Artificial intelligence is changing the field of hydroelectric power by improving forecasting, maintenance, and operational decision-making. This review collects research published between 2022 and 2025 on the use of data-driven methods in hydropower plants.
Effect of AI on Hydroelectricity Over the Past Three Years
Raj Shah, Justin Zheng, Mathew Stephen Roshan, Beau Eng, William Chen | Koehler Instrument Company
Abstract
Artificial intelligence is changing the field of hydroelectric power by improving forecasting, maintenance, and operational decision-making. This review collects research published between 2022 and 2025 on the use of data-driven methods in hydropower plants. Modern forecasting models use machine learning to capture nonlinear relations between weather, inflows, and production, which yields more accurate generation and water supply predictions than conventional statistical methods and supports improved reservoir management and energy market participation. Predictive maintenance strategies combine continuous sensor monitoring with learned patterns of normal and faulty behavior, allowing earlier and more precise intervention, less unplanned downtime, and reduced maintenance cost. Digital twins of turbines, sensors, and entire plants provide virtual replicas that support anomaly detection, operator training, and scenario analysis, and their coupling with deep learning and fuzzy logic controllers has been shown to increase efficiency, reliability, and fault detection performance. Operational optimization methods use reinforcement learning and other search techniques to coordinate water dispatch, unit commitment, and multi-reservoir management while respecting environmental constraints and reducing mechanical stress during startup and ramping. The review also identifies key challenges related to data quality and availability, cybersecurity risk, integration with legacy infrastructure, and the need for new skills among plant staff. Overall, the literature indicates that artificial intelligence and digital twins can enhance the efficiency, flexibility, and resilience of hydroelectric power systems.
Introduction
In recent years, AI, artificial intelligence, has significantly expanded and integrated into a wide variety of sectors. One such field is hydroelectric power, where AI models are used to enhance forecasting, maintenance, and operations. Hydroelectric power is one of the main sources used in global renewable electricity generation, generating approximately 60% of all renewable electricity and 14% of overall electricity generation [1]. Traditional hydroelectric systems are rule-based or calendar-driven, having operators manually monitoring equipment performance and scheduling interventions [1]. There is a modern-day push for integrating digitalization into hydropower systems to keep up with increasing demands [2]. It exchanges the adoption of a more complex system for significant contributions to the stability and flexibility of future energy supplies [2]. As global requirements for efficient and clean energy systems grow, traditional hydroelectric systems prove increasingly insufficient [2]. These models are unable to track issues, resulting in untimely interventions that waste time and money [3]. Influx of fragmented data, where essential information is scattered across disconnected systems, also complicates the learning curve, requiring long-term employees to supervise for the sake of familiarity [3]. AI models can resolve such issues by constantly analyzing large volumes of this data to identify patterns, detect anomalies, and predict failures in real time, delivering precise insights into system performance and potential faults [4]. Notably, recent AI models include the integration of digital twins that create real-time virtual replicas of hydropower systems and deep learning algorithms, which is what allows AI to process large amounts of data [4]. Digital replicas called digital twins can be fashioned to simulate systematic failures and optimize resource management [3]. This paper explores the integration and benefits of AI applications in hydroelectric power systems, focusing on research published in the past three years. It primarily addresses forecasting, predictive maintenance, digital twin implementation, and optimization.
