GreenDataNet
D1.6 - Energy Storage and PV
Architecture design
Final
Eaton Electrical, Power Quality & Electronics Division (PQED)
TABLE OF CONTENTS
TABLE OF CONTENTS ........................................................................................................................................ 2
REVISION SHEET ............................................................................................................................................... 4
2.
3.
INTRODUCTION ....................................................................................................................................... 5
2.1
Purpose of the document ...................................................................................................................... 5
2.2
Document overview .............................................................................................................................. 5
GREENDATANET UPS SYSTEM REQUIREMENTS ....................................................................................... 6
3.1
4.
3.1.1
Grid Operation .............................................................................................................................. 7
3.1.2
Stored Energy Mode ..................................................................................................................... 7
3.1.3
PV Operation ................................................................................................................................. 7
3.1.4
Load Sharing.................................................................................................................................. 7
3.1.5
Load Peak Shaving......................................................................................................................... 8
3.1.6
Backfeed to Utility......................................................................................................................... 8
GREENDATANET UPS SYSTEM TOPOLOGY SELECTION ............................................................................. 9
4.1
PV integration in Data Centre Power Architecture................................................................................ 9
4.2
Alternative system topologies ............................................................................................................. 10
4.2.1
PV Source with a Grid-connected inverter ................................................................................. 10
4.2.2
PV Source with a DC/DC-converter ............................................................................................. 11
4.2.3
PV Array Connected directly to the DC-bus of A UPS ................................................................. 12
4.3
5.
Use Cases ............................................................................................................................................... 7
Overview for the selected topology .................................................................................................... 13
INTERFACE WITH PV SOURCE ................................................................................................................ 15
5.1
PV field specifications for the PV integration ...................................................................................... 15
5.1.1
Review of PV field topologies ..................................................................................................... 15
5.1.2
Selected topology for GDN case ................................................................................................. 16
5.1.3
DC-DC converters for PV – UPS interface ................................................................................... 17
5.1.4
MPPT Control .............................................................................................................................. 18
5.2
Optimization of the PV – UPS interface ............................................................................................... 18
5.2.1
Optimization objectives definition .............................................................................................. 20
6.
5.2.2
Optimization parameters ............................................................................................................ 21
5.2.3
Interface modeling for optimization procedure ......................................................................... 22
5.2.1
Optimization results .................................................................................................................... 24
INTERFACE WITH THE NISSAN BATTERIES ............................................................................................. 27
6.1
Outline of the Battery interface .......................................................................................................... 27
6.2
Nissan Battery Pack and 93PM battery converter Specifications ........................................................ 28
7.
DEMONSTRATOR .................................................................................................................................. 30
8.
REFERENCES .......................................................................................................................................... 32
REVISION SHEET
Revision
Number
Date
Brief summary of changes
Rev 0.1
30/09/2014
First concept
Rev 0.2
30/10/2014
Baseline document as presented in Conference Call
Rev 0.4
19/11/2014
Concept document
Rev 0.5
19/12/2014
Final version
Rev 0.6
30/12/2014
Integration of Optimization explanation
Rev 0.7
09/01/2015
Introduction modification and Minor changes
Rev 0.8
07/03/2015
Comments added
2.
2.1
INTRODUCTION
PURPOSE OF THE DOCUMENT
The aim of this document is to describe higher level requirements related to the UPS system for the urban data
centre environment as targeted by the GreenDataNet project. This document describes requirements for the
UPS and its interfaces for the PV energy source and EV Li-Ion battery. It also describes operational concept and
use cases of the integrated UPS system in energy efficient data centre environment.
2.2
DOCUMENT OVERVIEW
This document is organized as follows. The chapter 3 provides technical and functional requirements as well as
use cases. The chapter 4 introduces alternative UPS system topologies and presents the selected topology. The
chapter 5 focuses on the interface between PV array and the UPS system. Interface between Nissan battery
and Eaton UPS is presented in the chapter 6. The last part details the system architecture of the Demonstrator.
5
3.
GREENDATANET UPS SYSTEM REQUIREMENTS
IT power of the Data Centre is the key parameter when specifying requirements for the UPS system. Based on
definition of Urban Data Centre provided in deliverable D1.1, the GreenDataNet UPS (GDN-UPS) system for
type I Data Centre shall be specified to the power range from 10 to 40 kW. Accordingly, type 2 data centre
needs an UPS system rated from 20 to 250 kW IT load [19].
Figure 3-1 Capacity ranges of Type I and Type II Urban Data Centres (Source: ICTroom, GDN D1.1)
GreenDataNet targets to utilize up to 80% of renewable power and decrease average Power Usage
Effectiveness (PUE) must be taken into account also in GDN-UPS design. Overall efficiency of the UPS system
shall be as high as possible, typically up to 96% and even more, on double conversion operating mode. In GDNUPS there shall be also an interface for PV source. It is assumed to be located in vicinity of the data centre.
Capability to store energy into Nissan Li-Ion batteries is one of the key requirements in GreenDataNet.
Depending on the type of the data centre, requirement for battery back-up time varies typically between 5 to
30 minutes. There shall be a possibility to connect minimum four (4) Nissan batteries in parallel into one GDNUPS unit. Capacity of such a battery pack is 96 kWh, enabling load peak shaving operation during load peaks.
The GDN-UPS shall be scalable system, enabling external parallel connection of the units. Mechanical
dimensions and environmental specifications for the equipment shall follow the European DC Standards.
