The Solid-State Transformer Concept: A New Era in Power Distribution

Department of Electrical and Electronic Engineering University of Stellenbosch Stellenbosch, South Africa Email: wimvdmerwe@sun.ac.za

J.W. van der Merwe

Department of Electrical and Electronic Engineering University of Stellenbosch Stellenbosch, South Africa Email: dtmouton@sun.ac.za

H. du T. Mouton

AbstractThis paper considers the solid-state transformer concept as a viable alternative to the line frequency transformer in distribution circuits. The benets of the concept are presented in terms of improvements to problem areas currently experienced on distribution networks as well as envisaged future or fringe benets. Some barriers to universal acceptance of the concept are also identied, with the effects of these barriers analysed and discussed below.

I. I NTRODUCTION During the initial years of electric distribution a so-called war of currents was waged between the proponents of alternating (AC) and direct (DC) currents. On the one side Thomas Edison held several patents for DC distribution systems, the American standard at the time. The DC system had considerable advantages, for example easy integration with energy backup devices such as batteries to provide power backup during system failures. On the other side of the argument, AC, as advocated by Nicola Tesla and George Westinghouse, had the ability to transform high voltages, suitable for transmission over long distances, to low voltages for residential usage. Ever since the introduction of AC systems in 1887, the magnetically coupled transformer had been used for both voltage transformation and galvanic isolation. The basic construction of the transformer did not change much during the last century. Improvements in material processing and the creation of new alloys with high magnetic saturation densities and low hysteresis losses gave rise to more efcient transformers. Likewise, it became possible to design transformers with the aid of detailed simulation software such as nite element analysis packages. Currently transformer technology is very mature in that any increases in performance are marginal and come at great cost. The traditional line frequency transformer (LFT) provides a cheap and very efcient method for voltage level transformation and isolation. However, this direct transformation also introduces unwanted characteristics into the modern power grid, among which are the following: 1) The output voltage is a direct representation of the input. Any unwanted characteristics from the input, such as voltage dips or frequency variations, will be represented on the output.

2) The output current will likewise inuence the input current. In a practical distribution environment the input current will be a reection of the sum of the connected output currents. The presence of harmonics on the load currents will manifest themselves in the input current. For single phase residential customers odd harmonics especially the 3rd , 5th and 7th are notably present [1] with values in excess of 15 % of the fundamental. As with the total load the harmonics are rarely balanced among the three phases. With delta connected primary transformers the 3rd harmonic circulates in the primary winding and does not propagate to the network, but it does add to the primary winding losses. 3) In general, transformers are designed for maximum efciency at near to full load. This results in a high standing loss, or a loss under no load conditions. In a distribution environment a transformer operates on average at about 30 % load. 4) All LFTs suffer from non-perfect voltage regulation. In general, the voltage regulation of a LFT is inversely proportional to the power rating of the transformer. At distribution level the transformers are generally small and the voltage regulation is not very good. During the last century much was done to combat these negative effects. The output voltage regulation is counteracted through the use of automatic tap changing transformers and voltage regulating transformers. New NEMA standards for transformers stipulate that distribution transformers be designed for maximum efciency at 50 % load [2]. Many standards, regulating the consumer, with regard to the type of load have been accepted to combat the propagation of harmonic currents through the system. To improve power quality FACTS (Flexible AC Transmission Systems) devices have been developed for use in conjunction with LFTs. In 1980 the concept of a solid-state transformer (SST) was mentioned as an alternative to the LFT [3]. Although the technology for the successful demonstration of the concept did not exist at the time, the advantages of such a system were clear. With recent advancements in power semiconductors and understanding of multilevel converters, the concept became

more viable, resulting in renewed interest [4][6]. In this paper the SST concept will be discussed in terms of the perceived benets and changes it will bring to a next generation distribution system. The weaknesses of the SST concept and the barriers to universal acceptance will also be discussed. II. T HE SOLID - STATE TRANSFORMER CONCEPT The basic solid-state transformer topology consists of three main sections, i.e. a converter (or converters) to generate high frequency AC from the incoming line frequency AC, isolation through a high frequency transformer and nally a converter (or converters) to produce line frequency AC from the high frequency AC. The isolation barrier furthermore divides the transformer in two: the high and low voltage sides.
High Voltage Side 11 000 V 3 LF-ac to HF-ac Low Voltage Side HF-ac to LF-ac 380 V 3
Fig. 2.

