M.

Nakamura
Associate Professor of Systems
Control Laboratory.

N. Egashira Digital Control of Working Fluid

Postgraduate student of Systems
Control Laboratory. Flow Rate for an OTEC Plant
Department of Electrical Engineering, The role of control in operating an OTEC plant efficiently is of great importance.
Saga University,
This paper describes digital control of working fluid rate based on an adaptive con-
Honjomachi, Saga 840, Japan
trol theory for the "Imari 2" OTEC plant at Saga University. Provisions have been
made for linkage between the software of the adaptive control theory and the hard-
H. Uehara ware of the OTEC plant. We can obtain satisfactory control performance using this
Director of OTEC Laboratory,
digital control system.
Department of Mechanical Engineering,
Saga University,
Honjomachi, Saga 840, Japan

1 Introduction
The purpose of ocean thermal energy conversion (OTEC) is
to extract power from temperature differences existing in the
oceans. The principle of OTEC was first proposed by D'Ar-
sonval [1] in 1881. Since the oil-shock of 1974, there has been
considerable interest in producing a feasible OTEC system,
and a good deal of research has been devoted to the subject.
One of the authors has designed an OTEC plant, was the first
to succeed in extracting power using the OTEC plant [2], and
has improved the design of the OTEC plant based on steady-
state operating conditions [3]. The next phase of OTEC
research is the control of OTEC plants under varying condi-
tions. Owens [4] studied two types of control problems of
OTEC plants, which were modelled by nonlinear differential
equations; response and control of the OTEC plants were
analyzed by computer simulation. Because the thermal effi-
ciency in OTEC plants is generally low, on the order of 3 per-
cent, and the operating conditions, such as seawater
temperatures, are constantly changing, the role of control is of
great importance in operating an OTEC plant effectively.
Among the many control theories, adaptive control is the most
promising for this purpose. The adaptive control theory gives
a control algorithm which tunes its parameter on-line for the
plant. Recently, adaptive control theories have been greatly
improved [5, 6], and digital controls, using adaptive control
theories programmed in computers, have been successfully ap-
plied to industrial processes such as paper machine control,
ore crusher control, and so on [5], But to our knowledge, there
has not been any attempt to utilize digital control at OTEC
plants.
In this paper, we explain the digital control of working fluid
flow rate for the "Imari 2 " OTEC plant [7] at Saga Universi-
ty. The innovative feature of this study is the construction of a
digital control system and the implementation of digital con-
Contributed by the Solar Energy Division for publication in the JOURNAL OF
SOLAR ENERGY ENGINEERING. Manuscript received by the Solar Energy Division,
February, 1985.
trol for the real pilot OTEC plant. In the construction of the
system, a provision has been made for linkage between the
software of control theories and the hardware of the OTEC
plant. We obtained satisfactory control performance using
this digital control system.
2 OTEC Plant and Control Objective
The closed cycle OTEC plant, shown in Fig. 1, consists of
an evaporator, condenser, turbine-generator, and pumps,
which are interconnected by pipe lines. Warm surface water is
used to evaporate a working fluid, such as Freon 22 (R-22), in
the evaporator. The dry Freon 22 vapor then expands in a tur-
bine to produce power. Finally, cold subsurface water is used
to condense the Freon 22 vapor in the condenser.
The ultimate purpose of OTEC plant control is to provide
stable power efficiently by appropriately regulating the
seawater flow rates and the working fluid flow rate under con-
ditions of continually changing seawater temperatures.
The control objective is to find a process input signal which
will produce an appropriate working fluid flow rate from the
Turbine Generator
Warm surface
water
Cold subsurface
water
Fig. 1 Scheme for closed cycle OTEC plant

Journal of Solar Energy Engineering MAY 1986, Vol. 1 0 8 / 1 1 1

Copyright © 1986 by ASME
Downloaded From: http://solarenergyengineering.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jseedo/28189/ on 04/06/2017 Terms of Use: http://www.asme
 DIGITAL CONTROL SIDE ! OTEC PLANT SIDE
desired [~ ' Input signal part ' ' Controlled process '
Adaptive control ' ' Input signal '
value I
part ) [characteristics part)
QD U Vu Vu'VUi Vu
identification a ll
u/v evapo 9e
control algorithms chara chara [ j * ® ^ amp v/u pump rator "]"'H t u r b i n e M ? .tor
n?
iu J I j

