant response over a wide voltage range. This allows for the development of new families of ultra-low power capacitive piezotronic/piezophototronic devices for touch sensing [29]. AI supports the development of a sustainable, long-term power planning model with compromise solutions in the area of seasonal and regional factors and air pollution with flexible management in peak loads [30]. AI-based energy prediction is becoming a key factor in fuel cell electric vehicles (FCEVs), where maintaining the appropriate energy production intensity is necessary to improve the efficiency of energy production and prevent it from falling into a state of inefficiency due to a lack of energy. This helps maintain the optimal state of charge (SOC) [31]. AI support even covers such simple applications as controlling the wearing of protective clothing or a protective helmet [32]. To solve optimization applications in energy acquisition and consumption, the Bird Swarm Algorithm (BSA) was proposed, implementing four search strategies associated with five simplified rules [33]. Improving energy efficiency is becoming necessary even in such large agglomerations as Beijing (China), optimizing the effects of energy reuse in urban and rural households using AI depending on the type of services/goods and energy reflection for each of them [34].

Accordingly, the Scopus search results were as follows: nine documents have been published since 2019 (Figure 7). Seven publications were observed that were not previously listed in the WoS database search results [35,36,37,38,39,40,41].

Most of the publications were reviews (Figure 8).

The main research areas were engineering, computer science, and chemical engineering (Figure 9).

Most studies were conducted in India, the USA, and African countries (Figure 10).China and Taiwan were not in the majority here as in the WoS database. It is clear from the above that the source (despite the same keywords) is important for the results of the review.

Based on the search results, there is no doubt that the primary goal of implementing green energy is to support economic development, improve energy security, increase access to energy, and fight global climate change. Sustainable development is achieved through the wider use of renewable energy sources and providing residents with access to affordable, reliable, and modern energy. Automation, additive manufacturing, AI, and other technologies can facilitate the transition to green energy and, in turn, to a cleaner environment by gradually improving the quality of the environment, controlling and monitoring infrastructure, and upgrading hardware and software [35]. AM supports the production of environmentally friendly (clean and renewable) energy generation and storage devices with geometric precision and very high operational efficiency. However, this is still not part of the mainstream of low/zero emission energy equipment production, which should be quickly addressed by leveraging the synergies between AI and AM [36]. AI can optimize not only energy production, but also water treatment; equalize access to resources and services; and, where possible, improve prosperity and quality of life. Therefore, AI support may primarily set priorities and even directions for actions aimed at facing climate change and protecting biodiversity [37]. This may require rethinking development directions, adapting or developing new development strategies to support economic productivity based on technology, innovation, and smart digital services, rather than on traditional sectors: from agriculture to traditional industrial production [38].

Based on the above results of the literature review (Figure 11), they do not distinguish authors who have made significant contributions to this field. Both the number and distribution of studies indicate that the researched area is at the beginning of its development, the leaders have not yet emerged, and many centers may become known with the results of their research. Therefore, this is an area of research that is a good debut area, where the fight for the upper hand is still ongoing.

Table 2. Compartment of three main research.

Feature	FedPT-V2G [35]	GGNet [36]	Multi-Node Load Forecasting [41]
Type	Security-enhanced federated transformer learning for real-time V2G dispatch with non-IID data	A novel graph structure for power forecasting in renewable power plants considering temporal lead-lag correlations	Multi-node load forecasting based on multi-task learning with modal feature extraction
Technologies	Deep learning Federated learning Regularization to align local models	Granular-based graph neural networks Two-dimensional convolutional neural network Graph attention network	Gated temporal convolutional network
Algorithm	Proximal algorithm and Transformer model to handle Non-IID data in V2G tas	Graph structure can adapt to the lead-lag characteristic arising from the air flow, allowing it to dynamically capture the effects of lead and lag among RPPs.	IGTCN module is designed to extract the coupling features from different node loads
Dataset	End-to-end learning based on historical data to make V2G decisions in real-time	Real-world datasets, with wind power plants and photovoltaic power plants	Data from the New Zealand distribution network and AEMO
Application	Load shifting and PV self-consumption under diverse uncertainties	Power forecast for each renewable power plant (RPP) in the renewable energy clusters	Multi-node load forecasting in the power system
Performance	Achieve similar performance to centralized learning on both IID and Non-IID data	The proposed graph structure can reflect the lead-lag characteristics among RPPs caused by the atmospheric flow, obtaining better correlation representations among RPPs.	The multi-task deep neural network exhibits the accuracy of multi-node load forecasting
	Accuracy for balanced datasets: 98.93%; imbalanced datasets: 92.20%	Forecast steps decreased on average by 48.95% and 18.75%, respectively	MAPE decreased by 17.04% and 3.92% in non-aggregation and aggregation situations, respectively

shows recent leading research in the aforementioned area.

