Page 1 of 7 2017-METC-0708

Directly Feeding Large DC Loads Through PV Power Plants: a New Concept

Applied to Industry

Flávio A. G. Rodrigo A. M. Matheus P. Porto Braz J. Cardoso

Lacerda Gomes Departamento de Engenharia Member, IEEE
Student Member, IEEE Programa de Pós-Graduação Mecânica Departamento de Engenharia
Programa de Pós-Graduação em Engenharia Mecânica matheuspporto@gmail.com Elétrica
em Engenharia Elétrica ramg.gomes@gmail.com braz.cardoso@ieee.org
falavio@gmail.com

Universidade Federal de Minas Gerais

Av. Pres. Antônio Carlos, 6627 - Pampulha, Belo Horizonte - MG, 31270-901, Brazil

predominantly in ac, the consumption of energy in its dc

form has grown in the last decades, due to the popularization
Abstract — The energy generated by PV sources is inherently of electronic devices such as computers, mobile phones, TVs
produced in direct current, but since the electrical distribution and other household appliances [2]. In the industrial
systems are mostly built on alternating current, the commercial environment, particularly, there are a number of processes
solutions of converters for solar applications are based on the
dc/ac conversion. This paper proposes that in industrial that demand power in dc and in large scale. Electrolysis, for
applications where the power feed into the process is in dc form instance, a process used in the production of metals, usually
and there is a desire of using a PV system to contribute in the requires high dc currents, which is obtained by the use of
energy production, a dc/dc converter should replace the transformers associated with ac/dc converters, namely
inverter, allowing the generated energy to be injected directly in "rectifiers".
the process, thus reducing drastically the losses associated. In Therefore, an industrial plant that performs electrolysis
order to illustrate this concept, a case study is developed to
establish a project methodology which will reveal the key and plans to build a PV system to reduce power consumption
aspects and constraints. The PV system is designed according to from grid during solar irradiance times would invariably have
the available area, and the voltage and current levels are set to make its solar generation connected to the ac grid, due to
taking into account distances and insulation limits on the solutions available in the market. In other words, the
semiconductors. Converters should be designed to perform energy would be generated in dc in the solar panels,
maximum power point tracking (MPPT) and to balance voltage converted to ac by the inverter, transmitted through the grid
levels between the PV system and the process. Simulation results
demonstrates the effectiveness of the designed solution. The to the transformer, and converted again to dc by the rectifier.
process electrical system must be able to sense and react to the Knowing that the inverter, the transformer and the rectifier
current injected by the PV system, in order to guarantee that are all non-ideal devices and have electrical losses, it comes
the load current is kept stable even during quick changes in to the conclusion that a significant part of the generated
irradiance conditions. The case study shows that the proposed energy is promptly loss in these elements, due to the dc/ac/dc
solution is technically feasible. transformations. Fig. 1 illustrates this conventional system,
where the arrows indicates the power flow direction.
Index Terms — Photovoltaic (PV) systems, dc loads, dc/dc
converters, industry applications, electrolysis This paper proposes that, in such cases, a dc/dc converter
should replace the inverter, allowing the generated energy to
be injected directly in the process and reducing drastically the
I. INTRODUCTION losses associated. The loss reduction will come from three
main aspects. First, the generated power will be subject to
E NVIRONMENTAL concerns have attracted a lot of
attention in recent years to renewable energy sources,
among which photovoltaics (PV) stands as one of the most
less conversion devices, performing no ac transformation.
Second, the dc/dc converter will inherently have a higher
efficiency than the inverter of same power, since the dc/dc
successful [1]. The energy generated by PV sources is
converter is typically a first conversion stage on the complete
inherently produced in direct current, but since the electrical
commercial inverter for solar applications. And third, as the
distribution systems all around the world are mostly built on
process will be able to sense the injection of current from the
alternating current, the commercial solutions of converters
solar source, it will reduce the current drained from the ac
for solar applications are based on the dc/ac conversion, the
grid, decreasing losses associated in the duo transformer-
so called "inverters".
rectifier. Fig. 2 illustrates the proposed solution.
However, although the distribution systems are
To demonstrate how this solution can be more power