QF QF
rriF
<e
heat
rate
thermo-
equation
couple
chara
I Detection characteristics part i

Fig. 2 Digital control system for working fluid flow rate

working fluid pump, so that the heat rate of the working fluid working fluid flow rate and the working fluid temperatures.
QF at the evaporator will behave according to a desired value. The working fluid flow rate is detected by an electro-
The controlled process consists of the plate type evaporator oscillation type flow meter and an amplifier. The
and the working fluid pump. The plant must also be equipped characteristics of the amplifier are expressed as
with an input signal part and a detection part. The digital con-
trol part consists of a computer (TOSBAC 600) programmed V„ = V*„,(V) (1)
with adaptive control theories and equipped with analog- and the relation between the flow rate mF and the voltage Vm
digital (A/D) converters and a digital-analog (D/A) converter. is given by
mF = 0.833 F,„(kg/s). (2)
3 Digital Control System
The working fluid liquid temperature TFI at the inlet and the
The digital control system for working fluid flow rate, working fluid vapor temperature Tvo at the outlet of the
depicted in Fig. 2, consists of the OTEC plant side and the evaporator are detected by thermocouples and amplifiers. The
digital control side. The two sides are connected by A / D con- amplification factor is 5000, and the characteristics of the
verters and a D/A converter, and they have a symmetric rela- amplifiers are written
tionship at the converters in Fig. 2 between a real and a mirror
image. Amplifiers in the OTEC plant side are used to stand- VT = 0.200 V*r(mV). (3)
ardize each of the analog values to the restriction of the A / D The thermocouples are characterized by the following
converters (0-10.24 V) and to that of the D/A converter polynomial equations of VT based on the least squares
(0~ 10.24 V), respectively. The digital control side consists of method:
the following blocks: a detection characteristics part, an adap-
7>/ = 2 . 1 1 6 x l 0 - 2 + 2.506xl0 1 K 7 F + 6..151xl0- 1 K 7 y 2
tive control part, and an input signal characteristics part. The
components of each part, programmed in the computer, are - 8.173 x lO" 1 VTI? + 1.805 x 1 0 ' 1 * V ( ° C ) , (4)
described below in detail. 3 1 2 2
Tvo = -6.700X 10" +2.539X 10 VTV-2.065 x\Q- VTV
3
3.1 Detection Characteristics Part. The heat rate of the -4.520x 1 0 - ' VTV + 1.114 x 1 0 - ' VTV* (°C). (5)
working fluid given at the evaporator is determined by the The heat rate of working fluid QF in the evaporator is deter-

Nomenclature

unknown parameter
of process
«i ( * ) estimated parameter Q = heat rate (kW)
of process r2, r3 = parameter of Superscripts
b0, bl = unknown parameter observer-and- * = standard value for
of process (kJ/kg) regulator the restriction of
b0(k),bx(k) estimated parameter T = temperature (°C) A / D D/A converters
of process (kJ/kg) U(k) = process input (kg/s) i = power
C(k) estimation matrix V = voltage (V, mV) T = transposition
specific heat W(k) = process disturbance
(kJ/kgK) (kW) Subscripts
m mass flow rate (kg/s) x(k) = estimated state of C — cold water
h specific enthalpy observer (kW) D = desired value
(kJ/kg) z~x = backwards shift F = working fluid liquid
I identity matrix operator m = mass flow rate
K(k) gain matrix z(k) = input-output vector H = hot water
k control stage \ , , X2 = pole of regulator I = inlet
N number of control /* = pole of observer O = outlet
stages p = index of exponential S = condensed liquid
p(k) = estimated parameter weighted least T = temperature
vector squares method V = working fluid vapor

112/Vol. 108, MAY 1986 Transactions of the ASME

Downloaded From: http://solarenergyengineering.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jseedo/28189/ on 04/06/2017 Terms of Use: http://www.asme

 1 1 1 1 | , , i i | , , , ,
1 ' ' ' ' .

