Farrah Gazi Mohammed
| Ghada Ghalib Abdulwahab | Mahmood Hussein Mustafa*
© 2025 The authors. This article is published by IIETA and is licensed under the CC BY 4.0 license (http://creativecommons.org/licenses/by/4.0/).
OPEN ACCESS
Against the background of intensive urbanization, the growing cities are faced with considerable difficulties in the area of sustainability, but they are also promising fields of innovative development. This paper questions how artificial intelligence is used in renewable energy systems, focusing on the design of such systems, their implementation, and the empirical evaluation of smart energy management models specific to the kind of needs of the emergent urban environment in less developed countries. Our study, based on a series of AI-based experiments, includes structured surveys, field surveys, and advanced statistical reviews, proving that it is possible to achieve up to fifteen percent energy efficiency, up to twelve percent grid reliability, and user satisfaction, which is more than an order of magnitude higher when compared to traditional methodologies. This empirical assessment is a detailed overview of the latest literature that is indexed in Scopus in the period 2022-2025, which shows the current development and points out the significant gaps, especially in the social and organizational aspects. The systems theory, optimization theory, and principles of responsible artificial intelligence are the methodological basis of the approach, where the inquiry is both technically sound and ethically sound. Its results highlight the potential changes in artificial intelligence that can be applied in the provision of scalable and reliable management of renewable energy sources in resource-constrained urban settings. The conclusion of the paper is to suggest ways of integrating the policy, to make AI human-oriented, and to provide future research directions that will support sustainable urban development.
artificial intelligence, renewable energy, smart energy management, developing cities, optimization algorithms, smart grids, systems theory
1.1 Research problems
The cities in development experience acute problems related to the transition to the sustainable energy system, including intense urbanization, lack of infrastructure space, monetary constraints, and irregularity of renewable energy. The traditional approach to energy management can hardly be used to manage the complexity and variability of modern urban energy networks. There is, therefore, an urgent need to develop new scalable and contextually responsive solutions that can maximise energy production, transmission and consumption and still remain reliable and cost-effective. The potential of artificial intelligence in managing renewable energy is quite significant.
H1: The implementation of AI-driven smart energy management models significantly improves the efficiency, reliability, and scalability of renewable energy systems in developing cities compared to conventional management approaches.
1.2 Hypothesis testing approach
The research design in this study is a mixed-method research design, which combines both quantitative and qualitative types of evidence to provide a systematic critique of the research hypothesis. The approach includes:
According to the latest literature, the potential of artificial intelligence in the management of renewable energy resources is transformative in its potential (especially in the context of developing regions), as mentioned in Table 1.
Key findings include:
Table 1. Summary of the literature review (2022–2025, Scopus-indexed)
|
Ref. |
Region/Context |
Focus Area |
Key Findings |
Implications/Contributions |
Added Contribution of this Study |
|
[1] |
Global/ |
AI explainability, governance |
The research identifies 15 governance parameters and is segmented into technical categories. |
It provides a basis to elaborate methodology of research in AI and energy issues. |
The study transcends the regular principles of governance and, in fact, applies Responsible AI on a pilot project of real, expanding cities, which makes all things understandable and in line with the interests of everyone. |
|
[3] |
Africa, Asia, Lat. Am. |
Optimization, forecasting |
The use of AI simplifies the workload and increases the accuracy of the processes. Studies show good results. |
More investigations on social and regulatory issues, and technical development are required. |
This paper is a combination of technical controls as well as how individuals themselves can be involved in it, which forms a connection between efficiency and social approval. |
|
[2] |
Developing Cities |
Demand response, real-time management |
AI is used to assist with demand response and load forecasting; deficit of information and confused policies are the issues. |
AI-driven models look good at making renewable energy systems work better. |
This study takes a head-on collision with the challenges mentioned above by considering the performance of the lightweight artificial intelligence architectures in the situation where there is lack of data, and thus, proves their technological feasibility through empirical studies. |
|
[4] |
Africa, Asia, Lat. Am. |
Decentralized energy management |
Blockchain and AI enable peer-to-peer trading and transparency. |
Blockchain complements AI for transparency and decentralization in energy management. |
The proposed study focuses on the implementation of AI-controlled modular microgrids in the informal settlements, and it thus offers a more local, socially inclusive system of managing energy. |
|
[1] |
Global |
Parameter discovery, analytics |
BERT models are useful in locating improved parameters of energy systems. |
Mentions the necessity of applying tailor-made strategies and engaging parties to adopt them successfully. |
The paper goes beyond the paradigm of algorithmic refinements and integrates the concepts of explainable artificial intelligence (XAI) and participatory co-design approaches, thus guaranteeing not only the quantitative accuracy but also the development of a strong social trust. |
2.1 Theoretical framework
The developed research is situated in a complex of interdependent theoretical frameworks that jointly shape the analysis of both technical and socio-technical aspects of AI-based renewable energy management:
2.1.1 Systems theory
According to the systems theory, the electrical power grid is viewed as a dynamic interdependent system that unites generation, storage, transmission, and consumption modules. The use of artificial intelligence algorithms is aimed at optimizing the performance of the whole system, feedback, and adaptive feedback to the real-time data [5].