Forecasting and Inflow Prediction
Forecasting in hydroelectric systems refers to the prediction of future water inflows, which directly influences reservoir levels and electricity generation. Having accurate forecasting is essential, as it allows for more efficient management of limited reservoir storage and better optimization of decisions in response to changing hydrological and meteorological conditions. Machine learning algorithms assist in this process by facilitating precise forecasts in resource allocation and grid management [5]. In a 2025 study performed by Di Grande et al., several machine learning models were developed and tested for the prediction of hydropower generation in water distribution systems [5]. One of the approaches tested was the Random Forest (RF) model, a learning method that accumulates predictions from multiple decision trees, which was particularly effective at modeling long-term patterns due to its ability to capture nonlinear relationships [5]. In contrast, the Temporal Convolutional Network (TCN), designed for sequential data, was better suited for short-term forecasting because it can learn sudden changes in data more effectively [5]. The primary metric they used to determine this result was by finding the Symmetric Mean Absolute Percentage Error (SMAPE), Mean Absolute Percentage Error (MAPE), Mean Absolute Scaled Error (MASE), and the Root Mean Squared Scaled Error (RMSSE) produced by each model [5]. These metrics allow for the examination of the difference between predicted and observed values, with lower values representing smaller prediction errors, in turn meaning better model accuracy [5]. MAPE and SMAPE measure relative percentage errors, which are useful for comparing forecasting performance across different scales [5]. MASE and RMSSE measure error magnitude relative to a control forecasting method [5]. These values are demonstrated in Table 1 below, which depicts the lowest values obtained from their model experiments.
Table 1. Performance of three forecasting models in a water distribution system. Reproduced from Di Grande et al. (2025).
|
Model
|
Aggregation |
Forecasting |
SMAPE |
MAPE |
MASE |
RMSSE |
|
RF
|
Monthly |
One step |
8.076 |
7.718 |
0.768 |
1.073 |
|
RF
|
Two weeks |
Two steps |
6.757 |
6.501 |
0.659 |
2.015 |
|
TCN
|
Two weeks |
One step |
4.939 |
4.881 |
0.886 |
1.365 |
The low values across these models demonstrate the improved accuracy that AI models can reach in forecasting, relative to traditional methods that struggle with the unpredictability of weather conditions and water inflow. Such methods include persistence and simple statistical models, producing MAPE and SMAPE values exceeding 15%, in comparison to the AI-based models which reduced these errors to below 8% [5]. Accurate data recording produces prominent benefits: allowing for better energy sale planning, information gathering for maintenance scheduling and water management, and efficient balancing of energy supply and demand [5]. Operators will be able to optimize reservoir releases and unit commitment, which involves scheduling individual turbines and generators to operate at the lowest electricity costs [5]. By aligning generation with up-to-date market conditions, hydroelectric systems can minimize costs while maximizing generation efficiency [5].
This sort of technology is already bearing fruit with Hydrogrid, a Vienna-based start-up founded by Janice Goodenough. They specialize in the integration of AI into hydroelectric systems, reporting that machine learning promotes proactive designs over traditionally reactive plans, such as predicting seasonal trends and water conditions [6]. Contemporary incorporation includes specialized models, like the Hydrogrid Insight Platform, a multi-model methodology that enables short-term forecasts and longer-term projections for seasonal reservoir strategies [6]. Building on the importance of AI in practical designs, in a 2024 study done by Fleming et al., machine learning models were trained using the MODSCAG fractional snow-covered area (fSCA) as an input variable for water supply forecasting [7]. Snow cover is a critical indicator of future runoff, and the integrating fSCA data allows the model to better capture seasonal melt dynamics that traditional forecasting methods often overlook [7]. As a result, late-season water supply forecast accuracy improved by 10–25% compared to conventional approaches [7]. Shown in Figure 1 are the locations of their test models. Ultimately, by employing AI in forecasting models, hydropower operators can optimize generation scheduling and reservoir management by providing more reliable insight into future water availability and energy production. These AI-driven forecasting systems have the potential to dramatically improve the efficiency of hydroelectric operations by enabling a more proactive decision-making network and reducing uncertainty in resource planning. Over time, such advancements could lead to a hydropower sector that is autonomously capable of adapting to changing climate and market conditions.
Figure 1. Location map showing four test basins across the western US. Reproduced from Fleming et al. (2024)
Predictive Maintenance and Asset Monitoring
Traditional hydropower maintenance schemes are typically calendar-based and reactive. This leads to interventions that occur prematurely or after failures have already developed. A 2025 blog from Ngue, CEO of Quebec-based start-up EN Solutions Hydro that creates specialized software solutions for hydroelectric power plants, suggests that up to 40% of scheduled maintenance actions are unnecessary, with reactive maintenance accounting for a substantial share of unplanned outages and cost overruns [3]. To improve efficiency, new AI applications aim to transition into a more proactive strategy. One such way is the continuous analysis of sensor data to detect discrepancies and optimize maintenance timing. Ngue describes the “Virtual Maintenance Expert" platform, which makes AI assistance readily available to operators [3]. This integration method employs intelligent workflows with AI algorithms constantly generating work order requests, creating a knowledge assistant called Retrieval-Augmented Generation (RAG) [3]. This system combines relevant historical maintenance records, technical documentation, and real-time data to support decision-making along with proactive anomaly detection systems [3].