978-1-5090-4894-6/17/$31.00 © 2017 IEEE

 2017-METC-0708 Page 2 of 7

Fig. 1. Conventional system topology of a PV plant feeding a dc industrial Fig. 2. Proposed solution of a PV plant directly feeding an industrial process
process. with dc.

efficient in comparison to the commercial option available, The inverter is replaced by a dc/dc converter with higher
consider Fig. 3 (a) where an example of industrial system is efficiency, and the generated power is directly feed into the
shown. The process load demands the equivalent to a 100% dc process. The total demand from grid is 75.8%, which
power, and it is fed by a grid-connected transformer and a means that the total losses in the system are 5.8%. In addition
rectifier. The efficiency of such equipment can vary to a decrease in demand from the grid, there was a reduction
depending on the rated power and the operation conditions, of 35% of total losses in comparison to the typical
but for the sake of this example, it will be considered commercial solution.
respectively 98% and 95%, which are typical values for these Although just illustrative and with arbitrary values, the
devices. Knowing that the PV system is not connected, then example stated shows that the proposed solution can promote
the total power demanded from the grid is 107.4%, which a significant benefit in terms of total system efficiency. It is
means that a total of 7.4% of power is lost due to the worth saying that if the efficiencies considered or the total
transformer and the rectifier (conduction losses neglected). PV system production were different values, then the
Consider now the situation illustrated in Fig. 3 (b), where a reduction of 35% in losses could be a different value.
PV system with instant power generation of equivalent 30% However, the total losses would still be lesser in the proposed
is connected to the grid through a 95% efficiency inverter solution, due to the use of lesser conversion devices.
(typical commercial solution). In this situation, the power Other studies in the literature had suggested similar
generated by the solar plant is fed to the process in ac, solutions for this problem, but in different contexts. Some
reducing the demand from the grid to 78.9%. Given that the articles have mentioned the use of PV energy to feed loads
generated power is 30% and the load consumption is 100%, directly in dc, but as a variation of a particular converter
then there was 8.9% of total losses. Despite the demand from topology or in a microgrid [2], [3], [4], [5], [6]. Others
the grid had largely decreased with the PV generation, the suggested the feeding of dc loads directly from PV systems in
total losses have actually increased, due to the addition of the standalone applications [7], and some even considered small
inverter. scale electrolysis processes [8], [9].
Then Fig. 3 (c) shows the solution proposed in this paper. As per the present paper, what is being proposed is an

(a) (b) (c)

Fig. 3. Grid power consumption: (a) PV system not connected, (b) PV system connected to the AC grid and (c) PV system directly connected to the DC load.