The GDN-UPS shall provide interface for communication with IT equipment and facility systems.
6
3.1
USE CASES
3.1.1
GRID OPERATION
In grid operation (also called as normal operating mode) the GDN-UPS supplies all power to the load from the
utility mains. The PV source is disconnected if available power is too low (low irradiation) or there is a fault in
PV array. Batteries are on the float charge or on the rest mode, depending on the battery state of charge.
Battery charger is only turned on when the batteries need to be charged.
In normal grid operation the GDN-UPS can be in following modes:
a)
Double conversion mode: the mains AC power is converted first to DC by rectifier. Then inverter
derives regulated DC power from rectifier and produces regulated and filtered three-phase AC power
to the load.
b) Energy Saver System mode: load is supported securely by utility power through the static switch with
double conversion available on-demand with typically less than 2 ms transition time.
c) Bypass mode: the load is supported directly by utility mains through the UPS static switch.
3.1.2
STORED ENERGY MODE
On the stored energy mode (battery mode) the GDN-UPS is supplying power to the load from the battery. The
UPS transfers to the stored energy mode automatically when the mains power fails or if the utility power does
not confirm to specified parameters. In the grid operation the GDN-UPS can share the battery power when
supplying the load if the mains utility input voltage is out of specification.
3.1.3
PV OPERATION
On the PV mode the GDN-UPS is operating independently without utility mains feed. Following use cases are
supported:
a) PV is supplying power to the load, battery is on the rest mode
b) PV is charging battery, load is disconnected
c) PV is supplying power to the load and charging battery
In case the GDN-UPS is shutdown PV arrays are also disconnected.
3.1.4
LOAD SHARING
The GDN-UPS is able to share the load between utility mains, PV source and the battery. The PV source has a
highest priority to provide power to the load. Following use cases are supported:
a)
Utility mains and PV source are both supplying the load
7
b) Utility mains is lost, battery and PV source are both supplying the load
c) Utility is regaining, GDN-UPS resumes utility mains and PV power share
3.1.5
LOAD PEAK SHAVING
On the load peak shaving operation the battery energy and PV source of the GDN-UPS can be used to reduce
peak demand and thus reduce the electricity cost by discharging a stored energy during load peaks.
Recharging of the battery is executed during the low demand periods, typically at night hours, but also when
daytime demand is low and solar energy is available.
Commands for peak shaving operation must be provided as an external signal to the GDN-UPS signal input or
through the remote monitoring interface.
This functionality was not planned in the initial description of work, but the team has considered that it would
be worth to implement it as the it was not time-consuming to implement while it provides added value for the
integration with the Smart Energy system that will be implemented in D3.7 – Smart Energy management
system where the overall energy control is detailed .
3.1.6
BACKFEED TO UTILITY
Power back feed to utility is executed only in a case neither the load nor the battery is able to demand all the
power supplied from the PV source.
Basically, power back feed to utility would enable data centre owner to support utility network in short peak
demand periods by discharging stored energy and feeding solar energy to the utility network. However, even if
the Rectifier and the inverter are both bi-directional, these use cases are not fully implemented to the GDNUPS.
8
4.
GREENDATANET UPS SYSTEM TOPOLOGY SELECTION
The power distribution architecture within a data centre affects also the integration of PV source and energy
storage. Following factors shall be taken into account when selecting the most feasible solution:
a)
b)
c)
d)
the targeted size of the data centre
the critical IT load to power
required back-up time for the load
the variable energy available from the PV panels
Three alternative topologies for integration of PV source and a UPS were studied in order to find the best
trade-off between total system efficiency, power density and system feasibility.
4.1
PV INTEGRATION IN DATA CENTRE POWER ARCHITECTURE
The first phase of the study involved a literature review of the state of the art of the UPS system and its
environment 'Data Center' to understand the problems of photovoltaic integration and the performance of the
overall system.
The PhotoVoltaic system can be integrated in different positions of the power chain of the Data Centre as
described in Figure 4-1.
The first possibility would be to connect the PV panels directly at the input of the IT equipment, but by
definition, the PV production is not constant, and it could not represent a reliable source to feed the IT power.
Another possibility to integrate the PV power within the Datacentre is to connect PV production system in
parallel with the grid which feeds the DC, via a PV inverter, which converts the DC power produced by the Solar
system into an AC signal at the frequency of the grid. Therefore, Data Centre power consumption from the grid
would be reduced thanks to produced PV power. This configuration is interesting especially when PV panels
peak power could be higher than Datacenter power requirements, so the extra power would be re-injected
into the grid.
Considering the case of GreenDataNet, where the aim is to allow Urban Data Centres (Urban DCs) to reach 80%
of renewable power and decrease their average Power Usage Effectiveness (PUE), PV panels would be installed
at building roof, so available surface for PV panels installation would never exceed datacentre surface.
Taking into account that:
-
2
Average power densities range in Urban Data Centres (Urban DCs) vary between 0,5 to 2,5 kW/m [19].
2
Maximum PV panel power density is 0.2kW/m at optimal conditions of 1400W/m² of solar irradiation,
and panel efficiency of 14%.
Consequently, PV power could never be re-injected into the grid because it could not be greater than
datacentre power requirements, moreover, this power could not even be greater than IT load power
requirements in a DC at PUE > 2 [20].As a result, if PV panel is integrated in parallel with the grid, conversion
losses will occur in order to feed IT load as seen on Figure 4-1 Possible positions for PV power injection in the
DC power infrastructure, which will reduce PV integration efficiency.