2.1.2 Optimization theory
Optimization theory underpins the use of AI for maximizing efficiency, minimizing costs, and ensuring grid reliability. AI models (e.g., reinforcement learning, deep learning) are formulated to solve complex operational problems such as energy dispatch and storage management under multiple constraints [6].
Even though the Random Forest algorithm [7], the Long Short-Term Memory network [8], and the Reinforcement Learning [9] have always been considered as the giants of the machine-learning canon, the current study stands out since it trains the older techniques in new structures specific to the requirements of emerging urban settings. More specifically, we include lightweight, federated learning components [10] and edge-AI designs [11] which, in combination, result in decentralized, privacy-preserving, and resource-efficient solutions, thus responding to the limitations of such environments. Moreover, the project is the first to deploy tiny AI models to edge devices and, as such, challenges the infrastructural and connectivity challenges that are typical of developing cities. In this way, it expands the frontier of the previous research, which has focused mainly on centralized architectures.
2.1.3 Responsible AI and socio-technical frameworks
The modern debate concerning the framework of responsible AI puts a primary focus on the ethical, legal, and social implications that accompany the implementation of artificial intelligence in the energy system. These frameworks endeavor to tailor high-intelligence AI models to the desires of their stakeholders. Such desires are unobtrusive yet must be urgent since most areas will soon be rapidly urbanized. The two things that are important in this case are the society's perception of the work as legitimate and the work being in compliance with the laws. These two things not only render the projects useful but also make the idea of energy workable and acceptable [12, 13].
2.2 Energy transition framework
The energy transition framework provides a complete picture of moving from fossil fuels to renewable energy, and it strongly focuses on the important roles of technology, laws, and social progress (Table 2).
It is thought that artificial intelligence can strongly help break down integration barriers and build sustainable energy systems in cities (Table 3) [14].
Table 2. Demographic distribution of questionnaire respondents by geography, income, and user type
|
Category |
Subcategory |
Percentage (%) |
Number of Respondents |
Notes |
|
Geographic Distribution |
Central Urban Areas |
42 |
147 |
Includes city center areas |
|
Peripheral Urban Areas |
58 |
203 |
Includes suburbs and developing areas |
|
|
Income Distribution |
Low Income |
43 |
150 |
Represents economically limited groups |
|
Middle Income |
41 |
144 |
Middle-class households |
|
|
High Income |
16 |
56 |
Higher income groups |
|
|
User Type |
Residential Consumers |
52 |
182 |
End-users' energy |
|
Utility Staff |
29 |
102 |
Workers in energy utilities |
|
|
Decision Makers/Policy Makers |
19 |
66 |
Regulatory and policy bodies |
Table 3. Theoretical frameworks and their application
|
Framework |
Application in Study |
|
Systems Theory |
This study is about improving smart grids. It implies real time adaptive control measures to make the system far more dependable and productive. |
|
Optimization Theory |
The work relies on AI to plan energy, operate storage, and organise demand response in the way of constructing an electric network that balances and optimises use of resources by them. |
|
Responsible AI |
The success of the solutions that are proposed will be based on the principle of ethical rigor, clear methodology, and a resolute stand on the interests of stakeholders, thus making every innovation consistent with societal values and presenting the broader population with a solution that is fair and accountable. |
|
Energy Transition |
We outline an extensive mapping approach that facilitates the urban shift towards renewable energy, which offers a way forward to cities aspiring to be sustainable in their development endeavors. |
3.1 Questionnaire tool
A good questionnaire was designed in order to gather primary data among the key stakeholders and these were the energy consumers, utility managers as well as the policy makers of the sampled cities.