978-1-5090-4894-6/17/$31.00 © 2017 IEEE

Page 3 of 7 2017-METC-0708

alternative use of a well-known converter topology, in a real have large rooftop areas and underused land, such as inactive
industry application, and to feed the process in a mining areas or tailings dams, which can be employed for the
complementary way with the grid - a different approach from purpose of installing the solar panels.
all the papers referenced. The goal is to identify the benefits
and to introduce a concept that is not commercially explored III. PV SYSTEM DESIGN
yet, but have a large potential. A case study will be used to The first step on designing the PV system is the solar panel
develop the solution and to reveal the restrictions involved, selection. For this project, it will be considered a typical
and a conceptual project will be presented as a methodology polycrystalline silicon panel, easily found in the market [10].
to be followed by other large-scale real projects. Table I shows the main characteristics of this type of panel.
This work has the support of Votorantim Metais, a metal The next step consists on determining the total capacity of
company that is investigating the feasibility of such systems the PV plant, therefore, the amount of solar panels that could
in order to apply its concepts in a real electrolysis plant. be installed in the available area. In contrast with rooftop
installations, where almost all the space available is used
II. CASE STUDY directly to install panels, in large areas where the panels must
In order to assess the proposed concept, a case study will be installed on the ground a significant part must be let
be carried out. A fictitious industrial plant will be considered unused. This is due to the need of space for circulation of
and, although it is not real, the parameters were chosen to maintenance personnel and vehicles, and also to guarantee
represent a good approximation of the values found on real that the panel sets will not shade each other depending on the
operating plants. The same decisions and restrictions taken irradiation angle. Only a detailed project that takes care of
care in this case study can be used later on a real these specific restrictions of the site could then determine the
implementation, therefore it can be seen as a project exact proportion of the total area that could be used directly
methodology. for installing panels.
The industrial plant has an electrolysis process and its In order to simplify this analysis, some inferences were
electrical system consists of a grid-connected transformer made over an NREL report regarding land-use requirements
feeding a thyristor rectifier. The output current of the rectifier for solar energy in the United States [11]. The objective was
can be regulated by controlling the gate firing angle, and its to find a typical ratio of the area effectively occupied by
value is determined by the process in order to achieve the panels, over the total PV plant area. Although the report does
best performance of the metal production. Likewise, the not mention this factor directly, after a few assumptions, it
output voltage of the rectifier is determined by the was concluded that it lies in the range from 31% to 58%. For
electrolysis cell voltage, which varies depending on the cell the purpose of this case study, it will be assumed that the
area and the material being produced, although it remains effective PV panels area correspond to 50% of the total area,
approximately stable during regular operation. The rated a ratio that seems reasonable in comparison to the data
output values are then selected as 400 Vdc and 50 kA, available in the report and also with known real PV projects.
summing up to a rated power of 20 MW. The panels presented on Table1 have the dimensions of
It will be considered also that the PV power will be 0.99 m x 1.64 m (1.62 m²). Considering that they will be
injected upstream of the current measuring point, allowing installed with a tilt angle of 20º, an assumed value for the
the rectifier to sense when the solar energy is being produced. fictitious plant latitude, it comes to a total of 16,386 panels,
By doing so, the rectifier can control the current drained from or 4.096 MW of total peak DC power capacity.
grid to reduce its value and equalize the load current in a The next decision to be made concerns on how the panels
fixed value throughout the operating cycle. It should be will be associated in sets of series and parallels. The series
mentioned that in real existing plants, where the connection connection determines the operation voltage levels, which
of a PV system upstream of the measuring point may not be from one point of view must be as high as possible, to
possible due to space restrictions, an alternative solution is to decrease conduction losses on the long cables inherent to
build a data network to inform the rectifier control system of
the solar-generated current value, allowing the rectifier to TABLE I
MAIN ELECTRICAL CHARACTERISTICS OF THE PV PANEL
actuate reducing its drained current as well. There will be no
need of a battery bank since all the power will be instantly Description Symbol Quantity Unit
used in the process. Maximum power Pmax 250 W
Lastly, it is assumed that the industrial plant has an area of Open-circuit voltage Voc 37.6 V
50,000 m² available to be used for the installation of the new Short-circuit current Isc 8.92 A
Voltage at Pmax Vmp 29.8 V
PV system. This assumption is likely to occur in real world Current at Pmax Imp 8.39 A
plants due to the fact that companies that performs Efficiency PCE 15.4 %
electrolysis (generally Mining and Metals companies) often