9
Figure 4-1 Possible positions for PV power injection in the DC power infrastructure
Taking into account the size of the targeted Data Centre which could be small to mid-size, the critical charge we
have to power ‘IT Load’, and the variable available energy on photovoltaic panels, the most efficient way to
take advantage of PV energy would be to connect them directly to UPS systems, so we could reduce the
amount of loss due to power conversion steps.
4.2
4.2.1
ALTERNATIVE SYSTEM TOPOLOGIES
PV SOURCE WITH A GRID-CONNECTED INVERTER
A grid-connected PV inverter in the UPS input would be a simple and straight forward solution to integrate PV
source to the Data Center power distribution system. However, the overall efficiency of the power system is
low because of four power conversion stages from PV array to the load. In a typical grid-connected PV-inverter
a DC/DC-converter and a DC/AC-converter are connected in cascade. In this topology there are additional two
power conversion stages inside a UPS.
10
Figure 4-2 PV source with a grid connected inverter
Typical weighted efficiency for the commercial grid connected PV inverter is about 96 – 97%. Accordingly,
typical efficiency in double conversion mode for premium class UPS’s is around 96% [21]. For external cabling
and connections about 1 % losses should be estimated. As a result, approximately 91-92% total power
conversion efficiency for the PV system can be achieved with this topology.
Another disadvantage in this topology is the space occupied by separate PV inverters. Also additional cabling
and connections between separate units are increasing total system losses.
4.2.2
PV SOURCE WITH A DC/DC-CONVERTER
Alternative solution to integrate PV source and a UPS is to replace inverter with the DC/DC-converter and
connect it directly to the DC-bus inside the UPS device. Compared to previous topology there are two power
conversions stages less in this topology, meaning higher total system efficiency. This topology also provides the
possibility to integrate the PV interface into the same hardware platform with the UPS and energy storage. It
will enable higher power density and smaller footprint in Data Center facility. Reduced power cabling and
connections improves also total system efficiency.
11
Figure 4-3 PV source with a DC-DC converter
Typical efficiency for the DC/DC converter in the PV interface can be as high as 98 – 99% [22] and it can be
integrated both electrically and mechanically directly to the DC-link of the UPS. Efficiency of the DC/AC
conversion inside a UPS can be higher than 98%. For external cabling and connections less than 1% losses can
be estimated. As a result, approximately 96-97% total power conversion efficiency for the PV system can be
achieved with this topology.
4.2.3
PV ARRAY CONNECTED DIRECTLY TO THE DC-BUS OF A UPS
The third topology alternative is to connect PV arrays directly to the DC-bus of the UPS. In this topology there is
only one DC/AC power conversion stage from PV source to the load. Theoretically, highest possible total power
conversion efficiency, about 97-98%, could be reached with this solution. Also integration rate and power
density of the system would be highest.
However, some drawbacks are related this solution. The DC-bus voltage of the UPS is regulated and voltage
level has to be kept in certain limits in order to perform its basic operation as a UPS device. In this topology PV
array voltage should be specified to that same narrow voltage range.
PV panel is inherently a nonlinear current source and its output power is dependent on ambient temperature
and irradiation. Based on that, there would be only few operating points and occasions when the PV source
was able to generate power to the load. So, this topology alternative is not a feasible solution to be used as we
are trying to maximize the HW platform efficiency.
12
Figure 4-4 PV array directly connected to the DC bus
4.3
OVERVIEW FOR THE SELECTED TOPOLOGY
The conclusion of the topology analysis is that the most efficient and feasible method to take advantage of the
PV energy is to connect it directly to the DC-bus of the UPS by using DC/DC-converter. This decreases losses
due to reduced number of power conversion stages, increases power density and provides flexibility for PV
array specification.
The selected system topology is presented in the figure below. It enables integration of PV source interface and
energy storage capacity into one HW platform. A PV and energy storage interfaces design is discussed more
detail in the following chapters.
The system topology should include the following functional blocks and sub-systems:

Rectifier for the utility interface

Static switch for the utility bypass operation

Inverter for the load interface

Battery charger for the energy storage interface

The EV Li-Ion battery

DC/DC converter for PV interface

PV array
13
Figure 4-5 Functional blocks of the HW system
The GDN-UPS design is based on the integration of Eaton 93PM UPS family and Nissan Leaf EV batteries.
14
5.
INTERFACE WITH PV SOURCE
The interface between a PV array and a UPS is done by means of DC/DC-converter. Following factors are
specifying the interface between the PV array and the UPS:
a)
b)
c)
d)
e)
Maximum allowed PV system voltage
DC-bus voltage of the UPS
DC/DC-converter topology selection
Grounding of the PV array system
PV system efficiency requirements
5.1
PV FIELD SPECIFICATIONS FOR THE PV INTEGRATION
With the wide expansion of PV installations, different arrangements of PV panels with their associated power
converters have been developed to increase power production and reliability of the PV generators, Figure 5-1:
PV Field Topologies. In the following chapter, most interesting system topologies will be presented, the most
adapted to GDN case will be retained, then the needed power electronic converters will be studied.
5.1.1
REVIEW OF PV FIELD TOPOLOGIES
Today, the most common topology is the central inverter [1], which consists of using a single inverter as
interface between the PV field and the grid. In this configuration, the number of PV panels connected in series
is determined to correspond to the inverter input voltage range, then several strings are connected in parallel
to meet the power constraint of the installation. The central inverter has usually a variable voltage input range,
which is possible thanks to a DC-DC converter installed at the input of the inverter.
