648
IEEE Transactions
on
Energy Conversion,
Vol.
7,
No.
4
December
1992.
A KNOWLEDGEBASED APPROACH TO THE DESIGN OF INTEGRATED RENEWABLE ENERGY SYSTEMS
R. Ramakumar I. Abouzahr Senior Member, IEEE Member, IEEE Director Research Associate Engineering Energy Laboratory Oklahoma State University Stillwater, OK USA 74078
Keywords:
Integrated Renewable Energy Systems (IRES) Design; KnowledgeBased Design Approach; StandAlone Energy System Design; Design of Renewable Energy Systems; Hybrid Systems; WindElectric Conversion Systems; Photovoltaic Systems; Biogas Usage.
ABSTRACT
Integrated Renewable Energy Systems (IRES) utilize
two
or more renewable energy resources and enduse technologies to supply a variety of energy needs, often in a standalone mode.
A
knowledgebased design approach that minimizes the total capital cost at a preselected reliability level is presented. The reliability level is quantified by the loss of power supply probability (LPSP). The procedure includes some resourceneed matching based on economics, the quality of energy needed, and the characteristics of the resource.
A
detailed example is presented and discussed to illustrate the usefulness of the design approach.
INTRODUCTION
Global environmental concerns coupled with steady progress
in
renewable energy technologies are opening up new opportunities to hamess and utilize different manifestations of solar energy
[13].
Nowhere are these opportunities more pertinent than in powering remote loads requiring a mixture of different grades of energy 1461. Integrated Renewable energy systems (IRES) utilize
two
or more renewable energy resources and enduse technologies to supply a variety of energy needs [7,8]. Such systems are typically operated
in
a standalone mode. However, they can also function
as
easily in conjunction with conventional energy systems such as dieselelectric generators and/or utility grid connection. The most effective use for
IRES
s believed to be in the standalone mode to energize the
two
million or
so
remote villages in the world with no grid connection [9111. Obviously, there are many other situations where
IRES
can contribute by providing energy without increasing the associated environmental burdens. The renewable energy resources that require consideration are (1) biomass,
(2)
insolation, (3) wind, and (4) hydro. All the energy needs can
be
consolidated
into
four categories
as
given below:
1.
mediumgrade thermal energy (100' C to
300'
C),
2.
lowgrade thermal energy (less than 100' C),
3.
rotating shaft power (fixed or variable location), and 4. electricity
92
WM
0372 EC
A
paper recommended and approved by the IEEE Energy Development and Power Generation Committee of the IEEE Power Engineering Society for presentation at the IEEE/PES 1992 Winter Meeting, New York, New York, January 26

30,
1992. Manuscript submitted April 23, 1991; made available for printing January 9, 1992.
K.
Ashenayi Member, IEEE Assistant Professor Dept. of Electrical Engineering The University of Tulsa Tulsa,
OK
USA 74104 Basically, there are two options to supply energy
to
a variety of loads by utilizing different resources in tandem.
ll
the resources can be converted to one versatile form (typically electrical) for storage and supply to the users. While this may be convenient, it is not always economical. The alternate approach is to match the xsources, devices, and the needs and achieve integration of benefits at the user end. In most cases, the latter approach results in
an
econoinically viable option [12].
IRES
Design The ultimate objective of any design procedure employed is to obtain the sizes and ratings of the various energy conversion and energy storage devices needed to supply energy to the different loads. Moreover, energy supplies are typically required at a pre selected reliability level and at minimum cost. Since
IRES
s very capital intensive and inputs are typically free or inexpensive, minimizing the capital cost will essentially result in a system with minimum annual cost. Some of the resources (wind and insolation) are highly stochastic and sitespecific. Others (biomass and hydro) are more predictable, though they have seasonal variations and are
also
site specific. Among the loads, some will be more variable than others and some of them can be easily predicted. The design procedure should consider all these factors and include them without unduly expanding he amount of calculations equired. Previous works
on
IRES design have proposed chronological simulation [13], linear programming[l4,15], goal programming [16], and a probabilistic approach involving the loss of power supply probability (LPSP)
as
a measure of the quality of the power supply [17]. Chronological simulation requires extensive data on the resources and loads which may not always be available. Linear programming and goal programming approaches are deterministic and they employ seasonal or annual average values in the analyses.
In
addition, they do not consider the quality (reliability) of the power supply in the design methodologies. Conventional probabilistic approaches require a considerable amount of computations which may not
be
justifiable. This paper presents a knowledgebased approach to the design of an
IRES.
t is a modification of the approach presented in Reference 17 in that a knowledgebase and a search algorithm are used
in
the last stages of the design program instead of a series of long and timeconsuming computations. The stored data that
go
with the knowledgebase have been developed by employing the results documented in a series of two IEEE papers
[18,19]
on the calculation of the loss of power supply probability of standalone windelectric and photovoltaic systems.
n
example is presented to illustrate the proposed design approach.