5. Discussion

The main problem of our results is the great diversity and number of topic areas in which solutions exist. This results from the connections between demographic problems and senior policy with the pro-future Green Deal and sustainable development policy (including sustainable production systems and green supply chain management), which were not obvious a decade ago, but are now completely clear. Of course, this results from a number of simplifications, such as changing consumer attitudes in aging societies or new trends in energy management in production and services, with particular emphasis on artificial intelligence-based demand forecasting related to the pandemic or the war in Ukraine. Similarly, waste disposal is becoming closely linked to energy management in the production process due to rising energy costs. Therefore, demographic and awareness changes in societies can affect customers’ priorities (e.g., by placing a high value on their sense of security), making the profitability factor only one of many important factors in decision making (cf. changes in the popularity of electric cars and attempts by China to expand into European markets in this area).Hence, in the review, references are made to alternative propulsion systems and their technical details or the recycling of waste cooking oil (WCO) to produce solid alcohol fuel (SAF) as a new fuel or green energy source, which is an important alternative given the amounts of such oil used in the economy. It points out that the growing importance of remote work and shortening the working day (to 6 h) and the working week (to 4 days) means that classic jobs are losing their importance, and the fact that AI can indirectly optimize energy consumption for water treatment is gaining importance. This is due to the fact that wealthier people are more willing to think about sustainable development and the future of future generations, because equal access to resources and services and improved well-being and quality of life allow them to take care of broader perspectives and values than just surviving another day or month.

Production requirements (including in the area of the individualization of production), rising energy costs, increased emphasis on environmental protection, and changes in consumer preferences promote ecological production. This requires a new economic, logistic, and production approach, as well as intelligent solutions based on AI in areas such as transport, management, optimization, and prediction. It is known that the demand for energy in IoT applications will increase and intelligent energy management will become an increasingly important issue due to the need for energy savings and the possibility of an energy crisis. We are moving towards intelligent structures enabling the pooling of reservations and increasing energy efficiency in IoT-based smart industries [26].

Thanks to advances in computational analysis and access to publications, there have been advances in bibliometric analysis and scientific mapping. Scientists and engineers, as well as governmental and non-governmental organizations, can determine the directions of development; the strength of the influence of researchers, research centers, and countries; and the importance of individual research, improving the broader understanding of individual research fields. This will facilitate further analysis despite the increase in the number of scientific publications in recent years. The results of this study also indicate a growing awareness of the importance of energy conservation and sustainable development, which is consistent with global efforts to mitigate the impact on the environment and optimize the use of resources, but also attempt to reverse the current changes in the human natural environment related to industrialization [42,43,44].

In light of the current results, it is difficult to determine the impact of the language barrier. Due to the possible advancement of research, e.g., in China or India, many scientific articles may be published in languages other than English, and these works may not be included in the databases. The problem discussed here concerns the insufficient representation of non-English publications/research in bibliometric analyses. However, it should be noted that the mainstream of publications in exact sciences and technology focuses on publishing research results in international journals (usually in English), also as part of research by interdisciplinary and international teams. One of the key activities highlighted in the literature is the optimization of techniques. The strategies proposed here should be verified each time based on reliable long-term data and DT and AI/ML models (and, if possible, on a test stand) for typical current and planned control schemes and the use of the production line and its individual devices [1,2,3]. This will minimize energy costs related not only to the current but also to future (potential) types of production activities and prepare for their implementation, including minimizing costs.

The research, planning, implementation, and development/modernization of a sustainable production system will be an increasing challenge due to increasingly strict regulations requiring not only energy-saving production activities but also the reduction in environmental waste generated during production while maintaining a cost balance favorable to the company. In lean management conditions, this will require the use of precise measurement and monitoring techniques and advanced computational models based on AI, taking into account uncertainties, risk factors, and the diverse scenarios of the use of production lines (including changes in recipes resulting, for example, from the temporary unavailability of some materials) [4,8].

The integration of IoT with AI already offers numerous benefits, such as increased efficiency, reduced costs, and a greater potential for sustainability and development, as well as real-time or near-real-time monitoring and control. Based on knowledge, experience, and data-driven automation, operators today can minimize unnecessary energy consumption, mitigate operational inefficiencies, and optimize production efficiency, thereby increasing the overall energy efficiency of individual processes. The importance of modernizing equipment and ensuring continuous and secure access to data is also growing. This makes achieving a dynamic balance between energy demand, energy production efficiency, and environmental sustainability possible, but requires a comprehensive, interdisciplinary approach using researched tools and science-based methodologies. A key element, in addition to optimizing processes and improving energy management systems, is the implementation of training programs for employees—this helps to explain and promote energy efficiency [44,45,46].