978-1-5090-4894-6/17/$31.00 © 2017 IEEE

 2017-METC-0708 Page 4 of 7

such applications. In the other hand, from the converter point

of view, voltages too high may cause the need of special and
more expensive semiconductor devices. The decision then is
to set the maximum voltage of the panel strings in 1,000 V,
allowing the use of regular semiconductors and at the same
time have relatively low conduction losses.
In order to do so, each string will be composed by 27
series-connected panels, which comes to a total open-circuit
voltage Voc of 1,015.2 V and a rated voltage at maximum
power point Vmp of 804.6 V. Considering the total number of
panels, the plant shall have 606 parallel strings with a total Fig. 4. Clusters division and solar panels association.
rated current at maximum power point I mp of 5.09 kA.
At this point, it is worth to highlight that conventional Due to the pulsating property of the input current on Buck
large-scale PV plants often uses decentralized inverters. By converters, a capacitor must be inserted before the switch to
doing so, it is possible to convert the power of portions of the assure the output current of the PV panels is continuous. On
total plant to ac separately, and use transformers to rise the the other hand, since the converter will work as a current
voltage levels and transmit the power with less conduction source and the output voltage is regulated by the process
losses to the connection point. As per the solution proposed itself, there is no need of an output filter capacitor.
in this paper, it must be remembered that the use of dc/ac Additionally to this topology definition, it will be used three
conversion and transformers is avoided, then the problem of parallel Buck converters for each cluster, with the objective
conduction losses must be addressed differently. of making each converter to handle an even smaller power
This application presumes industrial use and large areas, so and current than the total cluster. The three Buck converters
the transmission of currents of kilo amperes, even in voltages will be built over a single six pack IGBT module, to ensure
near 1 kV, could result in high losses. Additionally, since the best balance among them, and will be controlled using and
process has a lower voltage than the PV plant, it is preferable interleaving configuration. The proposed solution for the
to locate the converter as near as possible to the process. All converter and its topology is shown in Fig. 5, where it can be
that led to the decision of dividing the total area into smaller seen that the injection point is upstream of the thyristor
areas, or "clusters", that will be handled by individually rectifier measuring point.
separated sets of cables and converters, hence reducing B. Control Strategy
currents and losses. For this specific project, it will be The selected topology led to a control strategy where the
considered the division in 20 clusters, which allows the controlled variable is the input capacitor voltage, and the
farther one to have a maximum conduction loss on copper of MPPT provides the reference for a closed-loop feedback
less than 2% (considering the use of 240 mm² cables and the proportional-integral controller using pulse width modulation
area approximately squared with 200 m x 250 m). (PWM). The interleaved configuration means that each of the
Given this division, the PV plant will have 20 clusters, three Buck converters will be controlled in equally
each composed by 30 parallel strings of 27 series-connected distributed phase angle of the PWM carrier, considering the
panels. The cluster's output rated values are then 804.6 V, switching period [12].
251.7 A and 202.5 kW. The corrected total number of panels There are a number of different MPPT techniques in the
is 16,200, with a power capacity of 4.05 MW. Fig. 4 literature that can be more suitable for one or another
illustrates the division in clusters and the panel associations. application. For this project, the technique selected was
Perturb and Observe (P&O), which has shown a good
IV. CONVERTER DESIGN performance on experimental tests against other algorithms
A. Topology Selection [13]. As per the PI controller, its parameters were calculated
A converter is needed between the PV plant and the process using state-space averaging [14], tuning to a pass bandwidth
for two main reasons: make voltages in both sides compatible of 1/20 of the switching frequency (defined as 5 kHz).
and perform the maximum power point tracking (MPPT). As C. Simulation Results
informed before, the process voltage is set to 400 V, and the
To illustrate the system operation, a simulation was set
PV plant have a rated voltage of 804.6. It means a step-down
using MATLAB Simulink. In order to reduce computational
topology must be used, and since the rated duty ratio is nearly
effort, instead of simulating the entire PV plant and its 60
0.5, there is no need to use a transformer-isolated option. In
converters, it was considered only one Buck converter, which
this case, the more suitable alternative is the Buck converter,
handles one third of the power of one cluster (or 1/60 of the
since it requires less components and therefore have lower
whole plant, 67.5 kW). To compare the energy production of
losses and cost.