A1 A2 A3

Cell Cell Cell

AN

Cell

B1 B2 B3

Cell Cell Cell

ia Vdc

va

A B

3 Phase Inverter

ib ic

vb

BN

Cell

vc

C1 C2 C3

Cell Cell Cell

control

ia,ib,ic

CN

Cell

A1out ANout B1out BNout C1out CNout

Main Controller

A1_in AN_in B1_in BN_in C1_in CN_in

Realisation of SST without an accessible MV DC bus.

Fig. 1.

Generalised solid-state transformer circuit. Fig. 3. Detailed circuit of a single cell.

The problematic area of the power conversion process as described in Fig. 1 is the high voltage side, both in the conceptualisation of the converter needed to create the high frequency signal, due to the high voltage, and in the isolation transformer due to isolation and parasitic issues. Although semiconductor technology improved much during the last two decades, the voltage blocking capability of power semiconductors is much lower than the required 22 kV for the high voltage side. This limitation can be circumvented through the use of different multilevel topologies [7], all utilising several switches with lower voltage capabilities in series, to allow for the high voltage requirement. The main multilevel topologies are the ying capacitor, diode clamped, cascaded converter and the input series converter. As a rst step in the SST development, the basic characteristics of the different topologies were compared, especially in terms of switch, diode, capacitor, driving circuitry and control requirements [8]. Although it might be possible to realise the SST using a single AC-AC converter, the availability of a LV DC bus for integration of additional energy storage systems is a signicant advantage. The most feasible conguration of the SST is of the form AC-DC-AC-DC-AC, either with the MV DC bus available and stable [5] or without an accessible MV DC bus as used in [9]. The total SST circuit diagram, for a system without an accessible MV DC bus, is shown in Fig. 2. The MV side is created through a cascaded connection of series cells detailed in Fig. 3. This modular MV AC grid connected topology is also suggested for other systems, such as drives for the German 15 kV, 16 2 Hz train systems [10] and FACTS 3 devices [11]. Should bi-directional power ow be required, the

passive rectier on the output could be replaced by an active rectier. III. T HE BENEFITS OF THE SST IN TERMS OF EXISTING
PROBLEMS

The SST concept can address some of the problems experienced in distribution systems. In a direct comparison with the LFT drawbacks, as mentioned in the introduction, the SST addressess all the problem areas. The perceived benets of the SST concept will be discussed in terms of power quality, protection and deep rural feeder (and SWER) applications. A. Power quality The SST concept will improve power quality. Although the major improvement will benet the costumers connected to the SST, the concept will also marginally improve the power quality of other users on the same network. The power electronic converters on both sides of the isolation barrier, see Fig. 1, effectively separate the input voltage shape from the output voltage shape. This separation brings immediate benets to the consumer connected to the low voltage side of the SST. Some of the more important advantages include: 1) The SST operates with an intermediate DC energy storage capacitor from where the output voltage is created. The voltage of the DC capacitor can be regulated by the front end converters through a wide range of input

Fig. 4.

Implementation of the SST in a possible future grid.

voltages. Therefore the output voltage will be immune from input voltage sags and dips within a designed band. 2) The output inverter operates with an output voltage control loop, implying that the output voltage will be regulated throughout the load range. This results in a near perfect voltage regulation. 3) The output voltage will also be pure sinusoidal and free of power frequency harmonics, regardless of input voltage shape. For practical purposes the input of the SST will mimic a variable resistance, whereby the resistance will vary according to the required output power and the input current will be sinusoidal in shape and in phase with the voltage. Since any current harmonics generated by the low voltage load do not propagate back to the high voltage side, any other loads connected to the high voltage side will not suffer the detrimental effects of the harmonics. The active three phase rectier also isolates any phase imbalances on the secondary from the primary. The system would reect a pure balanced three phase load to the primary network, even in the event of a single phase load connected to the secondary of the SST. The isolation between the primary and secondary requires some form of intermediate energy storage, normally added on the LV DC bus. For example, when only a single phase load is connected, the output power will have a pulsating waveform. The MV side converters will operate with a constant bus voltage and under current mode control, hence delivering constant power to the LV DC bus. Assuming the inverter delivers power to a single phase resistive load, the output