The tool focused on demographics, perceptions of the users and barriers to adoption, and the validated survey modalities namely Likert scale questions, multiple choices and open-ended questions were used.
The validation of the questionnaire was one of the steps that were undertaken before the questionnaire was rolled out and ensured that the questionnaire was clear and reliable.
3.2 Data collection
Data were collected through questionnaires, monitoring systems, and supporting city reports.
We used stratified random sampling to ensure that the sample represents the main demographic groups, as the number of people who responded to the survey was 350. The population was varied and mixed, with people living in the city center and the outskirts, representing 42 and 58 percent, respectively, geographically. Respondents have been separated into high (16%), middle (41%), and low (43%), and there is a range of economic of the population. The vast majority of the respondents were homeowners (52%), followed by utility employees (29%), as well as policymakers (19%), and this represents numerous different types of people who make energy choices in cities. Such a precise demographic portrait renders the study findings applicable and reliable in the rapidly expanding cities.
3.3 Data analysis methods
The methodology of the analysis is loyal to the high standards in every academic journal, hence, transparency, reproducibility and validity are guaranteed [15].
The qualitative data analysis was derived on 350 samples which were organized into an organized Excel spreadsheet to aid the analysis.
The table has various columns which include text excerpts, coding or classification, researcher notes and the degree of confidence and reliability and the detailed sample data such as age, gender, location, sample type, and date of collection. The construct helps a researcher in order to systematize information, streamline the coding process, and compare samples, as well as to perform an accurate review and analysis of the results.
Remark: The detailed Excel spreadsheet on all the qualitative data analyzed is available to those who could view it and to the reviewers or the researchers who could carry out additional analyses of data.
Design and implementation of smart energy management models
The research introduced smart energy management systems that use AI in planning new cities and focused on adding these technologies to the city’s current power systems (Table 4, Figure 1):
Table 4. Model component, relationships, & functions
|
Component |
Description |
Key Elements & AI Role |
|
Data Acquisition |
IoT sensors, smart meters, weather stations, and grid sensors collect real-time data. |
IoT sensors, smart meters, weather data “ feeds real-time info to AI. |
|
AI Processing |
Data analytics, predictive models, optimization engines, and AI control systems work together to examine data for making predictions, improving processes, and controlling systems automatically. |
Forecasting, optimization, anomaly detection, autonomous control. |
|
Energy Sources |
Integration of solar, wind, traditional grid, and microgrids to diversify and stabilize supply. |
Solar, wind, grid, microgrids “ coordinated by AI for optimal mix. |
|
Distribution Systems |
Smart grid infrastructure, energy storage, distribution networks, and load balancing ensure efficient and reliable delivery. |
Smart grid, storage, load balancing “ AI ensures efficiency and reliability |
|
End Users |
Residential, commercial, industrial, and public service sectors receive and interact with energy services. |
Residential, commercial, industrial, public interact via dynamic pricing, DR, etc. |
|
Challenges |
Legacy infrastructure, financial constraints, and skills gaps can hinder deployment and operation. |
Infrastructure, finance. |
|
Social Factors |
Public awareness, cultural acceptance, and affordability influence adoption and impact. |
skills, awareness, acceptance, affordability. |
|
Benefits |
Improved energy efficiency, cost reduction, and sustainability are the primary outcomes. |
Efficiency, cost savings, sustainability. |
Figure 1. Illustration of a city integrating renewable energy sources, smart grids, and AI-driven management systems
Moreover, a co-design process was established with the community members, the local utility workers, and the policymakers who continue to participate in the workshops that are conducted on a regular basis. This ensured that the AI system was technologically powerful but palatable to the community and also aligned with the local priorities [17].