This topology is very attractive because it is inexpensive, simple to maintain and monitor. On the other hand,
the use of a single MPPT for the entire field does not make the optimum field power extraction; especially
when it is partially shaded.
Figure 5-1: PV Field Topologies
15
The String Inverter configuration [2] consists in one inverter for each array of PV panels , also known as string.
This has the effect of increasing the number of MPPT in the field and allows to have better continuity of service
comparing to the central inverter, upon failure of the inverter, for example. However, the overall cost of the
system increases and the string inverters efficiency are low when the solar energy is not enough. To increase
the efficiency of inverters at low insolation, improved concept called “team” was developed. The team concept
is to reduce the number of inverters that convert the power of the field by installing controlled switches
between the strings at the level of the inverter. Similarly, the multi-string inverter allows a single inverter[3],
while retaining the possibility of using a MPPT per string by using a DC-DC converter per string. The main
interest is to reduce the cost investment comparing with the use of string inverters by making the DC to AC
current conversion in a single element, the main inverter. On the other side, service continuity and scalability
of this topology are reduced because this main inverter is a single point of failure.
The literature proposes also other modular systems, closer to the energy source, such as individual inverter
assemblies (or Modular Inverter) parallel/series-connected DC-DC topologies [2]. The advantage of these
architectures is to reduce the impact of one panel on the overall operations of the field, because they always
extract the maximum energy from the PV field. In return, the increasing number of power electronics
converters considerably complicates the monitoring and the maintenance of the installation.
The modular inverter allows to provide power directly to the grid. Parallel topology uses a boost converter, per
PV panel/array, connected to a higher DC voltage (400 V), which is connected to an inverter. The seriesconnected topology also uses a DC-DC converter per PV panel/array, but they are connected in series to
increase the conversion efficiency.
Finally, centralized architectures have the advantage of being simple and inexpensive while the modular
structures may offer better scalability, continuity of service, and monitoring but are more expensive.
5.1.2
SELECTED TOPOLOGY FOR GDN CASE
Since the UPS already has an inverter in its structure, inverter based PV field topologies are not adapted to
interface with the UPS. Therefore, selected PV field topologies for GDN case are shown in Figure 5-2: PV Field
Topologies in GDN Case.
16
Figure 5-2: PV Field Topologies in GDN Case
The following study will focalize on these architectures to choose the most advantageous one regarding Data
Centre power specifications.
5.1.3
DC-DC CONVERTERS FOR PV – UPS INTERFACE
Protection equipment used for residential and commercial PV systems is rated up to 1000 VDC, so it is
important to make sure a PV array is configured so that this 1000-volt rating is not exceeded. On the other
hand, the DC-bus voltage of the UPS is inherently about 720V, which determines the absolute maximum
voltage for the DC/DC-converter output voltage.
A step-up converter topology could be a feasible solution for the PV arrays with maximum output voltage less
than 720V. Accordingly, a step-down converter topology should be selected if PV array voltage were above
720V but less than 1000V. So, step-up converter topology provides more flexibility and scalability for PV string
length and multi-string specification.
Transformer-based DC-DC topologies are interesting solutions, when a high set-up ratio is needed or when the
system needs to be grounded, but since PV panel on the market have an output voltage range between 30V100V; comparing to DC link voltage of about 720V, the maximum set-up ratio could be up-to 20; in case of
modular PV field topologies which has one converter for each PV panel. In addition, security standards in
Europe better define security issues, which allows to overcome galvanic isolation. Consequently, transformerless DC-DC topologies are preferred to transformer-based ones regarding the high efficiency target.
Among the various set-up DC-DC topologies, we can distinguish the three most interesting ones Figure 5-3:
Boost Based Topologies. These topologies is based on boost converter.
17
Figure 5-3: Boost Based Topologies
5.1.4
MPPT CONTROL
A good PV converter must be capable of obtaining the maximum power available from any PV panel. The PV
converter can track the maximum power being provided by the PV panel in any condition. When the power
output does not change dramatically, the converter is obtaining the maximum power from the panel. When the
power reading changes significantly, the converter is tracking power according to the variation of the sunlight.
In GDN-UPS design, perturb and observe (P&O) method is used for tracking maximum power point [6].
The GDN-UPS has two working modes. If the power produced by the PV array is lower than the load, the boost
converter controls the PV voltage and track the maximum power point. Rest of the power to the load is
supplied by the utility mains or the battery. If the PV power is higher than the load, the DC-bus voltage may
increase above the acceptable level. In that case, the UPS rectifier or the battery converter must take care of
the DC-bus voltage control.
5.2
OPTIMIZATION OF THE PV – UPS INTERFACE
Regarding the number of possibilities to build PV-UPS interface regarding PV field arrangements, power
converter topologies, power components technologies and power requirement according to Data Centre
installation, it is difficult to select the most efficient interface by using classic iterative or comparing methods.
Therefore, an optimization method has been adopted in order to find the best solution which could meet our
goals of increasing PV energy integration ratio and improve PUE within Data Centre. An optimization tool has
been built by EATON specifically for GreenDataNet project to help designing this power interface and choosing
the most powerful topology regarding power requirement, PV integration ratio and PV panels specification.
