**1. Introduction**

The role of power-converters is growing rapidly in importance. In part, this is due to an increased use of renewable energy, which gets injected into the grid as electricity that must satisfy minimum quality regulations (Hammons 2011).

One of the features of using multilevel converters, as opposed to one level converters, is that they operate with different output voltage levels and as a result, a lower harmonic distortion coefficient is achieved. Moreover, these topologies allow working with higher voltages than the transistor break-down voltage, therefore allowing higher power ratings (Bueno 2005).

NPC multilevel converters can be connected to photovoltaic panels in such a way as to optimize the efficiency. This means that fewer panels need to connect directly to the grid, as opposed to two-level topologies, while peak line voltage does not exceed DC-bus voltage.

Fig. 1. Three-phase NPC converter, connected to a pair of independent photovoltaic-panel arrays.

There is however a drawback that comes up in the use of photovoltaic panels. Only low voltage can be supplied by these devices. This means that they must be connected in series, in order to achieve the desired voltage. Unfortunately, the lowest current supplying element sets the maximum current generated by the array.

Minimizing the number of panels connected in a series improves array efficiency once the converter is directly connected to a 400V grid and NP is connected between the two halves of the panel array. Furthermore, as these two arrays are independent, both can be set to inject maximum power at any moment.

Recently, there has been some work that has focused on transformerless photovoltaic inverters, and the influence of current leakage (Gonzalez et al., 2008), (Kerekes et al., 2009) (Kerekes et al., 2007). This is especially important for human saftey. The German standard, VDE0126-1-1, deals with grid-connected PV systems, and gives the requirements for limiting ground leakage and fault currents. These works coincide in that the NPC is an ideal topology for regulatory compliance, and the algorithm proposed minimizes the NP ripple voltage, one reason why the current leakage exists. The voltage level achieved by the capacitors is studied in several works, which focus on the NP point ripple (Bueno et al., 2006), (Celanovic & Boroyevich 2000), (Ogasawara & Akagi 1993) (Qiang et al., 2003).

(Pou et al., 2007) proposes to eliminate low frequency ripple at the NP using two modulations which contribute to an increase in switching frequency for transistors, and therefore power losses. Work by (Cobreces et al. 2006), based on a single-phase inverter, applies a changing state strategy in a specific duty cycle which also contributes to an increase in switching frequency.

Usually, there is no intention to tackle the asymmetric supply issue since it is very common to implement the same voltage for both capacitors thus avoiding an independent power supply implementation.

This document takes a look at this problem and tries to increase the efficiency of renewable energy generation. In Section 2, system characteristics are determined, and we introduce an improved proposal for (Pou et al., 2007) reducing NP ripple voltage in ideal conditions: plugging symmetric or minimal asymmetric power supply into the DC-bus. Section III explains the underlying principle of the proposed algorithm and implementation details. Section IV shows simulations made under "ideal" power supply conditions, and analyzes their limits, characteristics and advantages. In Section V, the power supply imbalance issue, due to the fact that power supplied by PV1 and PV2 are different, is discussed: the imbalance tolerance limit will be shown analytically. Finally, in Section VI, conclusions will be drawn.