978-1-5090-4894-6/17/$31.00 © 2017 IEEE

Page 5 of 7 2017-METC-0708

Fig. 5. Converter topology and connection to process.

a single Buck, the process current was also reduced to 1/60 of decreases the current drained from grid when it senses that
the total rated 50 kA (833.33 A). Fig. 6 illustrate the results the total load current is exceeding its maximum value.
obtained, where the temperature was set fixed in 25ºC.
In the upper graphic of Fig. 6 it can be seen that as the V. LOAD CURRENT STABILITY
irradiance varies throughout the day, the power output tends One main aspect regarding the application suggested in
to accommodate in the maximum value possible. In this paper is the stability of the load current. As briefly
particular, when the irradiance is 1,000 W/m² (nominal described before, such processes need the load current to be
condition), then the power output sets in approximately 67.5 as stable and accurate in the target value as possible, since
kW, the expected rated power in dc side. This shows the this will affect the productivity and the quality of the material
implemented MPPT algorithm is working accordingly. It is being produced. Hence, the inherent variance in output
worth to mention that for the simulation the converters and current from the PV plant must be accommodated by the
cables were considered lossless, as the simulation had the regular current control, originally designed for the process (e.
purpose of verifying the solution technically and not to make g. in Fig. 5, the control performed by the thyristor rectifier
an efficiency assessment. together with the current measurement).
As per the lower graphic in Fig. 6, it shows that the load This variation is due to the fact that the MPPT technique
current is keep steady even when the irradiance and the PV forces the PV system output to follow as closely as possible
current raises. This is due to the rectifier current control, that the irradiance, seeking to extract the maximum power of a
given condition. That combined with a windy and cloudy
weather can provoke high di/dt. As an example, Fig. 7 shows
an experimental measurement of a PV array of 8.11 A rated
current during one day of operation (out of a set of hundreds
of days). The highlighted line shows the highest di/dt
registered for that day, a 6.5 A rise during 5 minutes, which
represents about 80% of the rated current. Although this data
only represents a particular event for a single PV plant, it
shows that the PV output can have a substantially high di/dt
in some extreme conditions.
On the other hand, such industrial processes will have a
current control system with a limited di/dt, usually due to the
technology employed in the converter. This must prevent the
PV plant to achieve any power capacity desired, since the
effects of the weather condition cannot overtake the
maximum di/dt tolerated by the process. In order to illustrate
this limitation, Fig. 8 shows the behavior of an example
Fig. 6. Simulation results: (a) Irradiance and PV output and (b) current from electrolysis plant that is commanded to reduce its output
the converter, from the grid and applied to the DC load, during a irradiance current. The control system limits the decay ratio, reaching
variation. its target value after about some time.

978-1-5090-4894-6/17/$31.00 © 2017 IEEE

 2017-METC-0708 Page 6 of 7

Fig. 7. Output current of an experimental PV array during a day.

maximum current variation observed in the PV plant must be
matched by the current variation from the process control.
This assures that the process will perform as if it was not fed
in parallel by a PV plant.
It is expected that this paper can encourage industry
owners and manufacturers to develop such systems in real
applications, being able to contribute with the growth of
renewable energy production and increase energy efficiency.
For instance, Votorantim Metais, supporter of this work, will
proceed with a pilot of this application in the near future.