Finding optimal system design is a sensitive task, which requires good knowledge of each component and its
effect on the system functionality. Optimization procedure is an iterative process in which system design
performance is evaluated for any variables vector many times, so modelling precision plays a principal role in
having best feasible solutions. Therefore, we chose to emulate real designing procedure, replacing simulation
software by analytical models which could accept real components, basing in their datasheets or further testing
information. Therefore, optimisation procedure gets up best feasible variables and components which fit
predefined objectives. Multi-objective optimization is adopted regarding its capability of bringing out trade-off
between several objectives, which is the case in industrial design, where the efficiency of a converter is not the
only selecting factor, and where cost and power density have often an equivalent interest with respect to final
design. The best efficiency can be obtained using oversized components, e.g. magnetic components, therefore
the aim of this work is to optimize PV-UPS interface regarding efficiency and power density of power converter.
18
There are several methods to deal with multi-objectives optimization [7], among these methods, Pareto
analysis, it does not give a single optimization result, but sums up the best compromises between several
objectives. Furthermore, Pareto solutions analysis could indicate the effect of design variables on each
objective.
The optimization procedure starts by defining desired interface specification, including Power requirement,
output voltage, DC-DC Boost topology, PV integration ratio and PV panels specifications. Also quality variables
limits are specified, as the maximal allowed variations of dc-link and PV panel output voltages and EMI limits,
which are the constraints during the optimization, other constraints related to design components are set
during the optimization in each model.
The free design variables and their limits are set, including the number of PV panel and DC-DC converters
connected in series/parallel, the switching frequency, number of power MOSFETs and/or diodes connected in
parallel and a database of several components based on their datasheets. Datasheet information is extracted
as single values for components limits and as curve fitting functions for dynamic variables. In Figure 5-4:
Optimization procedure flowchart a flowchart of the developed procedure for optimising the design variables is
shown.
Figure 5-4: Optimization procedure flowchart
19
Based on interface specifications and variables free vector, PV field arrangement is determined and
input/output voltage for each converter are derived, then Boost inductors value is calculated to produce
desired current and voltage waveforms in each component. Waveforms are then used in components models
to compute losses generate by each one, therefore, EMI filter and heat sinks are designed. Therefore,
objectives can be evaluated to find Pareto front.
In the following paragraph, optimization parameters and objectives are defined, then the models of PV panel,
the power converters, the semiconductors and the magnetic components as well as DC capacitor are briefly
summarised.
5.2.1
OPTIMIZATION OBJECTIVES DEFINITION
The MPPT control method used to track permanently the maximum power point is important for extraction the
maximum of PV array energy. But in order to transfer the maximum power to the load, converters efficiencies
must be significant. The design of high efficient power converter for PV application is difficult since the PV
power changes according to several parameters, such as irradiance, temperature, and meteorological
conditions during the year so, PV converters don’t work all the time at 100% of rated power. Therefore,
classical efficiency calculation based on 100% of power rate is not adapted for PV converters [5].
Since photovoltaic generators work in average between 30% and 100% of its rated power 80% of its operating
time, the European efficiency [4], Equation 5-1, is more adapted for the design of PV power converter, because
it takes in account the annual statistical analysis of different conditions.
Equation 5-1
Where the index of η represents the % of the photovoltaic delivered rated power and the coefficients
correspond to the statistic working time of each operation point.
Figure 5-5: Representation of the European efficiency expression as well as the efficiency characteristic evolution of a classical Boost [5].
20
In order to extend IT equipment, Data Centres owner seeks to improve power density in their installation. For
this reason, designing power converters must take into account this important factor by reducing the volume
needed for power conversion. Consequently, power density is set as objective for optimizing PV – UPS
interface, so that PV integration don’t penalize the power density of the Data Centres.
5.2.2
OPTIMIZATION PARAMETERS
Optimizing result depends on optimizing parameters or system specification and the limits fixed in variable
definitions, i.e. the total number of PV panels and PV panel specifications will limit the number of PV panels
connected in parallel/series. Figure 5-6: PV Field variables shows the graphic relation between optimization
parameters (PV panel spec & number of panels) and the optimizations variables, where
-
N=number of panels connected in series in each string
M= number of panels connected in parallel in each string
Kserie= number of converters connected in series
Kparal= number of converters module connected in parallel
Figure 5-6: PV Field variables
Table 5-1 : Variables table shows some optimization parameters and variables and their definitions and
boundaries.
Parameter/Variable
Definition
Limits
Total number of PV panels
Fixed by Integration ration & PV panel Power
Ntot
21
1 < N < Nmax = Ntot
N
number of panels connected in series
in each string
M
number of panels connected in
parallel in each string
Kserie
number of converters connected in
series
Function of Vpv, Vdc, N et M
Kparal
number of converters
connected in parallel
Function of Vpv, Vdc, N et M
F
Switching frequency of converter
20 kHz < F < 100 kHz
ΔI
PV panel Current ripple
non
10 %sonore
< ΔI < 50 %
d’entrée du Boost
Vdc voltage ripple
ΔVdc
Table 5-1 : Variables table
5.2.3
module
1 < M < Mmax=Ntot/N
important
5 % < ΔV < 10 %
360< V <450
Tension nécessaire pour que l’onduleur puisse
INTERFACE MODELING FOR OPTIMIZATION
PROCEDURE
reconstituer la tension
du réseau
PV MODULES
Several models of PV cells exist and vary in complexity and accuracy [14]. The model that will be retained in this
work is the one-diode model that contains five parameters as shown in Figure 5-7: One-Diode Model [16].