power po (t) will be po (t) = P (1 cos(2t)), (1)

where P denotes the average power delivered to the load. If the output bus voltage is allowed to uctuate, to allow for energy exchange, the bus capacitance can be used to store the excess energy. The amount of energy that must be stored in the bus capacitance, Wc , can be found through integration of the power through the capacitor over an exchange cycle:
T 8

Wc

= =

2P
0

cos(2t)dt (2)

1 P

The energy stored in a capacitor is described as 1 CV 2 . If the 2 variance of the capacitor voltage is expressed as an fraction () of the nominal bus voltage, VN , the capacitance required to store a specied amount of energy, given the voltage limits, is: Wc C= (3) 2 VN Therefore the bus capacitance needed to sustain a single phase load, while reecting a constant power load to the primary (and hence a balanced three phase load), can be found as: C= P 2 VN (4)

The energy requirements are modest. For a 80 kVA system, operating with a 800 V DC bus, only 2.65 mF capacitance would be required to limit the bus voltage uctuation between

760 V and 840 V. Similarly, the bus capacitance to prevent a 15 % third harmonic current, at full load, from propagating to the primary side is found as 1 mF. However, the correction of voltage uctuations does require more energy storage, depending on the length and depth of the event. It is possible to correct such events through reecting a negative input resistance to the grid (increasing current as input voltage decreases) but such a practice on a large scale might adversely inuence network stability. B. Single Wire Earth Return Systems Single wire earth return (SWER) systems are used in many countries, such as South Africa, Australia and New Zealand to electrify sparseley populated areas. SWER systems are generally an economic choice for feeders with load densities lower than 0.5 kVA/km. However, the system has several inherent disadvantages such as: the imbalance of load currents on the primary distribution line, inability to provide three phase voltages at point of use and voltage regulation. The charging currents of a SWER system can be as high as 0.025 A/km for a 12.7 kV system [12]. This high reactive current component coupled with a relatively small load current results in a leading power factor. As a result, the load voltage is very dependent on the load current. To provide better voltage regulation, SWER distribution systems are often tted with a voltage regulator. The SST concept will benet distributions systems utilising SWER technology in that: 1) The SST can deliver rated voltage for a wide range of input voltages, resulting in a good overall system load regulation. 2) Using an SST to branch from the main feeder the SWER system would reect a balanced three phase load to the utility, unlike the single phase load reection in traditional systems. 3) Using a SST system with an available MV DC link, the power could be transmitted in DC. This not only eliminates the charging current and associated losses but also increases the power capability of the line for a certain conductor size and insulation capability. C. Protection As part of normal operation the different controllers of the SST measure all the relevant operational parameters, such as the input and output current and voltage. Using this information the controllers have control over parameters such as the output voltage magnitude. Through the implicit control of the SST the concept will bring the following advantages to the distribution protection system: 1) If one of the outputs experience a line to ground short or another fault, the inverter can continue to supply power to the other two phases. The SST can relay information of the fault to the utility should maintenance be necessary. The inverter can also monitor the phase through active impedance measurement techniques to