OVERALL DESIGN APPROACH
The year is divided into
as
many number of timesections (seasons) as needed such that the resources and the loads have reasonably similar daily variations or are constant during each period of time. The design procedure is applied for each timesection to find the ratings of energy converters and/or quantities of resources needed and the sizes of energy storage systems to satisfy the energy needs at the desired reliability level subject to resource availabilities
08858969/92$03.000 992
EEE
649
and at
Mum
apital Cost. The final design (for the entire year) is based on these seasonal designs and the prioritization of the various
SXWXN
If the final design values are selected based on the worst possible combination of resource availabilities and loads, then a considerable amount of excess energy will
be
available d&g most of the rest of the time. Unless this excess energy can be productively utilized, the overall design will fail to be cost effective. The information, knowledge, and data base used in the development of the design procedure
are
discussed next.
KNOWLEDGE BASE
Energy systems usually consist of several interconnected components performing together to satisfy a set of energy needs.
n
IRES
may consist of solarthermal collectors, water turbines and pumps, windmechanical conversion systems (WMCS), wind electric conversion systems
(WECS),
photovoltaic mays (PV), and different types of energy storage and reconversion systems and end use devices in one or more of many possible combinations to satisfy a variety of energy needs. Different energy resources are variable to different degrees and some of them are complementary over the annual cycle.
Loads
also vary to different extents and the quality of the energy supply can be characterized by the probability of loss of power supply to the loads. Energy conversion devices are derived from different technologies and are matched to specific situations based on the type of energy supply needed and the enduse scenarios.
A
knowledgebased design approach can consider
all
these factors by proper rules of assignment for resourcesneeds combinations [20,21]. Any energy resource can
be
used to satisfy all the energy needs by employing a string of energy conversion and interface devices. However, some resourceneed combinations are more logical than others because of cost and efficiency considerations. This point has been discussed in detail [22]. The design approach presented in
this
paper orders, prioritizes, matches, and finally finds the ratings of the various energy conversion devices and sizes of different energy storage components required. Wind Reeimes
nd
WECS Itis generally accepted that wind regimes can
be
modeled ushe a Weibull distribution exmessed in terms of a scale Parameter and shape parameter [23]. dese wo parameters can
&
obtained from the mean and standard deviation of the sample set. For developing the knowledge base, mean wind speeds are quantified into three distinct values: low
(4m/s),
medium (6m/s) and high
(8m/s).
With each of these mean wind speeds,
t m
possible values of standard deviation are considered. Thus nine different wind regimes are employed to characterize the wind resource. Table I lists the nine sets
of
values used. The choice of these values is based on typical wind regimes
that
exist around the world [24]. The power output
of
a WECS depends on the incident wind speed, WECS power rating, and the cutin, rated, and furling wind speeds. Based on a survey
of
the specifications of most of the available small(<100 kW) WECS, these wind
speeds
are
taken to be 3.5
m/s
10
m/s,
and 22
m/s
respectively in this work. Insolation Regimes and PV Although several complex models are available [25], insolation regimes are represented in this work in terms of a beta distributed random variable 'cloud cover' [26]. The
two
parameters of the beta distribution can be derived from the mean and standard deviation of the factor by which insolation
is
reduced
as
compared
o
the maximum possible value during the study period. Based on observed insolation data 1271, nine pairs of these parameters are employed to characterize this resource. Table
II
lists these values. Photovoltaic arrays consist
of
individual cells fabricated using one of several possible technologies: singlecrystal silicon, polycrystalline silicon, amorphous silicon, thinfilm devices, etc. The conversion efficiency of a PV module depends on the particular technology employed, insolation, and the module temperature.
In
this work, a simple model in terms
of
an average efficiency is used.