In combination with digital twin models, virtual replicas of hydropower systems, predictive maintenance strategies can be further expanded. For example, comparison between live sensor data and a digital twin’s expected behavior allows for detection when the real equipment’s behavior differs from the model's predictions, forecasting failures before they occur [3]. In a 2024 study conducted by Ersan and Irmak, a digital twin was developed for a real hydroelectric power plant in Turkey to replicate the expected behavior of a pressure sensor under normal operating conditions [8]. In their study, Ersan and Irmak compared real-time sensor measurements with the digital twin’s simulation outputs and reported that the model was able to correctly identify sensor-related discrepancies with an accuracy of more than 90%, tremendously reducing false fault occurrences [8]. This allowed for the prevention of unnecessary shutdowns, avoiding production loss [8]. Shown in Figure 2 is the structure of their digital twin. Considering everything, studies trend towards a significant reduction of unplanned downtime and maintenance costs when AI-driven predictive maintenance is utilized.
Figure 2. Basic structure of the digital twin model used. Ersan and Irmak (2024)
A 2025 study conducted by Tan et al. investigated the integration of deep learning algorithms with a digital twin framework to improve fault detection in hydropower systems [4]. They reported that integration of deep learning algorithms with a hydropower digital twin reduced fault detection time by 12.14%, increased overall system efficiency by 8.97%, and cut maintenance costs by 5.49% compared to traditional methods [4]. Similarly, research from a 2025 study performed by Zeng et al. examined a hybrid control framework that combines Type-2 Fuzzy Logic Controllers, neural networks, and a digital twin to reduce uncertainty in hydropower operations [9]. While the digital twin simulated system behavior, the Type-2 Fuzzy Logic Controllers’ purpose is to gain the ability to administer adaptive control under changing operating conditions. Compared to traditional strategies, this approach improved load management efficiency by 11.48%, reduced fault detection time by 12.64%, and lowered maintenance costs by 13.04% [9]. With all these theoretical studies on AI integration, practical developments have also occurred. For instance, VERBUND, one of Europe’s leading hydropower operators, uses AI in a real-time anomaly detection system to predict equipment failures before they occur [10]. This system conducts continuous comparison of sensor readings to predicted normal ranges and raises an anomaly score if deviation occurs [10]. Higher scores, 2-3 standard deviations from normal operating behavior, alert operators for further inspection, while lower scores are constantly monitored to avoid unnecessary false alarms [10]. AI is also capable of incorporating economic factors into maintenance decisions. AI systems can coordinate maintenance to match periods of low electricity prices or reservoir inflows that are too high to minimize revenue impact [11]. As these systems become more widely adopted, achievement of predictive maintenance strategies that enhance both the operational efficiency and long-term performance of hydroelectric power plants will be commonplace.
Digital Twins
Digital twins are essential for many of the AI applications discussed in this review. In the hydroelectric sector, it functions as a simulation of power plants or separate components informed through real-time data [3]. Traditionally, hydropower monitoring and control have relied on physics-based models to assess system performance and detect faults. While this approach is effective under more stable operating conditions, it sometimes leads to reduced diagnostic accuracy and delayed fault detection due to struggles with more complex interactions within the system. Many sources point to an amplification of effectiveness when a digital twin is paired with AI, as it enables the ability for AI to continuously learn from both historical and real-time data rather than only the traditional physics-based models. To investigate this claim, Tan et al. integrated a deep learning model with a hydropower digital twin [4]. The study revealed that the AI-enhanced digital twin achieved a 72% accuracy rate, outperforming the 65% accuracy observed with conventional techniques [4]. Accuracy rate in this study refers to the system’s ability to identify operating conditions and faults when compared against the observed plant data. Digital twins have been implemented for tasks such as sensor validation, virtual operation change testing, and providing operators with a virtual model to learn from, allowing less advanced engineers to gain the necessary experience needed with lower-stakes simulations.
