Hue University Journal of Science: Economics and Development

pISSN: 2588–1205; eISSN: 2615–9716

Vol. 128, No. 5B, 2019, pp. 13–21; DOI: 10.26459/hueuni-jed.v128i5B.4375

USING IMPROVED GREY FORECASTING MODEL TO

ESTIMATE THE ELECTRICITY CONSUMPTION DEMAND IN
VIETNAM

Phan Van Thanh*

Scientific management & International Affairs, Quang Binh University

312 Ly Thuong Kiet Street, Dong Hơi city, Quang Binh Province

Abstract: On the basis of the grey prediction models, this study uses the previous data (from 1980 to 2014)
from the website of the World Bank and applies two algorithm models to forecast the electricity
consumption in Vietnam. The simulation results show that Fourier Residual Modified GM (1, 1)
(abbreviated as FRMGM (1, 1)) is an effective model with an average accuracy of prediction at 99.13%.
Therefore, the FRMGM (1, 1) model is strongly suggested for forecasting the electricity consumption
demand in Vietnam.

Keywords: electricity consumption demand, GM (1, 1); FRMGM (1, 1), Vietnam

1 Introduction

Worldwide energy consumption is rising fast because of the increase in human population,
continuous pressures for better living standards, emphasis on large-scale industrialization in
developing countries, and the need to sustain positive economic growth rates. Given this fact, a
sound forecasting technique is essential for accurate investment planning of energy production/
generation and distribution. In Vietnam, the electricity consumption forecasting plays a
significant role in strategic planning of an electric utility company since many activities need to
be planned in advance such as location and construction of new substations, creating new
transmission and distribution networks, improving existingsystems, and/or construction of new
power generation plants.

The grey system theory, established in the 1980s by Deng [1], is a quantitative method
dealing with grey systems that are characterized by both partially known and partially
unknown information [2–5]. As a vital part of the grey system theory, grey forecasting models
with their advantages in dealing with uncertain information by using as few as four data points
[6, 7]. These models have been successfully applied in various fields such as tourism [8, 9],
energy [10, 11], financial and economic [12–14], and IC industry [15].

Because the grey forecasting model has two major advantages, namely fewer samples
requirement and simple computation, this study tries to apply two models to estimate the
* Corresponding:thanhkem2710@gmail.com
Submitted: July27, 2017; Revised: March 09, 2018; Accepted: April 27, 2018
Phan Van Thanh Vol. 128, No. 5B, 2019

Vietnamese electricity consumption. The results will be helpful for the top managers in
formulating policies as well as orienting development in the power sector. The remaining of this
paper is organized as follows. In section 2, the concept of the grey theory is presented and the
fundamental function of traditional grey forecasting model “(GM (1, 1)” and modified GM (1, 1)
by Fourier series are shown. On the basis of the fundamental function of GM (1, 1) and FRMGM
(1, 1), the empirical results are discussed in section 3. Finally, section 4 concludes this paper.

2 The concept of grey prediction

2.1 GM (1, 1) algorithm

GM (1, 1) is the basic model of grey forecasting modelling, a first-order differential model
with one input variable which has been successfully applied in many fields of research. It is
obtained as in the following procedure.

Step1: Let raw matrix X ( 0) stand for the non-negative original historical time series data
 
X (0)  x(0) (ti ) , i  1, 2,..., n

where x ( 0) (ti ) is the value at time ti, and n is the total number of modelling data points.

Step 2: Construct X (1) by one time accumulated generating operation (1-AGO), which is
 
X (1)  x (1) (ti ) , i  1, 2,..., n

where
k
x1 (tk )  x (0)
(ti ), k  1, 2,..., n
i 1

Step 3: X (1) is a monotonic increasing sequence which is modelled by the first-order

linear differential equation
(1)
dX
 aX (1)  b
dt

where parameter “a” is called the developing coefficient and “b” is named the grey input.

Step 4: In order to estimate parameter “a” and “b”, Eq. (4) is approximated as:

X (1) (tk )
 aX (1) (tk )  b
dtk

where

X (tk )  x (tk )  x (tk 1 )  x (tk )

(1) (1) (1) (0)

14
Jos.hueuni.edu.vn Vol. 128, No. 5B, 2019

tk  tk  tk 1

If the sampling time interval is units, then let t k  1 , using

z (tk )  px (tk )  (1  p) x (tk 1 ) , k  2,3,.., n

(1) (1) (1)

to replace X (1) (t k ) in Eq. (1.5), we obtain

(tk )  az (tk )  b , k  2,3,..., n

(0) (1)
x

(1)
where z (tk ) in Eq. (8) is the termed background value, and p is the production coefficient of
the background value in the range of (0, 1), which is traditionally set to 0.5.

Step 5: From Eq. (9), the value of parameter “a” and “b” can be estimated using the least-
square method. That is

a  1 T
b   ( B B) B Yn
T

 

where

 
  z (1) (t ) 1
 2 

B   z ( t3 )
(1)
1
 
...... .....
 (1) 
  z (t n ) 1 

and
T
Yn   x(0) (t2 ), x(0) (t3 ),..., x(0) (tn )) 

Step 6: The solution of Eq. (4) can be obtained after parameter “a” and “b” has been
estimated. That is

 b  a ( tk t1 ) b
xˆ (1) (tk )   x (0) (t1 )  e   , k  1, 2,3,...
 a a

Step 7: Applying an inverse accumulated generating operation (IAGO) to xˆ

(1)
(t k ) , the
predicted datum of x ( 0) (t k ) can be estimated as:

 xˆ (0) (t1 )  x (0) (t1 )
 (0) , k  2, 3,...