Also
a value of (ln)
is
assumed for the fractionofdaytime. Since the objective
is
to
'design' the system in the presence of a multitude of uncertainties and variabilities, it is believed that a simple model would be adequate for considering the performance of PV systems. Data Base The data base that is an integral part of the knowledge base consists of sets of data relating the amount of energy storage required to the resulting LPSP for various combinations of loads and energy converter ratings under different input (wind and/or insolation) regimes. The energy converter could
be
PV or WECS or PV and WECS. Nontracking flatplate PV arrays rated at 2 to 22 kW in steps of 2 kW
are
assumed. The rating of the WECS is also varied from
2
to 22 kW in steps of 2 kW. The electrical load is assumed to
be
unifonnly distributed with its maximum value
Lmax
ranging from 2 to 12 kW in steps of 2 kW, with et at
10%
of
Lax
he LPSP values considered are allowed to range from 0.02 to 0.18 in increments of 0.02. For the sake of uniformity, a study period of 1 hour is assumed for the data in storage. If all the hours during the study period are similar in the statistical sense
,
then the values corresponding to any other study period can be obtained by multiplying the hourly values in storage by the number of hours in the study period. For each combination of energy converter(s) rating(s), load, wind regime and/or insolation regime, assuming a study period of one hour, the following quantities
are
calculated (see References 18 and 19) and stored: (i)
ii)
E,, the amount of energy input to storage,
kwh
E,, the amount of energy in
kwh
required to
be
drawn from energy storage to achieve an LPSP of zero, and
i)
set of LPSP and corresponding
SSMIN
values where SSMTN=E2pE1 TABLE I
WIND
REGIME
WEIBULL
PARAMETERS FOR THE DESIGN EXAMPLE Pwmls
a,
m/s
a
m/s
P
4 2.00 4.52 2.10 4 2.21 4.5 1 1.88 4 2.78 4.42 1.46 6 6 6 2.50 2.77 3.47 6.76 2.57 6.77 2.30 6.74 1.79
8
2.93 8.96 2.97
8
3.24 9.00 2.66
8
4.07 9.03 2.06 TABLE
II
CLOUD
COVER
REGIME BETA PARAMETERS FOR THE DESIGN EXAMPLE
Ps
0.1 0.2
0.3
0.4
0.5
0.6 0.7
0.8
0.9
0s
0.0900 0.1600 0.2200 0.2610 0.288 0.2610 0.2200 0.1600 0.0900
a
1.0111
1.0500 1.0016 1.0093 1.0070 1.5139 2.3372 4.2000 9.1000
P.
9.1000 4.2000 2.3372 1.5139 1.0070 1.0093 1.001
6
1.0500 1.0111
650
in which E2 <E2 is the energy required to be drawn from storage and suppliecfto the load to achieve the assumed value of LPSP.
All
this information is systematically arranged and stored in data files for easy access. Appendix A shows typical data files for
SSMLN
versus LPSP curves. With
9
different wind regimes,
9
different insolation regimes,
6
different load levels,
11
different WECS ratings and 11 different PV ratings, a total of
58 104
data sets are calculated and kept in storage
as
a part of the knowledge base. Enernv Storage Calculatiom
As
mentioned in the previous section, the values of El, E2, and LPSP versus
SSMIN
data in storage are all based on a study period of one hour. These values can indeed be multiplied by the number of hours in other study periods under consideration as long
as
they have
similar
characteristics. The actual amount of energy storage required for a given study period is calculated using these values
as
discussed next. During any one timesection (season), the diurnal variations of wind and insolation resources can be assumed to
be
repetitive in the statistical sense.
In
addition there will be longterm variations as described by the distribution
(or
density) function.
A
close examination of the mechanism of energy storage and withdrawal leads one to conclude that the amount of energy storage required can
be
assumed to consist of two components:
1.
the
storage required on a daily (24 hours) basis, and 2. longterm storage to meet the specified LPSP requirements when the resource cannot supply the demand during the study period.
If
the energy converters (WECS and/or PV) are oversized for the study period under question, then El will be greater than E2 and the amount of energy storage required is simply equal to 24E2. Clearly, under these circumstances, some of the energy generated will
be
dumped and not utilized. If El is equal to E2, then the amount of energy storage required is equal to
El
or E2 times
24.
In
both cases, no additional longterm storage
is
needed. If E1 is less than E2, then the amount of energy storage required
is
equal to the sum of 24E1 and the longterm component (SSMINXT) based on the required LPSP. The amount of energy storage as calculated above is multiplied by a factor of 2 to account for the fact that energy storage systems should not
be
depleted below a certain level (assumed to be
50 )
to preserve their lifetime and reliability attributes. Resource and Need Prioritization Recognizing the fact that under the present economic conditions
IRES
will be most appropriate to energize remote areas, the energy needs and the resources
are
prioritized
as
follows: a. mediumgrade heat b. lowgrade heat c. mechanical shaft power (fixed or variable location) d. electricity a. biomass (biogas) b. solarthermal energy
C.
hydro power d. wind (WECS,WCS) and/or insolation (PV) Needs: Resources: Biogas obtained by anaerobic digestion of biomass is the cheapest energy resource at the present time. It can
be
burned directly to obtain mediumgrade and/or lowgrade thermal energy, and can be used in an internal combustion engine to obtain mechanical shaft power and/or electricity by coupling a generator. For these reasons, biogas is given the highest priority. Simple solar collectors can be used to obtain lowgrade thermal energy fairly economically. Hydro power can be used via turbines to obtain mechanical shaft power and/or electricity. Windmechanical converters also can be used to obtain rotating shaft power. Finally, though relatively expensive, wind and insolation can
be
converted to electrical form using WECS and PV systems respectively. Based on this brief discussion, the logic behind the ordering of the needs and the resources as given above is evident. The design procedure considered in the next section discusses the fallback positions for cases where resources are in short supply.