It is imperative to mention that the actual creativity of this creation is rooted in the combination of traditional algorithmic methods with the participatory co-design approach and blockchain-based transparency and mechanisms. This is a multi-dimensional synthesis that not only improves the aspect of technical performance but also at the same time, enhances the aspects of social acceptance and regulatory compliance- aspects that have never been tackled together in previous studies.
Statistical analysis of the collected data
Table 5. Key performance indicators before and after AI implementation
|
Indicator |
Pre-Implementation |
Post-Implementation |
% Change |
|
Avg. Energy Consumption |
1,200 kWh/month |
1,020 kWh/month |
-15% |
|
Grid Reliability Index |
0.85 |
0.95 |
+12% |
|
Peak Demand Events |
22/month |
18/month |
-18% |
|
User Satisfaction Score |
3.2/5 |
4.1/5 |
+28% |
Interpretation of results
The hypothesis is supported by real data. AI-based smart energy management has improved energy efficiency, kept the grid stable, and increased user satisfaction in growing cities. More effective forecasting and optimization would enable us to utilise renewable resources in a more efficient way and reduce the cost of operation. According to the survey, the users gave the system a warm welcome, particularly when it contained learning and engagement programs. However, the issues related to data quality, technical capacity, and regulatory alignment have also been noted, and the need to introduce context-specific adaptation and capacity amplification has been observed. This was reported by Javed et al. [20] who observed a 15 per cent improvement in grid reliability due to the implementation of AI-based optimisation algorithms in the cities of sub-Saharan cities, which is similar to what we have observed.
The literature on recent publications indicates that there has been an improvement in terms of grid stability and energy efficiency. As a case in point, Adewoyin et al. [21] have defined AI applications to enhance energy access and efficiency in emerging urban settings and explained the contribution of a sound data infrastructure and an effective policy. Second, our findings can be applied to the research by Słyś et al. [22], who showed that under dynamical urban conditions, mixed-optimization approaches could depict better models. These recent findings make our study the validation of not only the benefits related to AI in the renewable-energy management, but also refer to the challenges of developing cities, such as a lack of infrastructure and scarcity of resources.
Moreover, the integration of responsible and explainable artificial intelligence models aligns with contemporary best practices delineated in recent disciplinary review [2]. This fact proves that the suggested models are valid, open, and credible, which is the key property of gaining the trust of stakeholders and ensuring the authorization of the regulatory authorities in the developing countries.
Moreover, the use of the federated and edge learning paradigm in this study enables decentralized learning and local grid-level adaptations to reduce the issue of data privacy and to avoid infrastructural limitations typical of developing urban settings. Moreover, the integration of responsible and explainable artificial intelligence models aligns with contemporary best practices delineated in recent disciplinary reviews, including Henao et al. [2].
This fact proves that the suggested models are valid, open, and credible, which is the key property of gaining the trust of stakeholders and ensuring the authorization of the regulatory authorities in the developing countries.
Moreover, the use of the federated and edge learning paradigm in this study enables decentralized learning and local grid-level adaptations to reduce the issue of data privacy and to avoid infrastructural limitations typical of developing urban settings.
The explainable AI (XAI) models that we used to analyze the outputs of our model included SHAP and LIME, which helped to understand the model in line with the principles of Responsible AI. Figure 2 shows a SHAP summary plot that shows the relative contribution of divergent features to the prediction of energy-management outcomes. The presentation clearly shows how the model questions and evaluates each of the input variables.
Figure 2. Feature importance analysis using SHAP values
The figure shows the contribution of every feature to the end predictions of the model applied by the SHAP (Shapley Additive Explanations) approach. Where each point on the plot is an instance of data and:
Interpretation: To illustrate, when the feature Solar Generation has red dots, which correspond to high values, on the right side of the plot, the high anticipated values are associated with increasing the amount of solar power output. On the other hand, the blue dots, which depict low values, on the left side suggest that the less solar generation is generated, the fewer predictions are. This visualization allows understanding which characteristics are the most influential, and at the same time, it shows how different degrees of each characteristic, in turn, drive the model in opposite directions with its predictions.