Figure 3. Simulated view of the entire hydropower plant using a digital twin. Reproduced from Cao T. et al. (2025)
The integration of computational fluid dynamics (CFD) and AI-based modeling plays a critical role in improving the efficiency of digital twins in hydropower applications. CFD simulations provide physics-based representations of fluid flow within systems, leading to detailed analysis of pressure distribution, turbulence, and flow-induced stresses [12]. However, they can be impractical for real-time operation on their own, as they require hefty computational power. With AI, this limitation is nullified as AI can learn from historical simulation and sensor data, allowing the digital twin to estimate complex fluid dynamics smoothly during live operation. A 2025 study conducted by Machalski et al. describes a practical application of a digital twin for a hydro plant in Poland and examines the development of a digital twin for an operating hydroelectric power plant in Poland with the goal of improving performance assessment and operational decision support [12]. They trained AI-based models on CFD outputs. As a result, the AI-enhanced digital twin demonstrated prediction errors reduced by 15–20% compared to traditional hydraulic plant models, while achieving simulation times several orders faster than base CFD simulations [12]. Shown in Figure 4 is the process of how the authors developed their digital twins.
Figure 4. Digital twin development process. Reproduced from Machalski et al. (2025)
Digital twin initiatives in hydropower have already demonstrated concrete benefits. VERBUND’s “Hydropower 4.0” at the Rabenstein power plant in Austria, for instance, has improved anomaly detection, successfully deployed autonomous drones for underwater inspections, and enhanced maintenance planning efficiency [13]. A primary use for digital twins is predictive simulations, allowing operators to be alert of issues and proactively reinforce equipment before it can fail. Sustainability Directory, an industry-focused publication that reports on sustainability technologies, describes how a digital twin, acting as a “digital nervous system,” can anticipate distress in dam infrastructure well before failure by analyzing degradation in the simulation [14]. A twin can pinpoint, with high accuracy, when a specific turbine blade, generator, or spillway gate will require intervention [14]. Maintenance can become a precise and scheduled activity that maximizes lifespan and energy output. As previously mentioned, digital twins aid in operational optimization. With the simulated scenarios, AI can uncover optimal strategies that would be risky or impossible to discover manually. Many researchers and practitioners consider the development of next-generation predictive digital twins a key future direction for hydropower [2]. These developments include physics-based models, big data, and AI-driven analytics that would drive hydropower towards autonomy and resilience.
While digital twins offer significant advantages for monitoring and optimization, their implementation in hydropower systems also presents notable challenges. Machalski et al. demonstrates that the act of creating a useful digital twin requires working with its own unique set of issues [12]. One of the primary challenges is incomplete legacy data, which led to difficulty in model calibration and reduced initial accuracy [12]. Moreover, the authors also explained the need for modular system architecture that can collect real-time data. This is much needed flexibility, as digital twin models are at risk of becoming too rigid to maintain as new operational requirements add up. More broadly, research on digital twin implementation reveals data heterogeneity and integration challenges, as sensor data is collected from varying formats and sources [15]. This is a significant technical barrier to building real-time digital twin models for hydropower [15]. Figure 5 shows the integration of digital twins with other components, and Figure 6 shows the severity of risk digital twins face from each challenge.