5.1. Limitations of Previous Studies

A limitation of the research to date may be its focus on two databases in English. They may not cover all of the relevant literature on the subject outside of the global circuit. Key challenges and limitations in industrial green energy management based on demand forecasting and artificial intelligence include the following:

• Implementing AI and advanced demand forecasting systems requires significant investment in technology solutions and staff skills;
• Integrating AI-based systems with the existing production processes and legacy systems can be complex and time-consuming;
• Data quality and availability can be limited, and comprehensive data collection in production environments can be difficult—in addition, handling and processing large amounts of data for AI applications can create significant data privacy and security issues;
• Scaling AI solutions across different manufacturing plants and systems can be difficult due to varying levels of technological sophistication;
• AI/ML systems are difficult to understand and trust their predictions, including for critical manufacturing decisions;
• AI systems can consume significant amounts of energy;
• Rapidly changing market demands may exceed AI’s adaptability, leading to inefficiencies in energy management;
• The regulatory challenges for both the AI Act and green energy management can be complex and vary greatly from region to region/country to country;
• Achieving the economic benefits of integrating AI into green energy management can take time, making it difficult to justify the investment in the short term [47,48,49].

5.2. Directions for Further Research

The following key directions for further research can eliminate the current limitations and unlock new opportunities:

• Improved data collection methods, including advanced sensors and IIoT devices to improve the accuracy and detail of data collection;
• The creation of new ML algorithms tailored to complexity and specific needs;
• Hybrid models created by combining artificial intelligence with traditional forecasting methods to improve the reliability and accuracy of forecasts;
• Developing methods to increase the transparency and understanding of AI forecasts;
• Exploring ways to reduce the energy consumption of AI systems themselves;
• Developing a framework for integrating AI-based demand forecasting systems with the existing energy production and management systems;
• Strengthening cyber-security measures to protect against breaches and attacks on sensitive data;
• Exploring methods to ensure the effective scaling of AI-based systems [50,51,52].

Power outages and deficits will lead to the development of AI-based distributed power system condition assessment [53,54,55].

The number of cyber threats is increasing, and network intrusion detection systems (NIDS) cannot easily identify threats in many areas, such as the CAN bus system in cars. The proliferation of mobile intrusion detection system (OMIDS) to detect and categorize these threats with high classification accuracy is beneficial: for HCRL Car-Hacking, the VGG16 and XBoost classifiers achieved 97.8241% and 99.9995%, respectively, for classification results on five subcategories [56,57]. Model-predictive control (MPC) with a simple structure, good robustness and dynamic response, and hybrid control solutions based on a predictive model and traditional algorithms allows combining the advantages of nonlinear and linear control, simplifying operation and facilitating implementation [58,59]. This approach helps develop efficient, safe, and sustainable energy uses based on new production tools and wider use of platform chemicals and energy sources (e.g., biofilm-based) [42]. Assessing the impact of industrial technologies and strategies on decarbonization, the environment, and the economy is gaining importance allowing better optimization of complex production systems and assessing future impacts at the systems level [43].

6. Conclusions

The bibliometric analysis highlights the importance of energy-saving technologies in industry and the concerted efforts of researchers around the world to address this critical issue. By identifying key research areas, geographic locations, and emerging trends, the analysis provides information on future research directions and initiatives which aim to promote sustainability and efficiency. Of the 26 publications included in the review, most were in engineering and computer science, but the spread of disciplines is very large. Review articles predominate. No leading country, research institution, or researcher was observed. The best accuracy achieved by the model for balanced datasets was 98.93% for FedPT-V2G. Overall, the research provides valuable insights for researchers, policymakers, and industry stakeholders, informing future directions for research, collaboration, and the development of energy efficiency and sustainability strategies.

A key emerging trend in green energy management in manufacturing is the use of AI-driven demand forecasting to optimize energy consumption, reduce waste, and increase sustainability. AI’s ability to accurately forecast energy demand allows manufacturers to align energy consumption with production schedules, minimizing excess energy consumption and emissions. Research in this area is increasingly using advanced ML algorithms that can integrate real-time data from various sources, such as weather patterns and market demand, to improve forecast accuracy. This trend is significant because it supports the dual goals of reducing environmental impact and lowering operating costs, contributing to both sustainability and economic efficiency. In addition, AI-driven demand forecasting can enable more dynamic and responsive energy management systems, paving the way for smarter, more resilient manufacturing processes.

Thanks to this approach, further research can develop specific technological solutions better, more precisely, and more deeply, taking into account the innovations thus achieved, and their matching to needs and opportunities, socio-economic impacts, and the necessary regulatory framework. Continuously updating and enriching the flexible strategy will allow it to be better adapted to current, perhaps not yet fully identified, needs. The efficiency and sustainability of this will be better than a rigid, procedural approach.