Figure 5-7: One-Diode Model [16]
The one-diode model is the most widespread model used for PV cells and PV modules due to its low complexity
and good accuracy in the power generating quadrant. The PV cell model can be adapted to PV modules by
modifying parameter values Iph, Io, Rs, Rsh, and Vt. The single diode model leads to a transcendental equation
shown in Equation 5-2, based on PV module datasheet, this equation can be solved by numerical determination
using iterative algorithms[15].
Equation 5-2
22
A. SEMICONDUCTORS:
The structure of boost converter uses a switching cell having two semiconductors switches with anti-parallel
diodes, generally a diode and a MOSFET with anti-parallel diode, beside the switching losses of the MOSFET an
additional switching losses caused by forced commutation of the diode will be generated in both the diode and
the MOSFET, these losses is due to the reverse recovery of the diode. So the choice of the diode will affect the
total switching losses [8]. The models of the semiconductors take into account switching losses generated by
the recovery of power diode within the switching cell, which allows the comparison between standard diodes
and soft switching ones.
Based on the above calculated current and voltage waveforms, turn-on and turn off currents are determined,
then using curve fitting functions, extracted from datasheet for dynamic variables, conducting and switching
losses are computed in each switching cycle. As temperatures of semiconductors are optimizing variables,
losses in semiconductors are calculated for a given temperature without any iteration.
RMS and maximum value are also computed using numerical methods then they are compared to
semiconductors limits extracted from datasheet, therefore constraints on each semiconductors are evaluated.
B. HEAT-SINK:
In this version of optimization, a single source model has been used to design heat sink for each semiconductor
component. As mentioned above, to avoid electro-thermal model which needs a lot of iterations,
semiconductor temperature is set as an optimizing variable. Furthermore, based on models proposed by [9],
natural/forced convection and radiation coefficients are calculated regarding heat sink efficiency, which
depends on fins geometry. Then heat sink length is calculated to reach the given junction temperature.
B. MAGNETIC COMPONENTS:
The design of magnetic component is a particular case in power electronic designs regarding the multiples
phenomena inside, furthermore the fact that they does not exist as a single component with its own datasheet.
Design model has to deal with magnetic cores and winding wires to get a proper design. In our case for
inductor design, characteristics of both magnetic core and winding wire have been taken into account, toroid
cores have been chosen for their high power density ratio, and also solid wires and Litz ones have been
introduced. The model starts by calculating the number of turns needed to get out the inductor value based on
core information, and then it evaluates the feasibility of the design with the preselected winding wire. The next
step is to estimate the losses generated within the core “Iron losses” and those generated within the wire
“Winding losses”.
Based on inductor current waveforms and on core datasheet, time variation of the flux density has been
determined. Then the core losses can be calculated based on equations given in the datasheet [10], where
coefficients are utilised for characterising the iron losses within core material in function of the rate of change
of the flux density (dB/dt) and switching frequency. According to core both geometry and material, saturation
of core is set as an optimizing constraint.
23
For winding losses, Dowell 1D model has been used to describe the inductor [11], which approximates winding
layers comprising multiple turns of round wire with an equivalent rectangular conducting sheet, the current
distribution change caused by this approximation is considered equivalent to a difference in conductivity. As
toroid cores winding layers length are not equal Fig. 4, winding losses are calculated for each layer by devising
the losses into proximity effect and skin effect losses as proposed in [12], Dowell representation is accurate for
closely-packed windings [13], which gives us a design limit to be set as constraint in the optimization
procedure.
a) Winding layers in toroid core
b) Transforming litz wires into conducting sheets
Fig. 4 Dowell model apply to toroid core [12]
C. DC LINK CAPACITOR:
DC capacitor and its contribution to converter power density and efficiency is not negligible, particularly in UPS
system, where near-instantaneous protection of load from short power losses of the grid is claimed. The model
of capacitor evaluates each capacitor and determines the number of capacitors needed to minimise voltage
variation, and achieve a hold up time capability. Several constraints are set on voltage and current variations
tolerance within the capacitor, then capacitor losses are calculated based on ESR value given in the datasheet.
5.2.1
OPTIMIZATION RESULTS
The optimization tool gives the specifications to build the most efficient PV-UPS Interface. Table 5-2 :
Optimization Solutions Specifications contains some solutions specifications resulting from the optimization
tool, the number of parallel converter connecting PV panels to DC link of the UPS ’Nb_PV_P’, the number of
panels connected in series & in parallel at the input of each converter, switching frequency at which converter
works to boost conversion efficiency, and power components needed to build this converter, like magnetic
core & winding wire for boost inductor and Mosfet & Diode to reduce power losses within the converter.
