An Investigation of Photovoltaic Power Optimization

This work is concerned with the effects of low frequency switching with a non-linear step size DC-to-AC direct conversion on systems that employ photovoltaic (PV) power as an input source. These techniques exhibit dramatic reductions in circuit complexity and power dissipation compared to traditional pure sine-wave inverters. However, designing multilevel DC-to-AC inverter-less systems with maximum efficiency and reduced circuit complexity is challenging. This project has produced two novel multilevel DC-to-AC inverter-less systems for PV applications that reduces the power dissipation and circuit complexity to minimum. First, a DC-to-AC direct conversion system based on two ladders of switches structured as in the Golomb ruler is implemented at 600 Hz switching frequency. The technique has been evaluated analytically, experimentally and by simulation. The evaluation prove the reduction in series resistance loss of Golomb structure over the conventional contiguous block arrangement. It is also shown that Golomb structure can cause uneven PV panel utilization. In order to ensure even panel utilization and produce a good-quality sine-wave output signal, a novel cyclic selection multilevel inverter-less system is developed. The need for the magnetic materials is removed by selecting series and parallel combinations of PV cells. Closed-form cyclic selection expressions are derived and analytically investigated. Calculations prove that this system has exactly the same performance as a conventional magnetic core-based inverter. Laboratory and simulation results prove the efficiency of the system in PV applications. The task of determining the timing steps of the multilevel signal of known amplitude is resolved using a completely new mathematical method to minimize harmonic distortion. Closed-form expressions for the timing steps method are derived for three levels signal and the methodology is extended to an unlimited number of levels. This method gives the same results as the well-known Fourier series method but requires no carefully-chosen optimization starting point. Finally, a multiple step, forward-backward algorithm is employed to maximize the power into a given load. This work uses the second order central difference theory to establish an approximate equation for MPPT. This approach is unlike previous work in that it does not rely on a PV side tracking. Laboratory and simulation results prove the efficiency of the algorithm in rapidly varying insolation conditions.