determine if the fault is still present and can reinstate power delivery to the faulty phase should the fault clear. 2) The SST acts as a barrier to fault currents. Even in the event of a full three phase fault on the LV side, the fault will not propagate back to the MV side. The network would not see a short circuit, as the SST simply decreases the output voltage to limit the fault current on the LV side, and hence limits the power ow through the SST. The network will however be informed of the fault through the communication interface. IV. C OMMUNICATION The power electronic system and associated control system necessitates the measurement of most currents and voltages in the system. Having a controller with measuring capabilities in the network brings many advantages to the network, tting into the idea of a smart grid [13]. Some of the benets of the communication interface are: 1) The SST can function under most secondary side fault conditions as well as certain primary side fault conditions, such as voltage unbalance or voltage sags. Information on these events can be made available to the utility for remedial action. 2) Certain outputs of the SST could be remotely controlled. For example, with the perceived rise in plug-in electric vehicles a dedicated DC output could be made available from the SST for charging purposes. To atten the load prole, this output might be switched according to actual conditions on the grid. 3) All load and other data could be made available in real time to the utility. As computational power increases utilities will be able to use such data for preventative maintenance. 4) If the SST is equipped with sufcient secondary side energy storage, the power ow through the SST could be controlled remotely in order to reduce the load in certain areas temporarily. V. P OSSIBLE FUTURE BENEFITS OF THE SST The SST concept could also offer a number of options to the distribution grid. A. Integration with other systems The LV DC link in the SST topology provides a good and readily accessible integration point for renewable energy systems into the distribution grid. A unidirectional converter could be used when the load demand is much bigger than the renewable energy generation capabilities. Where the peak generation capabilities exceed the load demand during certain periods, the excess power could be fed back into the grid by using a bidirectional converter. However, the DC link also provides an opportunity to integrate of energy storage elements such as ywheels or batteries. In such a system the SST would only provide power during periods of high demand. The complete controllability of the SST negates the use of

complex renewable energy controllers and systems to prevent island forming. The low voltage AC can also be generated at any specic frequency and phase displacement. This allows the system to generate, for example, 60 Hz from a 50 Hz host system, or to interconnect systems with different phase displacements. B. DC as a Means of Power Delivery The SST concept is ideally suited to extend the use of DC, both in MV and LV applications. The difculty in interrupting a DC feeder under fault conditions is often cited as a major hurdle in the acceptance of DC distribution in MV applications. The use of the power electronic interface (SST) to generate the DC is a means of controlling the system and interrupting fault currents. If the SST topology provides a stable MV DC bus, the SST can be separated to provide for DC distribution lines. A two wire DC line can transmit the same power as a corresponding three wire 3 phase line with the same construction. Although the distribution distances of a normal distribution feeder is relatively short, the typical benets of DC power transmission are still applicable. DC power can also be made available on the LV side. Most modern household appliances make use of DC internally and would be able to operate from a 220 V DC supply. However, it is difcult to conceive that the future residential supply will change from AC to DC. An additional DC supply might however be benecial for future, dedicated, loads such the afore mentioned plug-in electric vehicle. VI. BARRIERS TO UNIVERSAL ACCEPTANCE The power system community are used to the unique characteristics of the LFT. Many standards and preferred practices have been developed and accepted to allow for a robust and reliable system. Although the negative effects of the LFT on a distribution system are well known, the lack of any other option in the past has forced the community to embrace the LFT with both inherent strengths and weaknesses. A direct comparison of the SST and LFT in terms of SST benets previously discussed shows that the two systems are like the proverbial chalk and cheese, supercially alike but fundamentally different. The strengths of the SST are the weaknesses of the LFT. Likewise, the strength of the LFT is the weakness of the SST. The following are highlighted: A. Cost Despite the recent rally in the resource prices, the cost of the LFT is much less than the cost of a comparable SST. The SST cost is mainly driven by semiconductor cost. The cost breakdown between, rstly, semiconductors and and, secondly, other elements such as heat sinks, PC boards and passive components, would roughly be equal. Investigation of semiconductors on a price-to-powercapacity basis has shown that the price of semiconductors have roughly followed an adaption of Moores law. Conversely the price of resources has rallied over the last decade. Neglecting

