This dissertation had investigated a high efficiency multilevel inverter topology which has great potential for application in stand-alone renewable energy systems. The presented study was focused on small systems (< 10 kWp) and it was showed that most typical system configurations store energy in battery banks in order to overcome the intermittence and peak power limitation of renewable energy sources. In this context, it was also identified that these systems require inverters with improved characteristics of reliability, capability to start heavy loads, efficiency and robustness. In fact, all these benefits can be achieved through the use of multilevel topologies.
Although several multilevel topologies have been successfully employed in industrial and power systems applications, it has been showed that only few topologies are suitable to implement inverters for small stand-alone systems, where usually only one single DC voltage source is available and the load is not precisely defined. According to these requisites and considering only the topologies based on low-frequency switching, it was found that the multiple-transformer and multi-winding-transformer topologies are the most efficient; the latter is the topology adopted in this work.
In both multiple-transformer and multi-winding-transformer topologies, major losses are associated to conduction losses in the switches and transformer losses. Knowing that transformer conduction losses are more representative than conduction losses in current available semiconductor switches, it can be concluded that the adopted topology can achieve the best feasible efficiency characteristic because the load current is shared among transformer coils and shunt switches.
The use of low-frequency switching (< 3 kHz) implies in negligible switching losses and consequently high efficiency. In addition, low-frequency operation allows the use of rugged snubbers that limit dv/dt variations without inserting appreciable losses. Considering that temperature rise and voltage stress are the main factors that decrease semiconductors lifetime, then it is expected that higher efficiency and individual switch protections can contribute to increase inverter reliability. Also, individual snubbers can be very effective against unpredicted fault conditions which can lead to inverter failure.
Apparently, the line-frequency transformer and relatively high number of components employed in the adopted topology appear to be serious drawbacks. However, additional weight and volume required by the transformer is not a problem for most systems, once these inverters normally remain in the same place after their installation. Also, it is well known that amount of components does not imply necessarily in reliability decrease and the use of more switches can be even a benefit if it is considered that power dissipation are distributed among them.
The adopted topology was investigated in detail and the presented results are useful for future designs. It was found that acceptable THD (< 5%) and voltage regulation (+ 5 %, -10%) can be guarantied in practice by structures with at least 4 output cells. Related to the transformer specification, it should be noted that its rated power must be higher than the inverter’s rated power (about 30% higher) because the particular voltage and current relations required by the adopted topology. It is also showed that optimum transformer design must consider that the load profile of typical stand-alone systems is concentrated around a small fraction of its rated power.
A new method to identify and correct unbalanced load condition was developed specifically to the proposed inverter. It allows the inverter to operate even with pure half-wave loads and also guarantees smooth inverter startup. The introduced self-control mechanism, which makes use of the H-bridge snubbers, implements an automatic fine balancing of the transformer magnetization current. Although the hold-on-at-zero intervals introduced by the balance-control method distort the output voltage waveform (THD is increased in about 2 %), it contributes to decrease the transformer magnetization current and no-load losses.
Ideally, each output-stage switch should be controlled independently and in accordance with the output voltage polarity and load current direction. However, in order to simplify the implemented circuits and to reduce cost, a simplified control was adopted, in which both switches that compose a bi-directional switch are controlled by the same signal. In consequence, the output-stage switch snubbers can be submitted to the full load current during dead-time periods. In this case, special attention must be given to the design of the output-stage in order to avoid voltage stress across the switches and over-current thought the snubber zenner diodes.
A 63-level (5-cell output-stage) / 3 kVA prototype was implemented to validate the proposed inverter. Experiments with adjustable resistive, inductive and capacitive laboratory loads were done to trace the prototype characteristic curves, and several standard appliances, varying from computers to heavy duty sawing machines, worked properly when supplied by the proposed inverter. The prototype was capable to feed experimental half-wave loads up to 1.5 kVA without any problem. Bi-directional power flow capability was demonstrated though the operation of highly inductive and capacitive loads and battery-charge-mode was validated by connecting grid-interactive converters directly to the inverter output.
Peak efficiency of 96.0% and no-load consumption of 18.6 W were measured at rated input voltage (48 V). As expected, good quality waveform was achieved, presenting THD of less than 4% and static output voltage regulation within + 5%, -6%. Maximum measured surge power capability was 4.500 W (1.5 times the rated power), which was limited by the inverter protections. However, according to the prototype design, it is expected that surge power up to 2.5 times the rated power can be achieved.
Considering a typical load profile and in comparison with high-quality commercial inverters of similar power and voltage ratings, it was found that the implemented prototype can achieve the best efficiency performance for any load demand greater than 270 kWh/month. Moreover, through solely modifications in the design of the multiwinding transformer, it was demonstrated that it is possible to achieve peak efficiency of 96.3% and superior performance for any practical load demand.
In fact, it is expected that the proposed inverter can reach very high efficiency (> 98%) as a cost of higher price and volume. This is possible because switching losses are negligible and conduction losses can be decreased as much as desired by using better and/or oversized components.
Regarding cost, slightly high cost is associated to the proposed inverter due to the relatively large number of switches and isolated drives required by the adopted topology. Considering that a high-quality 3 kVA inverter currently present an average price of €2100, then it is estimated that the proposed inverter can cost around 5% more. However, it is expected that cost difference can be justified by its improved characteristics.
The most relevant contributions of this work are:
♦ The proposed topology was implemented with modern components and experimental results showed that it can present superior performance when compared to similar high-quality inverters.
♦ Detailed analysis of the adopted topology is presented, giving support for future implementation and possible commercial application.
♦ A new method for preventing transformer-unbalancing was developed and validated.
♦ Software tools were developed to aid the design and simulation of high-resolution multilevel inverters.
