In a paper published in the journal Scientific Reports, researchers unveiled an advanced sustainable power management system for light electric vehicles (LEVs) featuring a hybrid energy storage solution (HESS) integrated with machine learning (ML) for precise control. This innovative system leveraged renewable energy sources like photovoltaic (PV) panels and supercapacitors to overcome traditional battery limitations.
Study: Machine Learning-based Sustainable Power Management for Light Electric Vehicles. Image credit: BLKstudio/Shutterstock
Notably, the ML-based control optimized power distribution among the battery, supercapacitor, and PV sources, ensuring stringent voltage regulation of the DC bus while minimizing torque ripple and transient response times. Simulation results validated its robust performance, promising a more sustainable and efficient future for LEVs.
Related Work
Past work in the field has extensively investigated the integration of HESS and ML-enhanced control for optimizing power management in LEVs. Traditionally, LEVs relied solely on batteries, which posed limitations in energy density, charging times, and lifespan. Supercapacitors emerged as a promising complement to batteries, offering high power density and rapid charge–discharge characteristics and integrating renewable energy sources like photovoltaic (PV) panels further enhanced sustainability.
However, effectively managing power flow between batteries, supercapacitors, and PV panels posed challenges, especially in dynamic LEV systems. Traditional control strategies often struggled to optimize real-time power flow, leading to suboptimal performance and reduced battery life.
Power conversion innovation
Researchers developed a hybrid power supply (HPS) to downsize the power conversion interface for electric vehicle drives, integrating battery power into a direct current (DC) bus in two cascaded stages and PV power in one stage. This innovative design aimed to maximize power absorption from the PV source and ensure efficient transfer to the DC bus, employing a boost converter for unidirectional power flow from the PV panels.
Furthermore, bi-directional converters facilitated power transfer between the DC bus and the battery/supercapacitor interface, offering flexibility in energy management. This topology yielded advantages in terms of reduced inductor size and minimized voltage stress on power switches compared to conventional setups. Additionally, the switched reluctance motor (SRM) power converter was streamlined with fewer switches, enhancing efficiency and simplicity in operation.
The researchers developed the mathematical models governing the dynamics of the HPS and SRM converter to ensure optimal performance. For the PV converter differential equations to capture instantaneous current and voltage, they derived, with maximum power point operation as a critical focus.
Similarly, the researchers established equations governing the switching of the supercapacitor and battery-supercapacitor interface converters, enabling precise power flow control. They elaborated on the dynamics of the SRM converter in parallel, emphasizing the importance of rotor flux space vector magnitude and position in designing direct torque control (DTC). Combined with advanced control strategies, these models aimed to enhance efficiency, responsiveness, and overall performance in light electric vehicle systems.
Intricate control strategy
The control strategy for the proposed system is intricate, comprising multiple interconnected layers to ensure efficient operation of the PV-assisted EV drive. Like a pattern recognition ML algorithm, the first layer sets instantaneous torque based on driving patterns, estimates PV power output, and tracks the maximum available PV power.
The second layer utilizes mathematical models to estimate motor speed and controls the HPS, adjusting power flow from PV, battery, and supercapacitor. The final layer coordinates power flow throughout the interface, optimizing distribution among sources to maintain stable DC bus voltage and regulate system response to load changes.
Mathematical models and model reference adaptive controllers estimate motor speed without speed sensors, while a sophisticated control scheme regulates power distribution within the HPS system. Coordinated control of the SRM drive ensures smooth operation and efficiency, with torque and flux hysteresis components informing instantaneous voltage vector determination and switch control for current flow optimization into the SRM.
Simulation and performance assessment
Researchers conducted simulation results and performance evaluation using MATLAB/SIMULINK to assess the proposed drive's efficiency across various scenarios mimicking real-world electric vehicle operations. The ML algorithm's accuracy in generating torque references and estimating PV power, coupled with its ability to identify maximum power points, was rigorously evaluated.
Researchers analyzed the regulation and power distribution of the HPS among sources, highlighting precise control and seamless adaptation to load variations. Moreover, the drive's response to torque and speed demands, including sudden reversals, demonstrated its agility and stability. Comparative analyses against existing power supplies and evaluations of component sizing underscored the proposed system's superiority in robustness, accuracy, and efficiency, setting a new standard for PV-assisted EV drives.
Conclusion
To sum up, the study presented a novel PV-assisted EV drive system with advanced control strategies and efficient power management. Simulation results validated the effectiveness of the proposed ML algorithm and control schemes, showcasing precise torque control and robust performance across various scenarios. Comparative analyses demonstrated the superiority of the proposed system in terms of DC bus regulation, component sizing, and overall efficiency compared to existing power supplies.
Future research directions included advancing control strategies, exploring innovative converter designs, improving battery management techniques, optimizing system performance, and accelerating commercialization efforts through collaboration and testing.
Machine Learning Enhances Water Quality Monitoring