Figure 5. Physical and digital architecture of a digital Twin system for renewable energy. Mbasso et al. (2025)
Figure 6. Distribution of digital twin implementations across renewable energy sectors. Mbasso et al. (2025)
Optimization and Smart Control
AI techniques are also being applied to optimize the long-term operation of hydroelectric plants, working with digital twins as well as forecasting to boost decision-making. In the future, AI will also be able to recommend the best course of action to maximize efficiency and revenue under strict constraints. A 2025 review by Mbasso et al. talks about industry-level implementations of digital twin technology combined with AI-driven optimization in hydroelectricity [15]. Their review assessed changes in energy yield and downtime by comparing digitalized operations with more traditional practices [15]. Mbasso et al. reported that industry implementations achieved 10–20% improvements in energy yield and up to a 25% reduction in downtime, demonstrating the effectiveness of AI-enhanced digital twins [15]. In a more practical sense, ANDRITZ, a global engineering company specializing in hydropower, has developed the Metris DiOMera digital ecosystem aiming to modernize hydropower plant operation [16]. The platform integrates digital twin models, AI-driven monitoring, and predictive maintenance algorithms to create a continuously adapting representation of hydroelectric plants [16]. This allowed digital twins to simulate the physical behavior of systems, while AI algorithms adjust model predictions as operating conditions change, allowing the plant to transform into a seemingly self-optimizing operation [16]. This ecosystem is designed to support the increasing flexibility demands by enabling faster responses [16]. AI also can optimize multi-reservoir operations, with research prototypes employing deep reinforcement learning agents to manage complex reservoir systems [2]. This involves scheduling water releases across interconnected reservoirs for flexibility on objectives such as flood control and water supply [2]. Traditional optimizations rely on deterministic optimization, which struggles with system complexity [2]. As such, contemporary prototypes have started to utilize deep reinforcement learning (DRL) agents, which learn optimal reservoir release policies through repeated interaction using digital twins rather than relying on predetermined procedures [2]. DRL-based controllers have been shown to outperform traditional optimization methods, especially when managing across multiple reservoirs under uncertain inflows [2]. Despite its early stage, promising results include learning optimal release policies that maximize long-term energy generation and being able to handle the objective nature of reservoir optimization while respecting environmental and safety constraints [2]. For example, a 2025 study conducted by Wu et al., where they applied a transformer-based DRL model to multi-reservoir operation strategies in the Colorado River Basin, increased electricity generation by 10.11%, reduced annual flow deviation by 39.69%, and improved water supply revenue by 4.10% compared to conventional methods, which is indicative of a more efficient multi-objective control when utilizing optimal release policies [17].
Optimization also includes automatic control tuning, as demonstrated by Zeng et al.’s hybrid fuzzy logic and neural network controller, which reduced operational deviations by 8.05% [9]. This behavior improves efficiency and reduces mechanical stress. Modern hydropower optimization models incorporate ecological flow requirements as well to ensure that water releases comply with environmental regulations while meeting energy production objectives. To adapt to these requirements, AI frameworks can adjust releases to meet downstream habitat needs at minimal cost to generation [2]. For instance, an AI scheduling tool can identify an optimal pattern of flows to protect aquatic life by analyzing thousands of predictive release trajectories and pinpointing the best solution to balance power and environmental performance. Despite the early adoption phases, a significant quantity of data demonstrates that AI-optimized hydropower plants can operate more flexibly and reliably than before. A 2025 study conducted by Muser et al. researched the use of a neural network to optimize turbine startup sequences in hydropower plants where traditional procedures usually induce significant mechanical stress [18]. The study found that the classic start-up trajectory totaled 3.213 × 10−22 points of damage, while the optimized trajectory incurred only 1.014 × 10−24 points of damage [18]. This means that there is an achievement of up to a 99% reduction in fatigue stress compared to the conventional startup procedure. Overall, optimization utilizing AI allows hydropower plants to operate more flexibly and efficiently. By developing optimization, hydropower plants will be able to respond adaptively to changing conditions, further solidifying hydropower’s role in energy production.