'Eff_Europ' 'Nb_PV_S'
98.5
15
98.1
15
97.9
15
97.6
15
97.5
10
97.4
12
97.2
15
Nb_PV_P' 'Nb_Conv_P' 'f_sw' 'Core'
2
2
58000 'T400-26B'
1
4
58000
'T520-52'
1
4
58000
'T300-40'
1
4
39000
'T400-30'
2
3
62000 'T520-40D'
1
5
60000
'T650-2'
1
4
58000
'T400-30'
'Wire' 'Nb_Mosfet'
2
1
2.8
1
1.7
1
1.25
1
2.65
1
2.36
1
1.25
1
'Nb_Diode'
2
2
2
2
2
2
2
24
96.6
96.5
96.0
95.9
95.7
95.6
95.4
5
10
5
15
5
15
6
2
1
2
2
2
1
1
6
6
6
2
6
4
10
51000
68000
51000
51000
60000
58000
53000
'T520-26'
'T520-40D'
'T300-40'
'T300-40'
'T300-40'
'T400-30'
'T520-40'
3.75
2.5
2
2
2
1.25
2.24
1
1
1
1
1
1
1
2
2
2
2
2
2
2
Table 5-2 : Optimization Solutions Specifications
The tool was tested for a PV plant of 10kW with 60 commercial PV panels (180W/72Cells/36V), DC-DC boost
topology was chosen to build the PV-UPS interface. The tool gives distributed results as a function of each
objective, here European efficiency & system size are fixed as objectives, so solutions can be evaluated
depending on the wanted performance. Figure 5-8 shows the different possible solutions, represented as blue
stars, to build the interface between PV field and the DC link of the UPS by using 2 parallel converter, matrix of
15 series*2 parallel PV panel was connected at the input of each converter. In the same figure, red circles
characterize Pareto optimal solutions, where best efficiency does not mean best power density, so a
compromise between this two objectives has to be made to choose the preferred solution.
Figure 5-8: Solution for 10kW PV planet having 2 PV matrix of 15*2 Panel connected to DC-DC converter
Running the optimization for different scenarios, with different number of parallel converter and different
possible connections of PV panels, as explained in paragraph “5.1.25.1.2”, gives results shown on Figure 5-9:
Pareto Front for different possible connections, where a Pareto Front for each scenario was plotted. The figure
indicate that Pareto Front with fewer power converters have better solutions. Also, increasing PV panels
connected in series improves power conversion efficiency.
25
Figure 5-9: Pareto Front for different possible connections
Therefore, to build an efficient PV-UPS interface for 10kW PV plant, a 2-parallel converter system would give
the best conversion efficiency with a minimum of power density .
26
6.
INTERFACE WITH THE NISSAN BATTERIES
The Nissan Li-Ion battery system is originally designed for electric vehicle (EV) application, while the UPS design
is optimized for stationary lead-acid batteries. Requirements for EV batteries are different compared to leadacid technology, which is conventional e.g. in Data Center batteries. Those differences are leading to various
challenges and trade-off in the interface design between Nissan battery and the Eaton UPS.
Major issues which are requiring specific attention in the battery interface design:
6.1

Voltage matching between the battery and the rest of the system is a key issue. Specifically,
because operating voltage range of Nissan battery is lower than the voltage typically used in 3phase UPS’s

Operation and functionality of the EV battery system deviate from the conventional UPS system
functionality

Communication protocols of the Nissan battery and 93PM UPS requires coordination

General safety issues related to Li-Ion batteries
OUTLINE OF THE BATTERY INTERFACE
Outline of the Nissan battery interface is described in picture below. It consists of up to four parallel connected
battery packs, the Battery Management System (BMS), power supply and a UPS. The master controller (MCU) is
physically located inside the UPS. Each battery pack includes internal high voltage relay which connect or
disconnect the pack following the demand.
27
The operation principle of the system is as follows. The BMS is monitoring and controlling a state of the battery
packs. Internal CAN-bus is used for communication between BMS and battery packs, while communication
between MCU and BMS is based on the RS-485. The demand to transfer to the battery mode operation is
initiated by the MCU. Accordingly, the MCU will send a command to the BMS when transfer back to other, e.g.
grid operation, mode is acceptable. The BMS will switch ON or OFF the high voltage relays inside the battery
packs according to MCU commands.
6.2
NISSAN BATTERY PACK AND 93PM BATTERY CONVERTER SPECIFICATIONS
Nissan battery pack has following technical specifications [18] :
Nominal voltage
Operating voltage range
Initial capacity for new battery (@ 0.2C discharge)
Estimated capacity for second-life battery
Max. discharge current
Max. charge current
Nominal cell voltage
Number of cells in series
360 V
240 – 403 V
24 kWh (~66 Ah)
17 kWh
400 A
200 A
3,8V
96
Battery converter
operating voltage
range
Specification for the battery converter of the 93PM UPS is presented in the table below. It shall be noted that
the UPS is automatically limiting charge current to 16.5A, if the load is more than 40 kVA. When the load is less
than 40 kVA battery charger is able to charge battery simultaneously at 29.3A current.
28
Charging battery
Operating voltage range
Maximum charge current
Charge current
Discharging battery
Minimum battery voltage
Minimum cell voltage in discharge
Maximum continuous discharge power
330 – 550 V
29.3 A (when load < 40 kVA)
16.5 A (automatically limited if load > 40 kVA)
336 V
3,50 V
50 kW
Increase of battery power and charge current capacity of 93PM UPS is possible by connecting more power
modules into the system.
The operating voltage range of the battery converter and Nissan battery are overlapping in the range from
330V up to 403V so this is voltage range used for charging and discharging. Battery capacity can be utilized in
that voltage range. Referring to the discharge diagram above, one can see that available capacity from the
battery is dependent on the discharge rate. With high discharge rate (3C) available battery capacity is only
about 70% from the nominal. That is equal to16.8 kWh from new battery and about 12 kWh from the secondlife battery. However, with the slow discharge rate (1C or less) even more that 90% of the battery capacity is
available. That corresponds to 21.6 kWh from new battery and about 15 kWh from second-life battery.
