Hybrid AI Controller Boosts Solar PV Efficiency to 99.8%
A new study published in Discover Energy presents a hybrid control architecture designed to improve the efficiency of solar photovoltaic systems under rapidly changing environmental conditions. The research, led by J. Viswanatha Rao and colleagues, proposes a Maximum Power Point Tracking (MPPT) technique that integrates an Adaptive Neuro-Fuzzy Inference System (ANFIS) with a Random Forest (RF) algorithm. This approach aims to address the limitations of traditional MPPT methods, which often struggle to maintain optimal power extraction when solar irradiance fluctuates due to cloud cover or partial shading. The proposed controller utilizes an adaptive decision-level fusion mechanism to combine the nonlinear approximation capabilities of ANFIS with the ensemble learning robustness of Random Forest.
The study focuses on the challenge of tracking the global maximum power point in large-scale rooftop solar installations, specifically those in the kilowatt and megawatt ranges. In such systems, irradiance and temperature can change unexpectedly, causing conventional algorithms like Perturb and Observe (P&O) and Incremental Conductance (IC) to fail. These traditional methods are prone to steady-state oscillations, slow responses to rapid irradiance changes, and the risk of getting stuck at local maxima during partial shading. The researchers argue that these drawbacks are particularly significant in systems with multiple MPPT channels, where fast and accurate tracking is crucial for maintaining overall system stability and efficiency.
To overcome these issues, the authors developed a hybrid framework that merges the adaptive nonlinear learning abilities of ANFIS with the strong prediction and decision-making skills of Random Forest. The ANFIS component is responsible for smooth nonlinear inference and steady-state control, while the RF component provides fast estimation capabilities during transient conditions. The two models operate in parallel, receiving preprocessed and normalized input data including PV voltage, current, irradiance, and temperature. An adaptive weighting mechanism then adjusts the contribution of each model based on real-time tracking error and operating conditions, ensuring a balance between dynamic response time and steady-state accuracy.
The performance of the proposed hybrid controller was validated through comprehensive simulations on a 540 kWp rooftop PV system. The simulation environment was designed to replicate realistic rooftop conditions, subjecting the system to rapidly varying irradiance levels ranging from 600 to 1000 W/m². This setup allowed the researchers to evaluate the controller’s ability to handle nonlinear relationships between output power and voltage under dynamic stress. The study also included a detailed analysis of the ANFIS membership functions and rule base, which were configured to ensure smooth interpolation between operating points while maintaining sensitivity to abrupt irradiance transitions.
The results demonstrate that the hybrid ANFIS–RF controller achieves a peak MPPT efficiency of 99.8%, significantly outperforming standalone methods. In comparison, the Artificial Neural Network (ANN) controller achieved an efficiency of 98.9%, the standalone ANFIS controller reached 98.5%, and the conventional Random Forest controller recorded 97.3%. The hybrid approach also demonstrated superior stability, achieving the lowest power ripple percentage of 1.2% and the fastest settling time of 1.0 second. These metrics indicate that the hybrid controller can track the maximum power point more accurately and stably than its individual components when used in isolation.
Further analysis of the tracking performance reveals that the hybrid controller significantly reduced the Root Mean Square Error (RMSE) to 190 W. This is a substantial improvement over the RF controller, which had an RMSE of 920 W, and the ANN controller, which recorded 510 W. The standalone ANFIS controller also showed an RMSE of 410 W, but the hybrid approach proved more effective in minimizing tracking error. The study notes that the RF-based MPPT controller, while fast in initial response, suffers from pronounced steady-state oscillations, resulting in a higher standard deviation of extracted power. The hybrid design mitigates this issue by leveraging the smooth control action of ANFIS during steady-state operation.
The instantaneous tracking ratio, defined as the ratio of extracted power to the theoretical maximum power, was another key metric evaluated in the study. The hybrid RF–ANFIS controller achieved an average tracking ratio of 0.999, with a maximum of 1.000 and a minimum of 0.990. This corresponds to an overall tracking efficiency of 99.9%. In contrast, the conventional RF controller averaged a tracking ratio of 0.965, the ANN controller reached 0.987, and the ANFIS controller achieved 0.992. The high tracking ratio of the hybrid controller confirms its ability to maintain precise and stable maximum power point tracking with minimal deviation under changing environmental conditions.
The adaptive fusion mechanism plays a critical role in the controller’s performance. During periods of rapid transition in irradiance levels, the system assigns more weight to the RF output to utilize its rapid estimation ability. Conversely, during steady-state or slow transitions, more weight is given to the ANFIS output to ensure consistent control action and limit oscillations. This dynamic adjustment allows the controller to adapt to the specific needs of the PV system in real-time. The training process for the models is completed offline, with only predictive and adaptive fusion performed online, which enhances computational efficiency and facilitates embedded deployment in actual PV installations.
The study also highlights the importance of data preprocessing in ensuring the reliability of the models. The dataset was filtered to remove noise caused by sensor inaccuracies, and missing or damaged data was handled to improve data quality. All inputs were normalized to fall within a common numerical range, preventing model bias towards variables with larger numeric magnitudes. This preprocessing step is especially important when irradiance conditions are constantly changing, as raw measurement data can exhibit significant variance. By ensuring consistent predictions and removing noise, the ANFIS and RF models can effectively learn strong, nonlinear relationships between input variables and the maximum power point.