the unknown effects of the current nancial crisis, every expectation would be for this trend to continue. The cost of the control circuitry, in comparison with the power components, is large. Further development of the concept could result in the development and use of ASICs (Application Specic Integrated Circuits) reducing the cost of the control circuitry. Although the rst generation SST prototypes will be very costly to manufacture, it is reasonable to expect that the costs will follow a downward trend in future. Not only will the net cost of power semiconductors decrease and the cost of control circuitry decrease through integration, but continued industrialisation will force the manufacturing costs down. In this respect the modular SST topology [8], [9] has an additional advantage in that the modular cells are very general and could be used for many different modular converters and could be manufactured in large quantities. B. Reliability LFT transformers are in general very reliable. Most applications see the LFT in use exposed to the elements and operational in very harsh environments without requiring periodic maintenance. Given the complex nature of the SST, the system is unlikely to be as reliable as the LFT. However, the total system reliability can be improved signicantly. The modular nature of the proposed topology [8] allows for the isolation (or bypass) of a faulty module [14]. The system could be designed to include redundant cells, whereby the voltage stress over each module is reduced such that, in the event of a fault, the faulty cell could be bypassed and the remaining cells would be able to handle the additional voltage stress. Although the inclusion of redundant cells would allow for continued operation of the SST in the event of a cell failure, the system would still require maintenance in that the faulty cell would have to be replaced. As with all systems, the reliability will increase as the technology matures. The inherent measurements of the SST system allow for the collection of health assessment information, such as capacitor ripple current and heat sink temperature. This information can not only be used to estimate and predict component failures for preventative maintenance purposes but will also aid in improving the design, hence increasing system reliability. With increased integration, especially of the control system through the use of ASICs, the component count and therefore the failure rate of the system will be reduced. With each iteration of the system design, the reliability will increase. As the system matures, the cooling system, for example, could be changed. The current demonstrator uses mechanical fans and forced air cooling, however, once the electronic design questions are addressed and the system topology matures, a dedicated water cooled system could be implemented, hence improving system reliability. Depending on the power rating and additional capabilities of the SST, such as harmonic isolation, the energy storage requirements can be reduced to a level where electrolytic

capacitors can be replaced by lm capacitors. For example, 4.5 mF, 1 100 V capacitors are currently available with life times in excess of 100 000 hours [15] (compared to about 12 000 hours for an electrolytic capacitor). This 4.5 mF capacitor will be more than adequate for a 80 kVA SST with power quality capabilities, as discussed in III-A. C. Efciency The efciency of a 80 kVA, 11 kV / 400 V LFT is in the region of 97 % (according to [2] the efciency for a 30 kVA LFT must be > 97.3 % to qualify for NEMA class 1 efciency level rating), while the efciency of a corresponding SST will be somewhat lower. It is not clear what the efciency of the mature SST concept will be, for systems reported in litterature efciencies vary from about 90 % [6] to up to 98.1 %, without the output inverter [16]. However, a straight comparison of the efciencies is slightly misleading. The losses in a LFT can be separated into two broad categories, namely core and winding losses. The nature of the losses are such that core losses are a function of the applied voltage on the primary winding while winding losses are determined by the power delivered. Therefore the core losses are often termed standing losses. The point of maximum efciency of the LFT is roughly where the core losses and the winding losses are equal. Often the LFT is designed for maximum efciency at full load, although according to new standard practices the point of maximum efciency should be at 50 % [2]. Although it is clear that the LFT is more efcient, several other factors must be considered, most notably the effect of non-linear loads. As stated earlier, the average residential load injects a large proportion of current harmonics into the distribution transformer [1]. Although no specied guide or test method exists to characterise the efciency of distribution transformers under non-linear load conditions, it is well known that both current and voltage harmonics adversely affect the efciency [17]. With the reduced transmission path length of the SST and the fact that the harmonic currents do not propagate to the primary side, the effect of non-linear loads on SST efciency should be minimal. Although small, the inefciencies of additional circuit components such as current measuring devices present in a LFT circuit, which are redundant in a SST circuit with inherent measurements will also affect the total circuit efciency. VII. C ONCLUSION Although the LFT is in global use, many of its characteristics are unwanted in the modern distribution grid. This article discussed the benets the SST concept could bring to the grid, most notably in terms of power quality. The SST concept also brings control and communication advantages to the grid. Although recent advancements in power semiconductor technology allow for renewed interest in the SST concept, and several prototypes reported in literature, it is difcult to foresee the SST as a direct replacement of the LFT with current technologies. In this article some of the drawbacks of