4UU h(Tvo) ~
300 -_

200 h(TFi) ~

100

I , , , , 1 , , , , I I I , , "
0 10 20 30 40
o 0.5 1.0 1.5 2.0 2.5
Temperature T [°CI
Working Fluid Flow Rate U ikg/si
Fig. 3 Characteristic curves of specific enthalpy for Freon 22
Fig. 4 Characteristic curve of working fluid pump input

mined by the working fluid flow rate mF and the specific en-
thalpy difference between h(Tvo) and h{TFI) as follows:
QF = mF{h(Tvo)-h(TFr)}(kW).
The characteristics of the specific enthalpies are formulated by
polynomial equations based on the least squares method for
the data in the Freon 22 Table [8] as
h{TV0) = 4.047 x 102 +3.592x
The index p is in the range 0 < p ts 1, and / is the identity
matrix.
(6)
(b) Control Algorithm. We adopt the observer-and-
regulator using an extended model [10] as the control
algorithm. Arranging the control algorithm, we obtain a
process input for the next stage k + 1 as
\0"lTvo k

-1.429xl0-37Vo2 U(k+l) = rl(k)x(k+l) + r2(k) X) {QF(')-QD(')}. (14)

5 3
-1.788xl0- 7Vo (kJ/kg), (7)
The value QD(i) represents a desired output value, and the
2 3 2
h{ TFl) = 2.000 Xl0 + 1.158 Xl0°7> 7 + 1.519 X l O " r w variable x(k+ 1) is calculated by the observer as follows:
+ 9.895 x 10- 6 7> 7 3 (kJ/kg). (8) x(k + 1)= - o j (k)x(k) + {hi (k)-d1 (k)b0(k)} U(k)
These curves are illustrated in Fig. 3. -r,(k){QF(k)-x(k)-b0(k)U(k)}. (15)
3.2 Adaptive Control Part. The controlled process, which The parameters (rx(k), r2(k), ri(k)) of the observer-and-
consists of the evaporator and the working fluid pump, is regulator in the above equations are
assumed to be formulated by a pulse transfer function of the /-,(/:) = (X, +X 2 + a, (k) - 1
first order as
QF(k) = {(b0 + blz-l)/(l + alz-i)}U(k) + W(k) (9) -b0(k)r2(k))/{bl(k)-al(k)bQ(k)}, (16)
where the process input U(k) is the input signal for the work- /• 2 (*) = (X,+X 2 + X 1 X 2 - 1 ) / { S 0 ( * ) - S 1 ( * ) ) , (17)
ing fluid pump, and the process output QF(k) is the heat rate r3(k)=n + di(k) (18)
of working fluid; W{k) represents a process disturbance, k
the kth stage of the sampling period, z" 1 a backwards shift where Xj and X2 are poles of the regulator, and n is a pole of
operator, and {ql ,b0,bx) unknown parameters of the process. the observer. These poles must be selected appropriately by
The adaptive control algorithm comprises combinations of a the designer beforehand. Since the controller parameters
recursive identification algorithm and an easy-to-control r\(k), r2(k), and r3(k) are calculated using the estimated
algorithm as follows: the unknown parameters of the process parameters of the process p(k), the control algorithm can
are estimated on-line by use of the identification algorithm, adapt properly to the process even under varying conditions.
and an appropriate process input U(k) is determined by the (c) Initial Condition. The initial value of the estimated
control algorithm using the estimated parameters (<J,, b0, bx) parameter vector is necessary as the initial condition for the
thus obtained. identification algorithm. Before digital control commences,
the initial value is calculated using input and output data
(a) Identification Algorithm. We adopt the exponential based on the exponential weighted least squares of batched
weighted least squares (EWLS) method [9] as the identifica- type as follows:
tion algorithm. The estimated parameters of the process are o o
given by
P= E p-'z(i)QF(i)/ £ p-'z(i)zU)T. (19)
r
p(k) = [dl(k)b0(k) £,W] 1 = 1-N i=l-N
This algorithm is equivalent to the algorithm of an on-line
= {I-K(k) z(k) }p(k- \) + K(k) QF(k), (10)
type, equations (10-12).
C(k) = {I-K(k) z(k)}C(k- l)/p, (11) 3.3 Input Signal Characteristics Part. The relation be-
T tween the pump input voltage Vv and the working fluid flow
K(k) = C{k- \)z(k)/{z{k) C{k- \)z(k) + p) (12)
rate U (the process input) is determined by a polynomial equa-
where p(k) is the estimated parameter vector. Both the
tion based on the least squares method as
estimated parameter vector p(k- 1) and the estimation matrix
Vy= - 6.000 X 10~2 + 6.286 x 10° U-l. 606 X 10~' t/2 (V).
C(k- 1) on the right side of the above equations have already
been determined in the previous stage k-l. The input-output (20)
vector z(k) is the column vector whose elements are the pro- This curve is illustrated in Fig. 4. The characteristics of the
cess input and the process outputs as amplifier used here are expressed as
z(k) = [QF(k- 1) U(k) U(k~ l)]7". (13) Vf, = 0.667 VV(V). (21)