ACKNOWLEDGEMENTS
Fig. 8. Example of the real current output on an electrolysis process after a
change in the desired control setpoint by the operator. The authors would like to thank the financial support from
These two limitations must then be used to calculate the the Brazilian federal government agencies CAPES and
CNPq, the Minas Gerais state government agency FAPEMIG
maximum power capacity of a PV plant directly feeding a dc
and the partner research company Votorantim Metais
industrial load. The designer of the system must investigate
(research grant no. 24985).
the process where the solar power will be injected in order to
make sure that the power variations due to weather
REFERENCES
conditions can be met by the current regulator, even if in
some cases this means the PV plant must have its capacity
[1] Kristin Seyboth, F. Sverrisson, F. Appavou, A. Brown, B. Epp, A.
decreased. Leidreiter, C. Lins, E. Musolino, H. E. Murdock, K. Petrichenko, T.
C. Farrell, T. T. Krader, A. Tsakiris, J. L. Sawin, J. Skeen, and B.
VI. CONCLUSION Sovacool, Renewables 2016 Global Status Report. 2016.
[2] H. Elfeqy, M. Shahin, A. Al-Rumaihi, A. Massoud, and A. Gastli, “A
This paper presents an alternative approach in terms of the highly efficient PV power system for DC MicroGrids,” in 2016 IEEE
use of PV systems to contribute in power generation for Symposium on Computer Applications & Industrial Electronics
(ISCAIE), 2016, pp. 183–188.
industrial plants. Particularly, for processes fed in dc, it was [3] N. Zhang, D. Sutanto, and K. M. Muttaqi, “A four-port DC-DC
shown that the proposed technology can promote a converter to integrate energy storage system and solar PV to supply
significant reduction in general losses, making such projects the grid and local load demand,” in 2015 Australasian Universities
Power Engineering Conference (AUPEC), 2015, pp. 1–6.
more economically feasible in comparison to the traditional [4] D. Debnath and K. Chatterjee, “A transformerless grid connected
and commercially available technology. This is especially inverter for solar photovoltaic systems having capability to negotiate
interesting for Metals and Mining companies, since they are DC loads,” in 2015 IEEE International Conference on Industrial
Technology (ICIT), 2015, pp. 2835–2840.
usually the biggest consumers of dc in the industry. [5] Aditya C, Sundararaman K, and Gopalakrishnan M, “Multiple-
The concept proposed was then used in a case study, where integrated converters for grid-connected PV systems with intermediate
its many aspects could be explored. The methodology and DC loads,” in 2016 2nd International Conference on Advances in
Electrical, Electronics, Information, Communication and Bio-
restrictions developed indicates that such projects are Informatics (AEEICB), 2016, pp. 544–548.
technically viable, even though some considerations may be [6] S. N. David, S. P. Santosh, and L. S. Narayanan, “Control Strategy for
different from regular PV plants. Simulation results Power Flow Management in a PV System Supplying DC Loads,” in
demonstrated that the procedure established can be followed 2014 International Conference on Power Signals Control and
Computations (EPSCICON), 2014, vol. 60, no. 8, pp. 1–5.
in real industrial plants. [7] U. Shah, P. Kadam, and G. Lahoti, “Control of standalone solar PV
Finally, one key aspect was addressed in order to system in varying operating conditions for DC loads,” in 2016 IEEE
guarantee the proper operation of the conjugated system: the

978-1-5090-4894-6/17/$31.00 © 2017 IEEE

Page 7 of 7 2017-METC-0708

International Conference on Engineering and Technology

(ICETECH), 2016, no. March, pp. 1181–1186.
[8] S. Dahbi, A. Aziz, N. Benazzi, and M. Elhafyani, “Optimised
hydrogen production by a photovoltaic - Electrolysis system DC/DC
converter and water-flow controller,” in 2015 3rd International
Renewable and Sustainable Energy Conference (IRSEC), 2015, pp. 1–
6.
[9] J. Appelbaum, K. Khouzam, and Y. Dagan, “Photovoltaic Systems for
Electrolysis Using a Maximum Power Point Tracker,” in 2006 IEEE
24th Convention of Electrical & Electronics Engineers in Israel, 2006,
pp. 21–24.
[10] Datasheet solar panel YL250P-29b Yingli Solar.
[11] S. Ong, C. Campbell, P. Denholm, R. Margolis, and G. Heath, “Land-
Use Requirements for Solar Power Plants in the United States,”
Nrel/Tp-6a20-56290, no. June, p. 47, 2013.
[12] E. T. Kangappadan and D. David, “Interleaved buck converter with
continuous supply current using OCC technique,” in 2016
International Conference on Electrical, Electronics, and Optimization
Techniques (ICEEOT), 2016, pp. 320–325.
[13] M. Aghaei, A. Dolara, F. Grimaccia, S. Leva, D. Kania, and J.
Borkowski, “Experimental comparison of MPPT methods for PV
systems under dynamic partial shading conditions,” in 2016 IEEE
16th International Conference on Environment and Electrical
Engineering (EEEIC), 2016, pp. 1–6.
[14] N. Mohan, T. M. Undeland, and W. P. Robbins, Power Electronics:
Converters, Applications and Design, 2nd ed. John Wiley & Sons,
1995.

978-1-5090-4894-6/17/$31.00 © 2017 IEEE