A strict bias audit was conducted to identify equity among relevant demographic groups represented in the data. When evaluating this, we considered major measures such as demographic parity and equalized odds to certify that the performance of predictors was equitable across different income levels and various areas of geographic coverage. The empirical evidence indicated that there is a slight difference only, hence, justifying the fact that the decision making of the model is fair among the sampled population.
This study has shown that AI-based smart energy management models are great solutions to developing cities that cannot avoid implementing sustainable, resilient, and efficient energy systems. Key findings include:
Despite the progress that has been made in this research, several limitations are evident, the first of which is the small sample size and the diversity of the city environments studied in this study, which limits the generalizability of our results to municipalities outside the scope of this study.
Besides, the variability was introduced into some parts of the data even by the technical complications and regulatory contingency that the research team could not control. In this regard, we recommend future studies that utilize larger samples and apply more refined and technologically advanced study designs in order to overcome these barriers.
7.1 Enhancing originality in proposed models
This study recommends the following advanced strategies to further strengthen the originality and contextual suitability of AI-driven energy management in developing cities:
7.2 Future research directions
Despite improvements made, there are gaps in research: Longitudinal Impact Studies: Carry out thorough assessments of the sustainability, resilience, and equity of the AI-managed energy management over several years.
The articles are based on methodological and reporting standards of the most popular academic journals in the discipline and any claims are supported by recent and high-quality sources recorded in the Scopus database (2022-2025) found in the literature review.
All primary data and statistical models used in this study are available to interested researchers upon request, in accordance with the privacy and data protection policies of the participants. The person interested can be contacted via the email listed at the end of the research to obtain the data.
The authors would like to thank Al-Mustansiriyah University (www.uomustansiriyah.edu.iq), Baghdad, Iraq, for its support in the current work. And also, thanks to the University of Technology (www.uotechnology.edu.iq), Baghdad, Iraq.
[1] Alsaigh, R., Mehmood, R., Katib, I. (2023). AI explainability and governance in smart energy systems: A review. Frontiers in Energy Research, 11: 1071291. https://doi.org/10.3389/fenrg.2023.1071291
[2] Henao, F., Edgell, R., Sharma, A., Olney, J. (2025). AI in power systems: A systematic review of key matters of concern. Energy Informatics, 8(1). https://doi.org/10.1186/s42162-025-00529-1
[3] Wang, Q., Li, Y., Li, R. (2025). Integrating artificial intelligence in energy transition: A comprehensive review. Energy Strategy Reviews, 57: 101600. https://doi.org/10.1016/j.esr.2024.101600
[4] Kuzior, A., Sira, M., Brozek, P. (2022). Using blockchain and artificial intelligence in energy management as a tool to achieve energy efficiency. Virtual Economics, 5(3): 69-90. https://doi.org/10.34021/ve.2022.05.03(4)
[5] Gbadamosi, S.P., Edward Ereh, R., Babatunde Ibrahim, A., Ashraf Tunde, U. (2025). The role of artificial intelligence in optimizing renewable energy system. IJEMT International Journal of Engineering and Modern Technology, 11(3): 209-229. https://doi.org/10.56201/ijemt.vol.11.no3.2025.pg209.229