Journal of Solar Energy Engineering MAY 1986, Vol. 108/113

Downloaded From: http://solarenergyengineering.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jseedo/28189/ on 04/06/2017 Terms of Use: http://www.asme

 = 500

400 400

300 300

200 200

a 100 100

0 100 200 300 400 0 100 300 200400

Time isi Time is]
Fig. 5 Input and output data fo the controlled process for off-line Fig. 7 The process input and output in the first control experiment
identification (constant conditions)

500 - 0. 300
" I , MM,
'"' "|l" I " " I " "

„ 400 o- "i - 200 c?"

-0.1 - 100
300 3 3 °i
-0.2 " -
Bo
200
M M
-0.3, I I M I M N I ; I I I I I M I I I I M I M M I M I
-100
^ 0 100 200 300 400
100
Time Isi
Fig. 8 The estimated parameters in the first control experiment (cons-
LnJn I tant conditions)
0 100 200 300 400
Time is)
Fig. 6 Model output for the process
flow rate was 20 t/h (5.56 kg/s). Other conditions for the ex-
periment were as follows:
4 Experimental Results • warm water temperature TH, 30 °C
We investigate an off-line identification result and three on- • cold water flow rate mc, 80 t/h (22.22 kg/s)
line digital control results for the "Imari 2 " OTEC plant. In • cold water temperature Tc, 10 °C
the experiments, warm fresh water heated by a boiler and cold • sampling period, 5 s
fresh water cooled by a refrigerating machine were used in- • number of control stages N, 90
stead of warm surface seawater and cold subsurface seawater. • poles of the regulator \i X2, 0.7 0.7
The turbine-generator was disconnected from the OTEC plant • pole of the observer JX, 0.5
for the purpose of the experiments. • index of EWLS p, 0.95
• initial value p(l), [-0.115 -39.11 252.20] T
(a) Off-Line Identification Result. Figure 5 illustrates the » initial value C(l), diag (0.0100, 1752, 1752) (diag:
input and output data of the controlled process for the off-line diagonal matrix)
identification. The sampling period was 5 s and the sample • initial value x(l), 160 Mcal/h (186 kW)
number was 90. The process inputs U(k) were given by ran-
dom numbers drawn from Gaussian distribution with a mean • initial condition ^ (Q F (/) - QD (/)), 250 Mcal/h
of 1.53 having a variance of 0.28. The estimated parameters / = — oo

using the input and output data based on the exponential

weighted least squares method (equation 19) are calculated as
r Figure 7 shows the input U(k) and the output QF(k) of the
p= [-0.115 -39.11 252.20] .
To verify the validity of the model (equation 9), one-step-
ahead model output was used
QF(k) = -a1QP(k-l) + b0U(K) + blU(k-l).
As seen from QF(k) in Fig. 6, the model output was approx-
imately equal to the true output QF(k) in Fig. 5, which allows
us to prove the validity of the model (equation 9). We adopted
the estimated parameter vectorp as for the initial valuep(\) of
the on-line identification algorithm in the following
experiments.
(b) Control Result {Constant Conditions). In the first con-
trol experiment, we investigated the digital control of working
fluid flow rate under constant conditions; the desired output
value QD(k) was 250 Mcal/h (291 kW) and the warm water
(291 kW)
(22)
digital control. As seen from Fig. 7, the output QF(k) ap-
proached the desired output value QD(k) in 80 s and main-
tained its value. The estimated parameters p(k) are shown in
(23) Fig. 8; they keep almost constant values for the whole control
interval.
(c) Control Result (Variable Desired Output Value). In the
second control experiment, the desired output value QD(k)
was changed from 250 Mcal/h (291 kW) to 300 Mcal/h (349
kW) at a time of 250 s. Other conditions were the same as
those adopted for the previous experiment. Figure 9 shows the
input U(k) and the output QF(k), and Fig. 10 illustrates the
estimated parameters p(k). It may be seen that the process
output QF(k) behaves appropriately according to the variable
desired output value Qo(k), and that the estimated
parameters adapt to the system under the changed conditions.