29
7.
DEMONSTRATOR
A prototype of UPS integrating the PV source and the Li-Ion batteries will be implemented in order to validate
the forecast performances. The architecture of the System is represented in next page This prototype will be
installed in the laboratory of the CEA-INES in Chambery.
According to the results of the analysis and of the optimization tool, a DC-DC converter will be integrated to the
DC bus of a 50kVA UPS in order to be able to inject the power coming from a 10kW PV field (representing an
integration of 20% of renewable self-production). This DC-DC converter will be composed by 2 Boost modules
of 5 kW in parallel connected to the DC bus of the UPS, this choice has been pushed in order to privilege the
modularity of the PV interface. On a first step, the 20% of self-production would allow GDN project to validate
the UPS modified infrastructure and measure the efficiency of the system, on a second step, further DC-DC
modules can be added in order to validate the simulations results produced by CEA in D1.3 and reach a higher
percentage of self-production.
30
Figure 7-1 GDN Communication Architecture
31
Figure 7-2 GDN HW System Architecture
GDN-UPS System Enclosure
Static Switch
Utility Mains
3~ 400/230V
Rectifier
Inverter
Load
3~ 400/230V
DC-link 720V
Utility Mains
3~ 400/230V
UPS
MCU
DC/DCConverters
Power
CPU
Battery
Converter
Communication
Bays:
Ethernet,
SNMP,
1 relay output, 5
signal inputs
4R BMS
PV Array:
2*5kW sub-arrays
Uin: 350 – 700V
Iin_max: 10A /array
7" touch
screen
NISSAN Leaf
battery pack
1 – 4 * 24 kWh
HMI
Communication
Interfaces:
UPS information
to external
Mgmt system
8.
REFERENCES
[1] M. Meinhardt, “Past, present and future of grid-connected photovoltaic and hybrid power systems”,
PESSM’00 (2000) pp.1283-1288
[2] A. Pavan et al., “Power electronics conditioning systems for industrial photovoltaic fields: Centralized or
string inverters?”, ICCEP’07 (2007) pp.208-214.
[3] W. Yao et al, “Improvement of Performance and Flexibility for Photovoltaic Module using Individual DC/DC
Converter”, IPEMC’09 (2009), pp.441-444.
[4] H. Haeberlin, L. Borgne, M. Kaempfre and U.Zwahlen. “New Tests at Grid-Connected PV Inverters: Overview
over Test Results and Measured Values of Total Efficiency”. 21st European Photovoltaic Solar Energy
Conference, Dresden, Germany, Sept. 2006.
[5] A. Berasategi, C. Cabal, C. Alonso, B. Estibals, “European efficiency improvement in photovoltaic
applications by means of parallel connection of power converters”, Power Electronics and Applications, 2009.
EPE '09. 13th European Conference.
[6] N. Femia and all “A Technique for Improving P&O MPPT Performances of Double-Stage Grid-Connected
Photovoltaic Systems”, IEEE Trans. Ind. Electron., vol. 56, no 11,November 2009, Pages: 4473-4481.
[7] K. Deb, “Multi-Objective Optimization Using Evolutionary Algorithms”, New York: Wiley, 2001.
[8] P. Haaf and J. Harper, “Understanding Diode Reverse Recovery and its Effect on Switching Losses”, Fairchild
Power Seminar 2007.
[9] Y.S. Muzych, J.R. Culham and M.M. Tvvanovich, “Thermal spreading resistance of eccentric heat sources on
rectangular flux channels,” Journal of electronic ackaging,Vol.125 June, (2003), pp.178-185..
[10] [Online]. Available: www.micrometals.com
[11] P. L. Dowell "Effects of eddy currents in transformer windings", Proc. IEE, vol. 113, pp.1387 -1394 1966.
[12] G. Lefevre , H. Chazal , J. Perrieux and J. Roudet "Application of Dowell method for nanocrystalline toroid
high frequency transformers", Proc. PESC\'04, pp.899 -904 2004.
[13] Xi Nan and C. R. Sullivan, “An Improved Calculation of Proximity-Effect Loss in High-Frequency Windings of
Round Conductors”, IEEE Power Electronics Specialists Conference, June 2003, pp. 853–860.
[14] Matos et al, “A model for semiconductor photovoltaic (PV) solar cells: the physics of the energy
conversion, from the solar spectrum to dc electric power”, ICCEP’07 (2007) pp.352-359:
[15] D. Sera, “PV panel model based on datasheet values”, International Symposium on Industrial Electronics
2007, pp.2392-2396
[16] K.L. Kennerud, “Analysis of performance degradation in CdS solar cells”, IEEE Transactions on Aerospace
and Electronic Systems Vol. AES-5 No.6 (1969) pp.912-917
[17] Hardcastle Jonathon: Forecast: Data Centers, Worldwide, 2010-2015. Gartner Research. (2011)
[18] Nissan Battery specifications (restricted)
[19] GreenDataNet Project. “Urban Data Centre Definition and Specification” (2014)
[20] GreenDataNet Project. “Renewable energy maps and Guidelines” (2014)
[21] EATON EATON 93PM UPS Technical specs (2014) Website
http://powerquality.eaton.com/Products-services/Backup-Power-UPS/93PM-UPS/93PM-specs.asp?CX=3
[22] Skvarenina, Tim L. (editor) The power electronics handbook, Boca Raton, Fla.: CRC Press, c 2002. p. 664
34
Download PDF
Similar pages