Performance comparison of solar PV fed PI, MPC & AI controller for fast battery charging system in Electric Vehicles

The road to a cleaner future is paved with electric vehicles, but for these cars to truly replace the internal combustion engine, they must overcome two stubborn hurdles: the anxiety of running out of power and the frustration of waiting hours to recharge. While the technology to store energy in batteries exists, the challenge lies in how that energy is delivered. One promising solution is to power charging stations directly with sunlight. However, the sun is an unreliable partner; its intensity shifts constantly as clouds drift by or the day progresses. This fluctuation creates a chaotic environment for the delicate chemistry inside a lithium-ion battery. If the flow of electricity is too erratic or too aggressive, it can damage the battery or trigger safety systems that shut the charge down. To navigate this, engineers rely on a "brain" for the charging system—a controller that constantly adjusts the flow of energy, much like a skilled driver modulating the gas pedal to keep a car moving smoothly over a bumpy road.

Researchers at the Babu Banarasi Das Institute of Technology and Management, along with colleagues from Maharishi University of Information Technology, set out to determine which type of controller is best suited for this demanding task. They focused on three distinct approaches: a traditional method known as the Proportional-Integral controller, a sophisticated predictive system called Model Predictive Control, and a modern, data-driven approach using Artificial Intelligence. The team did not just theorize about these methods; they built a detailed simulation of a solar-powered charging station and tested how each controller reacted when the sunlight suddenly changed, mimicking a cloud passing over the solar panels. They then took their findings a step further by testing the controllers on physical hardware connected to a simulated environment, a process known as Hardware-in-the-Loop testing, to see how they performed when faced with the messy reality of electronic noise and sensor imperfections.

The traditional controller, which has been the industry standard for decades due to its simplicity, struggled significantly in these tests. When the sunlight dropped, this controller reacted with a jerky, oscillating motion. It overshot the target charging current by about fifteen percent, causing the system to wobble back and forth before finally settling. This instability meant the system took nearly half a second to stabilize, a delay that feels instantaneous to a machine but represents a loss of precious energy and potential stress on the battery. In the real-world hardware tests, this jittery behavior became even more pronounced, with the controller "hunting" for the correct setting and wasting energy as heat. Its overall efficiency in these noisy conditions dropped to roughly ninety percent, a figure that suggests a significant amount of harvested solar power was being lost rather than stored.

A more advanced approach, Model Predictive Control, offered a marked improvement. Instead of simply reacting to errors as they happened, this system uses a mathematical model to look ahead, predicting how the battery and solar panel will behave in the next few moments. This foresight allowed it to adjust the charging current smoothly, avoiding the violent overshoots seen in the traditional method. In the simulations, it stabilized the system in about one-tenth of a second with almost no overshoot. However, this intelligence comes with a heavy price: the system requires a powerful processor to solve complex optimization problems in real-time. When tested on actual hardware, the time it took to crunch these numbers introduced a slight lag, and the overall efficiency settled at just under ninety-eight percent. While robust, the researchers noted that the computational demand might make it difficult to implement in systems that need to switch power on and off extremely fast.

The most compelling results came from the Artificial Intelligence controller, which operates differently by learning from data rather than relying on rigid mathematical rules or predictive models. This system was trained on thousands of examples of how solar panels and batteries behave under various conditions. When the sunlight changed during the tests, the AI controller recognized the new pattern instantly and adjusted the charging current without hesitation. It reached a stable state in less than one-hundredth of a second, effectively eliminating the overshoot and oscillation that plagued the other methods. In the hardware tests, which included the interference of real-world sensor noise, the AI controller maintained a near-perfect tracking curve, ignoring the electronic static that confused the traditional controller. Its efficiency soared to nearly ninety-nine point seven percent, meaning almost all the energy captured from the sun was successfully transferred to the battery.

The study concludes that while the traditional controller is cheap and easy to build, it is ill-suited for the dynamic demands of fast solar charging. The predictive controller offers a strong middle ground but is hindered by its heavy computational requirements. The Artificial Intelligence controller emerged as the most effective solution, combining the precision of advanced prediction with a speed that allows it to run on standard automotive hardware. By learning the complex, non-linear behavior of the system, it manages to filter out noise and maintain stability where other methods falter. For the future of electric vehicles, this suggests that the key to reliable, fast, and efficient solar charging may not be in building better mathematical models, but in teaching the system to learn from experience.