Conclusion

This work has dealt with the various aspects of MPP in PV and the basic implementation of a solar module optimizer. In this regard, attention was first drawn to the existing problem of shading and then the technologies for a solution were presented. The concept of power adjustment and the operation of the boost converter were explained. The change of the input resistance of the step-up converter by adjusting the duty cycle was discussed in more detail.

Subsequently, some requirements were elaborated and a concept for the operation of the solar module optimizer was developed. The idea of the concept is to feed the continuous measurement of current and voltage at the solar module to a programmable controller. Based on its programming, this controller then decides how to set the control variable for the boost converter in order to find the MPP.
According to this concept, the development of the hardware and the related issues were documented in detail. The coil and capacitor were sized for the boost converter. Solutions for measuring the current and voltage were developed, and a voltage divider was sized and a Hall-effect sensor was selected. A suitable mosfet was chosen as the switching element and a suitable gate driver was selected for it. Furthermore, a snubber cicuit was designed to dampen the unwanted oscillations when switching the mosfet. In addition, a low-pass filter was sized to reduce noise at the adc. To be able to program the solar module optimizer later, a microcontroller was selected. Furthermore, the hardware provides access to an SD card reader.

Next the software was implemented, for this a class was written to separate the basic hardware access from the MPP algorithm. In addition, various drivers for the peripherals of the solar module optimizer were developed. Here, a method was developed to drive the duty cycle for mosfet of the boost converter. Conversion functions for current and voltage measurements were developed. A class for filtering the measurement data was developed for further removing noise from the ADC. A PID was implemented to control the output voltage to set point and a recording function for the SD card reader was implemented to record the measurement data for later analysis. Subsequently, two algorithms were implemented to search for the MPP, first the PO which is most commonly used, and the VI scanner algorithm. With the developed hardware and software, the solar module optimizer was tested for its basic functionality, the ripple of the output voltage was determined and a VI-characteristic curve of a solar module was recorded. For the investigations under constant laboratory conditions of a vi-characteristic curve of a PV-module, a PV-emulator was developed. This provides constant characteristics of the emulated solar module in order to compare the algorithms.

The PV emulator was then used to investigate and compare both algorithms, once without shading and once with shading. Here it turned out that the PO method cannot find the global MPP and dwells in a local MPP. The VI scanner turns out to be the more effective algorithm, and finds the global MPP immediately.

One of the research questions was the feasibility of such a project using the Micropython programming language, the system needs appropriate real-time capability to reliably set the pwm with the frequency of $50kHz$. There must also be enough computing power to read the adc, process the data, and respond. The feasibility was demonstrated with the successful experiment of studying a VI characteristic. This has shown how much faster development occurs with Micropython than with a traditional language such as C++. The limitations of Micropython, such as significantly slower processing and a limited functionality of a few functions, are less relevant in the work, because the project does not require a fast response in the microsecond range. 

The consideration in the concept of the solar module optimizer to use a PI controller for the output voltage, and to supply it with a reference voltage from the MPP controller was a mistake. This lengthens the controlled system by several elements, resulting in unpredictable oscillations.

Overall, the goal of implementing a solar module optimizer for PV-modules was successfully realized, with minor and major challenges that were solved with good approaches. The final practical test, despite some weaknesses, highlights the superiority of the VI-scanner over the PO. Setting the MPP working point for the next $60$ seconds is one such point, if the characteristic changes during this time, it results in a loss of performance. A possible conclusion for this, is to combine both algorithms. Here, the VI characteristic could be sampled first, and then the PO-algorithm could be activated in a small range around the global MPP for the next $60$ seconds. Here, the step size could be much smaller than the sampling of the VI scanner.