Challenges
AI and digital twin technologies have already produced practical improvements in hydropower forecasting, maintenance, and operations, but there are also significant challenges to consider as AI integrates itself further into hydroelectricity. One major issue is data quality and availability. To create effective machine learning models, large volumes of high-quality data are required; it is common to use years of operational data to train a predictive maintenance algorithm [19]. Shown in Figure 7 is the AI modelling process. Many hydro plants are relatively new, and failure data can be uncommon, meaning that there are few examples to learn from. There are ways to bolster data volume, including techniques like synthetic data generation through digital twins [2]. However, the effectiveness of these approaches still depends on the currently available high-quality real-world data [2]. Sensor infrastructure is essential for capturing accurate, high-resolution measurements of practical plant data, which is the foundation for training AI models [2]. Equally important is effective data management, as this data must be securely stored for long-term usage [2]. There needs to be continuous investment in sensor infrastructure and data management so AI models can be trained with confidence. Another major challenge is integrating these advanced systems into traditional hydropower operations. As hydroelectric plants are critical infrastructures, operators are more cautious about risk-inducing changes. AI recommendations and automated controls require trust. In such high-risk applications, human oversight remains important, and AI systems should be designed to act as decision support rather than fully autonomous operators.
Figure 7. Model development process. Bechara H. et al. (2024)
Cybersecurity also remains a prominent issue, as increasingly digitalizing structures introduce vulnerabilities unique to modern systems [11]. There are many recent examples of cyber incidents. One is the 2023 cyberattack on Hydro-Québec, where internal computer networks were affected, forcing the plant to temporarily isolate parts of its systems to contain the damage [2]. It showed how digital systems for communication can become an entry point for attackers. Another is the 2022 attempted attack on the Grand Ethiopian Renaissance Dam, where attackers tried to gain access to industrial control and monitoring interfaces [2]. Such attacks call for the development of security frameworks, like the IEC 62443 standard for industrial control systems, which provide guidance on securing digitally connected infrastructure by managing access control, as the integration of AI and digital twins becomes more prevalent [2]. Fortunately, monitoring technology in hydropower plants can also function in security devices. Digital twins can be used to simulate and test cyber defenses, and AI algorithms can detect unusual patterns that might point to cyber intrusions. Introducing new skills to personnel is also necessary, as training staff to work with adopting AI tools could help address issues with AI implementation. While there may be resistance to AI integration in this field, the benefits AI provides are too great to give up. Predictions point to AI and digital twins becoming standard tools in hydroelectricity [2]. The continued adoption of AI allows for the transition of hydropower from a reactive operation towards a more autonomous and sustainable system in the future [2].
Future Work
AI applications connect strongly with broader goals, including smart grid integration and renewable energy optimization. Hydropower has a definite role as a flexible power source, making forecasting and real-time control with AI extremely valuable to operators. Contemporarily, there is a market push for efficiency and availability that prioritizes predictive maintenance to avoid outages and optimize output. Future applications could include scaling up digital twin frameworks to entire fleets of plants, integrating hydropower AI systems with energy market platforms and other generation assets to coordinate operations in a more holistic fashion, and using AI to enhance environmental performance. There is also current research to combine physics-based models with AI, merging both the trustworthiness of engineering models with the pattern-recognition power of machine learning [20]. Physics-based models offer consistency by enforcing known laws, but they often struggle with more complex, nonlinear behavior [20]. This fusion represents a key direction for AI development in hydropower, producing more trustworthy decision support as hydropower systems grow in complexity [20].
Conclusion
Over the past three years, the adoption of AI has been dramatically reshaping the future of hydroelectric power. The applications of AI in forecasting, predictive maintenance, digital twinning, and operation optimization for hydropower systems are all effective strategies for further integration and development within the field. According to the surveyed literature, which include studies by Tan et al., Zeng et al., and Mbasso et al., as well as implementations reported by ANDRITZ and VERBUND, it is revealed that AI techniques can noticeably improve the efficiency and reliability of hydropower operations. For instance, forecasting produces more accurate inflow and generation predictions, leading to better resource planning. Predictive maintenance systems utilize AI to anticipate equipment failures and plan interventions, which reduces unnecessary downtime and maintenance costs. Pairing AI and digital twins allows hydropower plants to run more optimally, which provides a more flexible and rapid response to grid demands. Real-world applications have already shown quantifiable benefits. The integration of AI will reinforce hydropower’s role as a reliable renewable energy source. However, introducing new technologies also brings challenges, including data quality and availability, cybersecurity risks, as well as the need for workforce training and organizational adaptation. Ultimately, the adoption of AI in hydroelectric power is essential for the future as we approach new horizons in technology.