114/Vol. 108, MAY 1986 Transactions of the ASME

Downloaded From: http://solarenergyengineering.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jseedo/28189/ on 04/06/2017 Terms of Use: http://www.asme

 500

400

300

o 200

V 100-

0,
100 200 300 400
Time is)
Fig. 9 The process input and output in the second control experiment
(variable desired output value)

a ° MM|MM| ill nil | i i ii H I ,M i | i i i i | i i n 300 o 100 200 300 400

Time ts)
Fig. 13 The temperatures around the evaporator (a) and around the
o- "i
200
condenser (b)
-
s -o.i 100
"i
I i n i|ii n |i i u | i in III 30
V. -0.2 " b0 - 0 2

11111111111 111111 11111 ,,,,1,1, mc

^ -0.3, I M l l l M l l 100
0 100 200 300 400 ms m?
/\ - 20
Time isi
Fig. 10 The estimated parameters in the second experiment (variable U
desired output value) -

500 mn

400
,,,!,, IMIIIIIIII I I I H I I I M Illll
100 200 300 400
300 Time is)
Fig. 14 The mass flow rates
200

600 ~ r . i i | i l l , I I M | M , | i n , M I | I M I , I Ml III

- 500 Qc
- QH -
200 300 /
u. 400 ^ - V N V A -
Time isi ^y^<.
Fig. 11 The process input and output in the third experiment (variable 300 r QF

warm water flow rate)

3 0.1 300
200

100
j jps -

0 200
CD
b_ _l_Lllll IIIMMIII IIIMM! ,,!,,,,! I l l l l l l
CD
100 200 300 400
| -o.i 100 „ Time Isi
a
o_ Fig. 15 The heat rates
"a -n P 2 - o •e.
b0 =
o
E constant at 300 Mcal/h (349 kW). Other conditions were the
LLJ U 100 200 300 400 -100 same as those for the previous experiment. Figure 11 shows
the input U(k) and the output QF(k), and Fig. 12 illustrates
Time is)
the estimated parameters p (k). It may be seen that the process
Fig. 12 The estimated parameters in the third experiment (variable
output QF(k) can maintain the desired value shortly after the
warm water flow rate)
conditions have been changed.
(d) Control Result (Variable Warm Water Flow Rate). In (e) Related Values. Related values for the third control ex-
the third control experiment, the warm water flow rate value periment are shown in Figs. 13-15. Figure 13 shows the
mH was changed from 40 t/h (11.11 kg/s) to 20 t/h (5.56 kg/s) temperatures around the evaporator and the condenser. The
at a time of 250 s, and the desired output value QD(k) kept mass flow rates mH, mc, mF, ms for the water and the work-

Journal of Solar Energy Engineering MAY 1986, Vol. 108/115

Downloaded From: http://solarenergyengineering.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jseedo/28189/ on 04/06/2017 Terms of Use: http://www.asme