[6] Hanif, M.F., Ahmad, R., Farooq, A.B., Liu, X., Ahmad, A., Iqbal, S., Mi, J. (2025). The solar AI nexus: Reinforcement learning shaping the future of energy management. WIREs Energy and Environment, 14(3). https://doi.org/10.1002/wene.70012
[7] Breiman, L. (2001). Random forests. Machine Learning, 45(1): 5-32. https://doi.org/10.1023/a: 1010933404324
[8] Hochreiter, S., Schmidhuber, J. (1997). Long short-term memory. Neural Computation, 9(8): 1735-1780. https://doi.org/10.1162/neco.1997.9.8.1735
[9] Sutton, R.S., Barto, A.G. (1998). Reinforcement learning: An introduction. IEEE Transactions on Neural Networks, 9(5): 1054-1054. https://doi.org/10.1109/tnn.1998.712192
[10] McMahan, H.B., Moore, E., Ramage, D., Hampson, S., Arcas, B.A.Y. (2016). Communication-efficient learning of deep networks from decentralized data. arXiv preprint arXiv:1602.05629. https://doi.org/10.48550/ARXIV.1602.05629
[11] Deng, S., Zhao, H., Fang, W., Yin, J., Dustdar, S., Zomaya, A.Y. (2020). Edge intelligence: The confluence of edge computing and artificial intelligence. IEEE Internet of Things Journal, 7(8): 7457-7469. https://doi.org/10.1109/jiot.2020.2984887
[12] Haidar, A. (2023). An integrative theoretical framework for responsible artificial intelligence. International Journal of Digital Strategy, Governance, and Business Transformation, 13(1): 1-23. https://doi.org/10.4018/ijdsgbt.334844
[13] Mustafa, M.H., Al-Saatee, H.H., Al-Bakry, H.M.J., Sabu, A.I., Al-Helli, M., Al-Graiti, A.A.A., Mohammed, A.A., Ismael, Z.A., Mohson, Z.H., Jassim, A.K. (2025). The importance of empowering the smart city in Iraq: A case study of Baghdad municipalities. International Journal of Sustainable Development and Planning, 20(3): 1133-1141. https://doi.org/10.18280/ijsdp.200319
[14] Magdalena, R., Sharif Karimi, M., Si Mohammed, K., Dumitru, M., Hagiu, A. (2025). Pioneering sustainable energy: A dynamic analysis of AI and green patents in renewable and nuclear power. Journal of Applied Economics, 28(1). https://doi.org/10.1080/15140326.2025.2519068
[15] Ali Khan, J., Raman, A.M., Sambamoorthy, N., Prashanth, K. (2023). Research Methodology (Methods, Approaches and Techniques). SAN International Scientific Publications.
[16] Alam, M.M., Hossain, M.J., Habib, M.A., Arafat, M.Y., Hannan, M.A. (2025). Artificial intelligence integrated grid systems: Technologies, potential frameworks, challenges, and research directions. Renewable and Sustainable Energy Reviews, 211: 115251. https://doi.org/10.1016/j.rser.2024.115251
[17] Zytko, D., J. Wisniewski, P., Guha, S., P. S. Baumer, E., Lee, M.K. (2022). Participatory design of AI systems: Opportunities and challenges across diverse users, relationships, and application domains. In CHI Conference on Human Factors in Computing Systems Extended Abstracts, pp. 1-4. https://doi.org/10.1145/3491101.3516506
[18] Lee, S. (2025). 7 surprising stats: ANOVA’s role in energy & utilities. https://www.numberanalytics.com/blog/anova-energy-statistics-utilities.
[19] AlGhanim, A. (2003). A statistical approach linking energy management to maintenance and production factors. Journal of Quality in Maintenance Engineering, 9(1): 25-37. https://doi.org/10.1108/13552510310466828
[20] Javed, H., Eid, F., El-Sappagh, S., Abuhmed, T. (2025). Sustainable energy management in the AI era: A comprehensive analysis of ML and DL approaches. Computing, 107(6). https://doi.org/10.1007/s00607-025-01485-0
[21] Adewoyin, M.A., Adediwin, O., Audu, A.J. (2025). Artificial intelligence and sustainable energy development: A review of applications, challenges, and future directions. International Journal of Multidisciplinary Research and Growth Evaluation, 6(2): 196-203. https://doi.org/10.54660/.ijmrge.2025.6.2.196-203
[22] Słyś, D., Stec, A., Bednarz, K., Ogarek, P., Zelenáková, M. (2025). Managing and optimizing hybrid distributed energy systems: A bibliometric mapping of current knowledge and strategies. Energies, 18(10): 2497. https://doi.org/10.3390/en18102497