Biographies
Dr. Raj Shah, is a Director at Koehler Instrument Company in New York, where he has worked for the last 25 plus years. He is an elected Fellow by his peers at ASTM, IChemE, ASTM, AOCS, CMI, STLE, AIC, NLGI, INSTMC, Institute of Physics, The Energy Institute and The Royal Society of Chemistry. An ASTM Eagle award recipient, Dr. Shah recently coedited the bestseller, “Fuels and Lubricants handbook”, details of which are available at ASTM’s Long-awaited Fuels and Lubricants Handbook https://bit.ly/3u2e6GY. He earned his doctorate in Chemical Engineering from The Pennsylvania State University and is a Fellow from The Chartered Management Institute, London. Dr. Shah is also a Chartered Scientist with the Science Council, a Chartered Petroleum Engineer with the Energy Institute and a Chartered Engineer with the Engineering council, UK. Dr. Shah was recently granted the honorific of “Eminent engineer” with Tau beta Pi, the largest engineering society in the USA. He is on the Advisory board of directors at Farmingdale university (Mechanical Technology), Auburn Univ(Tribology), SUNY, Farmingdale, (Engineering Management) and State university of NY, Stony Brook (Chemical engineering/ Material Science and engineering). An Adjunct Professor at the State University of New York, Stony Brook, in the Department of Material Science and Chemical Engineering, Raj also has over 725 publications and has been active in the energy industry for over 3 decades. More information on Raj can be found at https://shorturl.at/JDPZN.
Mr. Mathew Roshan and Justin Zheng are Chemical and Molecular Engineering Undergraduate Students at Stony Brook University and interns at Koehler Instrument Company, Holtsville, NY.
Mr. William Chen and Beau Eng have earned their graduate in Chemical Engineering from the State University of New York, Stony Brook, NY and is a member of the senior internship program at Koehler Instrument company in Holtsville.
References
[1] IEA. (2023, July 11). Hydropower. IEA. https://www.iea.org/energy-system/renewables/hydroelectricity
[2] Cao, T., Gao, S., Shi, X., Mo, L., Zhang, B., & Zhu, J. (2025). Core technologies for hydropower digitalization within the Energy Internet framework: a mini-review. Frontiers in Water, 7. https://doi.org/10.3389/frwa.2025.1681345
[3] From Reactive to Intelligent: How AI and Digital Twins are Reshaping Hydropower –Waterpower Canada. (2025, September 18). Waterpower Canada. https://waterpowercanada.ca/learn/blog/all/from-reactive-to-intelligent-how-ai-and-digital-twins-are-reshaping-hydropower/
[4] Tan, J., Raoof Mohammed Radhi, Kimia Shirini, Gharehveran, S. S., Zamen Parisooz, Khosravi, M., & Hossein Azarinfar. (2025). Innovative framework for fault detection and system resilience in hydropower operations using digital twins and deep learning. Scientific Reports, 15(1). https://doi.org/10.1038/s41598-025-98235-1
[5] Grande, S. D., Mariaelena Berlotti, Cavalieri, S., & Gueli, R. (2024). A Machine Learning Approach to Forecasting Hydropower Generation. Energies, 17(20), 5163–5163. https://doi.org/10.3390/en17205163
[6] HYDROGRID Blog: Real-Time Inflow Predictions in Hydropower. (2024). Hydrogrid.ai. https://www.hydrogrid.ai/blog/real-time-inflow-predictions-hydropower
[7] Fleming, S. W., Rittger, K., Oaida, C. M., & Graczyk, I. (2024). Leveraging Next‐Generation Satellite Remote Sensing‐Based Snow Data to Improve Seasonal Water Supply Predictions in a Practical Machine Learning‐Driven River Forecast System. Water Resources Research, 60(4). https://doi.org/10.1029/2023wr035785
