مشخصات پژوهش

In today’s world, power electronics systems play a crucial role in providing sustainable and efficient energy for various devices and systems. One of the most significant challenges associated with DC-DC converters is input current ripple, which can have numerous negative impacts on the performance of power supply systems and the stability of electrical grids. Input current ripple not only reduces system efficiency but can also generate electromagnetic interference (EMI) and decrease the lifespan of the power supply or solar panel connected to the converter. Therefore, reducing input current ripple is an unavoidable necessity for enhancing the quality of power electronic converters. Multi-phase DC-DC converters are recognized as an effective solution for reducing input current ripple due to their ability to distribute power among different phases. However, the design and control of these converters, especially under conditions of increasing phase counts or rapidly changing system loads, become complex challenges. In such scenarios, traditional control methods may not provide a quick and optimal response. Here, neural networks emerge as a powerful and flexible tool. Neural networks, with their high learning capability and adaptability, can model system behavior in real time and provide optimal control strategies for reducing current ripple. The fast processing speed of neural networks and their ability to analyze complex, nonlinear patterns make this approach highly applicable in modern power systems. In this research, neural networks will be employed to control multi-phase DC-DC converters to significantly reduce input current ripple. Another advantage of multi-phase converters using this approach is the possibility of employing multiple converters in parallel at the input. This capability enhances the system's flexibility and allows for greater power delivery without a significant increase in input current ripple. The use of parallel converters can also lead to uni