the SST were highlighted. However, with advances in power semiconductors and with each prototype the SST concept will mature. Given the increased capabilities of a distribution system equipped with a SST, it is possible that the SST concept might become economically viable in the near future. This statement can be strengthened if the cost of additional power systems components, such as measuring equipment and voltage regulators that can be replaced by the SST, are factored into the equation. R EFERENCES
[1] M. de Lima Tostes, U. Bezerra, R. Silva, J. Valente, C. de Moura, and T. Branco, Impacts over the distribution grid from the adoption of distributed harmonic lters on low voltage customers, IEEE Transactions on Power Delivery, vol. 20, no. 1, January 2005. [2] Guide for determining energy efciency for distribution transformers, NEMA Standards Publication TP 1-2002. [3] J. Brooks, Solid state transformer concept development, Naval Material Command, Civil Engineering Laboratory, Naval Construction Battalion Center, Port Hueneme, CA, Tech. Rep., 1980. [4] J. Lai, A. Maitra, and F. Goodman, Performance of a distribution intelligent universal transformer under source and load disturbances, in Conference Record of the 2006 IEEE 41st IAS Annual Meeting, vol. 2, 8-12 October 2006. [5] L. Heinemann and G. Mauthe, The universal power electronics based universal transformer, an unied approach, in Conference Record of the 2001 IEEE 32th Power Electronics Specialists Conference, 18-22 June 2001, pp. 504 509. [6] E. Ronan, S. Sudhoff, S. Glover, and D. Galloway, A power-electronic based distribution transformer, IEEE Transactions on Power Delivery, vol. 17, no. 2, pp. 537543, Apr. 2002. [7] J.-S. Lai and F. Z. Peng, Multilevel converters - a new breed of power converters, IEEE Transactions on Industry Applications, vol. 32, pp. 509517, May / June 1996. [8] J. van der Merwe and H. d. T. Mouton, Solid-state transformer topology selection, in Conference Record of the IEEE International Conference on Industrial Technology, 10 - 13 February 2009. [9] , Natural balancing of the two-cell back-to-back multilevel converter with specic application to the solid-state transformer concept, in Conference Record of the 4th IEEE Conference on Industrial Electronics and Applications, 25 - 27 May 2009. [10] L. Heinemann, An actively cooled high power, high frequency transformer with high insulation capability, in Applied Power Electronics Conference and Exposition, 2002. APEC 2002. Seventeenth Annual IEEE, vol. 1, 2002, pp. 352357 vol.1. [11] H. Akagi, The next-generation medium-voltage power conversion systems, Journal of the Chinese Institute of Engineers, vol. 30, no. 7, pp. 11171135, 2007. [12] N. Chapman, When one wire is enough, Transmission and Distribution World Magazine, April 2001. [13] A. Ipakchi and F. Albuyeh, Grid of the future, IEEE Power and Energy Magazine, March / April 2009. [14] E. Choudhary, V. Ledezma, R. Ayyanar, and R. Button, Fault tolerant circuit topology and control method for input-series and ouput-parallel modular DC-DC converters, IEEE Transactions on Power Electronics, vol. 23, no. 1, pp. 402411, 2008. [15] EPCOS, Film capacitors - power electronic capacitors, pcc series, Datasheet B25655, 2008. [16] L. Heinemann, Analysis and design of a modular, high power converter with high efciency for electrical power distribution systems, Power Electronics Specialists Conference, 2002. IEEE 33rd Annual, vol. 2, pp. 713718 vol.2, 2002. [17] A. Damnjanovic and G. Ferguson, The measurement and evaluation of distribution transformer losses under nonlinear loading, Power Engineering Society General Meeting, 2004. IEEE, pp. 14161419 Vol.2, June 2004.