While the results are promising, the study acknowledges that the validation was conducted through simulations rather than physical hardware implementation. The authors note that future research will focus on integrating deep learning, reinforcement learning, and advanced optimization techniques to further improve MPPT accuracy under dynamic and partial shading conditions. Additionally, real-time hardware implementation using FPGA/DSP platforms, combined with IoT-based monitoring and Digital Twin concepts, is planned to enable predictive maintenance and intelligent energy management. The current study provides a strong foundation for the development of smart grid PV applications, demonstrating that the hybrid RF–ANFIS approach offers superior robustness and scalability compared to traditional intelligent MPPT methods.
In detailing the structural design of the parallel framework, the study explains how the two models simultaneously evaluate preprocessed operational parameters comprising photovoltaic voltage, current, solar irradiance, ambient temperature, and state of charge. Within the neuro-fuzzy branch, input variables pass through fuzzification layers where premise and consequent parameters are optimized using gradient descent and least-squares estimation. To modulate the converter duty cycle, input variables such as array voltage and solar irradiance are categorized across low, medium, and high operating regions using triangular membership functions. Meanwhile, the ensemble branch employs random forest regression, creating numerous decision trees through bootstrap sampling and synthesizing their predictions via ensemble averaging. Although these decision trees quickly isolate optimal operating regions during abrupt environmental shifts, their discrete decision boundaries inherently produce steady-state ripple, which the continuous fuzzy inference branch helps smooth.
The simulation experiments also captured notable variations in power stability and operational variance across the evaluated architectures. Under dynamically changing solar irradiance, the proposed hybrid controller achieved a mean extracted power of approximately 121 kilowatts while constraining the standard deviation of extracted power to roughly 22 kilowatts. In comparison, the standalone neuro-fuzzy model exhibited higher power dispersion with a standard deviation of about 28 kilowatts, whereas the standalone random forest controller exhibited pronounced steady-state oscillations that increased power standard deviation to approximately 35 kilowatts. Dynamic stability assessments further reinforced these distinctions, with the hybrid architecture attaining a stability index of 31.5 compared to a score of 12.5 recorded for the conventional random forest baseline. Furthermore, the system maintained an instantaneous tracking ratio exceeding 0.95 for 89 percent of the total operational duration.
Examining the broader field of solar energy harvesting, the research team, which includes Jagarapu S. V. Siva Kumar, Juvvanapudi Sharon Rose Victor, Kiran Kumar Pulamolu, Deepthi Janyavula, and Rajanand Patnaik Narasipuram, contrasted their framework with existing maximum power tracking literature. The authors observed that although earlier efforts investigated combinations of artificial neural networks, fuzzy systems, and evolutionary algorithms, many prior methodologies suffered from insufficient quantitative benchmarking or were validated only on small-scale arrays under narrow test profiles. Standalone artificial neural networks require substantial training datasets and struggle with generalization in untested outdoor conditions, while fuzzy controllers rely heavily on predefined expert rule bases. By pairing random forest estimation for rapid transient tracking with neuro-fuzzy inference for smooth continuous modulation, the researchers aimed to resolve longstanding trade-offs between dynamic response speed and steady-state precision.
In their concluding evaluation, the authors highlight that the adaptive hybrid architecture minimizes power losses, maximizes energy harvesting, and mitigates physical stress on power electronics by stabilizing operational voltages under nonlinear conditions. Across extended comparative testing, the framework delivered an overarching tracking efficiency of approximately 99.2 percent, constrained power ripple to roughly 1.45 percent, and settled within 1.54 seconds. These operating characteristics reduced root mean square error by up to 38 percent against traditional baselines, improved transient response by an estimated 40 percent, and achieved an error reduction exceeding 80 percent compared to standalone random forest tracking.
Subject of Research: Hybrid ANFIS-Random Forest MPPT control for solar PV systems
Article Title: A next-generation AI-driven random forest MPPT control architecture for solar energy extraction
Article References: Rao, J. V., Siva Kumar, J. S. V., Victor, J. S. R., Pulamolu, K. K., Janyavula, D., & Narasipuram, R. P. (2026). A next-generation AI-driven random forest MPPT control architecture for solar energy extraction. Discover Energy, 6(1), Article 81. https://doi.org/10.1007/s43937-026-00176-x
Image Credits: AI Generated
DOI: 10.1007/s43937-026-00176-x
Keywords: Solar Energy, MPPT, Artificial Intelligence, Random Forest, ANFIS, Photovoltaic Systems, Control Architecture, Renewable Energy, Simulation, Energy Efficiency, next-generation, AI-driven
News Source: Faith Mcneil. (November 11, 2026). Hybrid AI Controller Boosts Solar PV Efficiency to 99.8%. Scienmag.
We bring you the latest biotechnology news from best research centers and universities around the world. Check our website.
Enter your email address to subscribe to this blog and receive notifications of new posts by email.
Email Address
Subscribe
Bioengineer.org © Copyright 2023 All Rights Reserved.
Login to your account below


Remember Me

Please enter your username or email address to reset your password.


Bioengineer.org © Copyright 2023 All Rights Reserved.

source