 ing fluid are shown in Fig. 14, and the heat rates QH, Qc, QF,
Qs are illustrated in Fig. 15. The heat rates were calculated by
the following formula:
Q„ = CpmH(THI-THO), (24)
Qc = Cpmc(Tco-TCI), (25)
QF = mF{h(Tvo)-h(TFI)}, (26)
Qs = ms{h(TVI)-h(TFo)}. (27)
The specific heat Cp is 4.186 kJ/kgK. The specific enthalpies
of working fluid vapor h (TVI ),h( Tv0) were calculated by the
polynomial (equation 7), and those of working fluid liquid
h(TFI), h(TF0) were calculated by the polynomial (equation
8). These curves in Figs. 13-15 can hopefully provide some
useful information for hardware designs of the OTEC plant in
the future.
5 Discussion
(a) Significance. The innovative feature of this study was
the construction of a digital control system and the implemen-
tation of the digital control of working fluid flow rate for the
real pilot OTEC plant. The process output of the digital con-
trol behaved appropriately according to a desired output value
under changing conditions by use of an adaptive control
algorithm of the digital control side programmed in the com-
puter. The advantages of using digital control in general are
summarized as follows:
(/) Any adaptive control algorithm can be selected by
changing the program of the adaptive control part, whereas
for analog type control, such as the PID controller, this is
rather difficult to do once the controller has been attached to
the plant.
(//) Linearity for detection characteristics and for input
signal characteristics is not necessarily requisite. The
polynomial equations for the characteristics, determined by
the least squares method, need only be programmed in the
computer.
(b) Remarks
(i) The heat rate of working fluid at the evaporator was
calculated by equation (6). A more reliable calculation might
be possible by using mv (working fluid vapor flow rate) in-
stead of mF (working fluid liquid flow rate) adopted here.
(ii) The process model (equation 9) used here had the
structure of the first order. A more appropriate order of the
process model might be found by using another technique,
such as AIC (Akaike's Information Criterion) [11].
(Hi) The values of the poles A,, X2, \i in the control
algorithm and the value of the index p in the identification
algorithm were selected a priori by computer simulation
before the control experiments. More appropriate values
might be determined through further experimental
investigation.
(iv) For an adaptive control theory, we adopted the ex-
116/Vol. 108, MAY 1986
Downloaded From: http://solarenergyengineering.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jseedo/28189/ on 04/06/2017 Terms of Use: http://www.asme
ponential weighted least squares method as an identification
algorithm and the observer-and-regulator as a control
algorithm. Other adaptive control theories, such as the
MRACS (Model Reference Adaptive Control System) [6],
could also be applied simply by changing the program of the
adaptive control part.
(c) Future Problems
(i) Remarks described in the previous subsection will be
considered in order to achieve the more precise digital control
of working fluid flow rate.
(ii) The "Imari 2 " OTEC plant has been designed based
on steady-state analysis. By using dynamic characteristics,
such as the curves shown in Figs. 13-15, the hardware designs
of the OTEC plant will be reconsidered in order to achieve
more efficient performance.
(Hi) Using adaptive control theory for multi-input and
multi-output systems, the control of the seawater flow rates as
well as the working fluid flow rate will be accomplished.
(iv) The total system control of the OTEC plant will be
established to provide generating power at a desired value.
(v) The OTEC plant in Imari is now planned to be con-
trolled by microcomputer from Saga University, approximate-
ly 40 km away, using ordinary public telephone lines.
Acknowledgments
The authors are grateful to Mr. T. Nakaoka (research assis-
tant), Mr. S. Nishida (research assistant), and Mr. Oishi
(postgraduate student, now with Mitzubishi Heavy Industry)
of Saga University for their useful comments and help in the
experiments, and to Dr. G. Jember of Saga University for
revising the English in this paper.
References
1 D'Arsonval, J. A., "Utilization des forces naturelles: Avenir de l'elec-
tricite," La Revenue Scientifique, Sept. 17, 1881, pp. 370-372.
2 Uehara, H., Kusuda, H., Monde, M., Nakaoka, T., Masuda, T., and
Nakahara, M., "Model of Ocean Thermal Energy Plant—Shiranui 3 , " Proc.
5th OTECConf., 1978, pp. V 117-146.
3 Uehara, H., Kusuda, H., Monde, M., Nakaoka, T., and Sumitomo, H.,
"Shell-and-Plate-Type Heat Exchangers for OTEC Plants," ASME JOURNAL OF
SOLAR ENERGY ENGINEERING, Vol. 106, Aug. 1984, pp. 286-290.
4 Owens, W. L., "OTEC Plant Response and Control Analysis," ASME
JOURNAL OF SOLAR ENERGY ENGINEERING, Vol. 104, Aug. 1982, pp. 208-215.
5 Astrom, K. J., Borisson, V., Ljung, L., and Wittenmark, B., "Theory
and Applications of Self-Tuning Regulators," Automatica, Vol. 13, 1977, pp.
457-476.
6 Landau, I. D., " A Survey of Model Reference Adaptive Techni-
ques—Theory and Applications," Automatica, Vol. 10, 1974, pp. 353-379.
7 Uehara, H., "Introduction to OTEC," Ohm, (in Japanese), 1982.
8 JAR., "Thermophysical Properties of Refrigerants (R22)," (in Japanese),
1982.
9 Nakamura, M., "Relationship Between Steady State Kalman Filter Gain
and Noise Variances," Int. J. Systems Sci., Vol. 13, No. 10, 1982, pp.
1153-1163.
10 Furuta, K., "Digital Control and Control Theory," Computwl, (in
Japanese), No. 2, 1983, pp. 16-25.
11 Akaike, H., "A New Look at the Statistical Model Identification," IEEE
Trans. Autom. Control, Vol. AC-19, No. 6, 1974, pp. 716-723.
Transactions of the ASME