DESIGN PROCEDURE
The knowledgebased design approach that has been developed is presented in flowchart form in Figure
1.
This computer program will
be
identified by the acronym
IRESKB.
Though this procedure considers a variety of energy resources and loads, the absence of any of these can be easily handled by setting the appropriate quantities equal to zero. The first step in the design procedure is to divide the year into
NY
timesections (or seasons) where each season is characterized by a set of resource availabilities and load requirements.
As
mentioned earlier, the design procedure
is
applied to each of these
NY
easons to result in
as
many seasonal designs. The final design will
be
selected from these seasonal designs. The next step
is
to find the maximum area cmax of solar thermal collector needed to satisfy lowgrade thermal energy requirements. This is followed by a determination of the maximum rating wmax of windmechanical conversion system (WMCS) needed to power
l
the rotating shaft loads at fixed locations. These two values are then divided into smaller areas/ratings
as
dictated by their market availability. The design process starts with a selected set of these two values and goes through a series of steps, culminating in an optimum combination of resource consumption, energy converter ratings, and energy storage device sizes that
minimizes
the total installed cost. Mediumgrade heat requirements are always satisfied by biogas, if available. Any portion that is not satisfied is added to the electrical load for that season with suitable conversion factors. Lowgrade heat requirements are satisfied by a combination of biogas and one of the solar collector segments chosen. The portion of lowgrade heat that is not satisfied is added to the electrical load for that season. In the next step, mechanical shaft power requirements at fixed locations are satisfied by a combination of biogasfueled engines and WMCS rated at a fraction of the maximum rating wmax calculated earlier. Any portion of
this
of load that is not satisfied by this combination is added to the electrical load. Mechanical shaft power at variable locations is satisfied by biogas, if still available. The part that is not fulfiied is once again added to
the
electrical load
as
appropriate At
this
stage in the design process, the only load left to
be
satisfied is electrical, which is an aggregate of
all
the additions discussed above and the loads that were electrical to start with. The electrical loads are satisfied by biogasfueledengine driven generators
if
there
is
still any biogas left. Any leftover loads are supplied by waterturbinedriven generators if available. Finally, the remaining portion of the electrical loads is satisfied by a combination of WECS
and/or
PV with energy storage. The design procedure has reached the point at which it is required to find the ratings of WECS and/or PV and the size of energy storage required to satisfy a certain electrical load at a predetermined reliability level (as quantified by LPSP) and at minimum capital cost. To accomplish this, the data base in storage is searched to find all combinations of
WECS
and/or
PV and energy storage that will meet the LPSP requirement. The capital costs of these combinations are calculated and the one that results in miniinum cost
is
chosen. This design process is repeated for each timesection (season) to obtain a set
(NY
n number) of designs. The design values for each season will consist of solar collector ma,
WMCS
rating, WECS rating, PV array rating, ratings of water turbines, biogasfueled engines and enginegenerators, ue of energy storage, amount of biogas required, and the amount of water storage. The entire design procedure described thus far is repeated for all the different and reasonable combinations of solar collector
areas
and
WMCS
ratings. The final (applicable for the entire year) design values are selected from these seasonal designs.
DESIGN EXAMPLE
A
remote agricultural village with no electrical grid connection and a population of
3 11
is chosen
as
an example for the application of IRESKB presented in this paper. The number of cattle
is
367.
The mount of animal and agricultural wastes available is seasonal as expected.
65
1
enera
need
are
M for MGH
L
forLGH R forMRS F for
FRS
E
for
EL
flat plate
sc
areas
:
=
0,cmax
WCS
atings
w
=
0,wmax
U
do c
=
o,cmax

o w
=
o,wmax
+
o
s
=
1,NY

atisfy M(s ) by BI(s )
I
II
RM(s )
=
0
BI(s )
=
remaining
E(s )
=
E(&)
+
RM(s ) BI(s )
=
0
satisfy
L
by
I
E@#)
=
E(&)
RL(s ) BI(s )
=
0
I(s )
=
remaining satisfy
F
by
.
RF(s )
=
0
E(s )
=
E( )
+
RF(s )
r
T
I=
I(s )
=
remaining E(s )
=
E@#)
RR(s )
BI(s )
=
0
t l
atisfy use database to
\
fiidr;tings optimum rating minimum cost xit with optimum design Figure
1.
IRESKB Plowchart.