[8] Mustafa Ersan, & Erdal Irmak. (2024). Development and Integration of a Digital Twin Model for a Real Hydroelectric Power Plant. Sensors, 24(13), 4174–4174. https://doi.org/10.3390/s24134174
[9] Zeng, Y., Hussein, Z. A., Chyad, M. H., farhadi, A., Yu, J., & Rahbarimagham, H. (2025). Integrating type-2 fuzzy logic controllers with digital twin and neural networks for advanced hydropower system management. Scientific Reports, 15(1). https://doi.org/10.1038/s41598-025-89866-5
[10] Eurelectric - The link between AI & clean energy transition. (2023). Eurelectric beyond Digital Website. https://digitopia.eurelectric.org/case-studies/verbund-hydro-power/
[11] The use of AI in hydropower plants is creating new challenges. (2024, December 3). Pv Magazine International. https://www.pv-magazine.com/2024/12/03/the-use-of-ai-in-hydropower-plants-is-creating-new-challenges/
[12] Artur Machalski, Szulc, P., Dominik Błoński, Nycz, A., Nemś, M., Janusz Skrzypacz, Janik, P., & Zuzanna Satława. (2025). The Concept of a Digital Twin for the Wały Śląskie Hydroelectric Power Plant: A Case Study in Poland. Energies, 18(8), 2021–2021. https://doi.org/10.3390/en18082021
[13] Gallo, A. T. (2024, September 13). Embracing Digital Transformation in Hydropower with Di-Hydro - Di-Hydro - Digital Hydropower. Di-Hydro - Digital Hydropower. https://dihydro-project.eu/2024/09/13/embracing-digital-transformation-in-hydropower-with-di-hydro/
[14] Sustainability Directory. (2025, August 18). Predictive AI for Sustainable Hydropower Component Life Extension → Scenario. Prism → Sustainability Directory. https://prism.sustainability-directory.com/scenario/predictive-ai-for-sustainable-hydropower-component-life-extension/
[15] Wulfran Fendzi Mbasso, Harrison, A., Idriss Dagal, Pradeep Jangir, Khishe, M., Hossam Kotb, Shaikh, M. S., Aseel Smerat, Donfack, E. F., & Kumar, R. (2025). Digital twins in renewable energy systems: A comprehensive review of concepts, applications, and future directions. Energy Strategy Reviews, 61, 101814–101814. https://doi.org/10.1016/j.esr.2025.101814
[16] Staff Report. (2025, November 8). Pumped storage emerges as the backbone of grid reliability. Electrical & Power Review. https://www.eprmagazine.com/power-talk/pumped-storage-emerges-as-the-backbone-of-grid-reliability/
[17] Wu, R., Wang, R., Hao, J., Wu, Q., & Wang, P. (2024). Multiobjective multihydropower reservoir operation optimization with transformer-based deep reinforcement learning. Journal of Hydrology, 632, 130904–130904. https://doi.org/10.1016/j.jhydrol.2024.130904
[18] Muser, T., Krymova, E., Morabito, A., Seydoux, M., & Vagnoni, E. (2025). Fatigue damage reduction in hydropower startups with machine learning. Nature Communications, 16(1), 2961. https://doi.org/10.1038/s41467-025-58229-z
[19] Bechara, H., Ibrahim, R., Ryad Zemouri, Bachir Kedjar, Arezki Merkhouf, Tahan, A., & Kamal Al-Haddad. (2024). Review of Artificial Intelligence Methods for Faults Monitoring, Diagnosis and Prognosis in Hydroelectric Synchronous Generators. IEEE Access, 1–1. https://doi.org/10.1109/access.2024.3502546
[20] Zhao, Z., Li, D., She, J., Zhang, H., Zhou, Y., & Zhao, L. (2021). Construction and Application of Digital Twin Model of Hydropower Plant Based on Data-driven. 2021 3rd International Workshop on Artificial Intelligence and Education (WAIE), 19, 60–64. https://doi.org/10.1109/waie54146.2021.00020
The content & opinions in this article are the author’s and do not necessarily represent the views of AltEnergyMag
Comments (0)
This post does not have any comments. Be the first to leave a comment below.
Featured Product