 xˆ (tk )  xˆ (tk )  xˆ (tk 1 )
(15)
(1) (1)

15
Phan Van Thanh Vol. 128, No. 5B, 2019

2.2 Fourier Residual Modified GM (1, 1) algorithm

The overall procedure to obtain the modified model is as follows:

Let x be the original series of m entries and v is the predicted series (obtained from GM
(1, 1)). On the basis of the predicted series v , a residual series named  is defined as

   (k ) , k  2, 3,...m
where
 (k )  x(k )  v(k ) , k  2, 3,...m

According to the definition of the Fourier series, the residual sequence of GM (1, 1) can be
approximately expressed as
 2 i  2 i
  a 
Z
1   1,2,3,., m
ˆ ( k )  a(0)  cos  ( k )   bi sin  (k )  , k
 m 1   m 1 
i
2 i 1

m 1
where Z  ( )  1 is called the minimum deployment frequency of the Fourier series
2
[13] and only takes an integer number; therefore, the residual series is rewritten as:
  P.C
where
1  2 1   2 1   2  Z   2  Z 
2 cos  2  sin   2  ... cos  2  sin   2 
  m 1   m 1   m 1   m 1 
1  2 1   2 1   2  Z   2  Z 
P  2 cos  3  sin   3 ... cos  3  sin   3
 m 1   m 1   m 1   m 1 
... ... ... ... 
1  2 1   2 1   2  Z   2  Z 
 cos  m  sin   m  .... cos  m  sin   m 
 2  m 1   m 1   m 1   m 1 
k  1,2,3,., m
and
C   a0 , a1 , b1 , a2 , b2 , ..., aZ , bZ 

Parameters a0, a1, b1, a2, b2, …,aZ, bZ are obtained by using the ordinary least-square
method (OLS) which results in the equation

P 
1
C 
T
P P
T
 T

Once the parameters are calculated, the modified residual series is then achieved on the
basis of the following expression:
 2 i  2 i
 a 
Z
1 
ˆ ( k )  a(0)  cos  ( k )   bi sin  (k ) 
 m 1   m 1 
i
2 i 1

16
Jos.hueuni.edu.vn Vol. 128, No. 5B, 2019

From the predicted series v and ˆ , the Fourier modified series v̂ of series v is
determined by:
vˆ  vˆ1 , vˆ2 , vˆ3 ,...., vˆk ,..., vˆn 

where
 vˆ1  v1
vˆ  
vˆk  vk  ˆk (k  2, 3,..., m)

2.3 Valuation performances

Regarding the evaluation performance of the volatility model for forecasting, there are
some common approaches, including the root of mean square error, mean absolute error, and
mean absolute percentage error (MAPE).

This study uses MAPE [16] to identify the grey prediction models with good
performance; small MAPE is taken to indicate good forecasting performance. MAPE is defined
as follows:
( k )  xˆ
n (0) (0)


1 x (k )
MAPE  (0)
 100%
n k 2 x (k )

Where x ( 0) (k ) indicates the actual value in time period k; xˆ ( 0) (k ) indicates the forecast
value in time period k.

And the grade of MAPE is divided into four levels [17]. More detailed shown in Table 1 .

Table 1. Grades of MAPE

MAPE ≤10% 10–20% 20–50% >50%

Grade levels Excellent Good Qualified Unqualified
In addition, indicator  is used to evaluate the predicted accuracy of the forecast model,
higher  is taken to indicate a good forecasting performance. This is defined as follows:
  100  MAPE

3 Data and empirical results

The data on electricity consumption from 1980 to 2014 in Vietnam were obtained from
the World Bank on December 12th 2016 [18]. From the historical data, parameters a and b are –
0.1323 and 0.7783 calculated through the algorithms of the GM (1, 1) model (section 2.1). The
fundamental of GM (1, 1) for electricity consumption is found as follows:

17
Phan Van Thanh Vol. 128, No. 5B, 2019

(1)  0.7783  0.1323( k 1) 0.7783

xˆ ( k )  3.29  
 0.1323 

e
0.1323

The residual series attained form GM (1, 1) is then modified with the Fourier series,
which results in the modified model FRMGM (1, 1) according to the algorithms stated in section
2.2. The evaluation index of GM (1, 1) and FRMGM (1, 1) is summarized in Table 2.

Table 2. Forecasted value of the electricity consumption (Units: kWh per capita)

Actual Forecasted value Forecasted value

Years % error % error
value by GM(1,1) by FRMGM(1,1)
1980 54.562 54.562 0.00 54.562 0.00
1981 55.115 39.942 27.53 53.756 2.47
1982 57.859 44.639 22.85 59.218 2.35
1983 57.541 49.889 13.30 56.182 2.36
1984 67.011 55.756 16.80 68.370 2.03
1985 70.344 62.314 11.42 68.985 1.93
1986 73.412 69.642 5.14 74.771 1.85
1987 80.146 77.833 2.89 78.787 1.70
1988 87.176 86.987 0.22 88.535 1.56
1989 94.158 97.217 3.25 92.799 1.44
1990 98.096 108.650 10.76 99.455 1.39
1991 102.004 121.429 19.04 100.645 1.33
1992 105.464 135.710 28.68 106.823 1.29
1993 116.334 151.670 30.37 114.975 1.17
1994 134.473 169.508 26.05 135.832 1.01
1995 159.302 189.444 18.92 157.943 0.85
1996 186.996 211.724 13.22 188.355 0.73
1997 210.613 236.624 12.35 209.253 0.65
1998 241.173 264.453 9.65 242.532 0.56
1999 261.369 295.555 13.08 260.010 0.52
2000 295.037 330.315 11.96 296.396 0.46
2001 335.345 369.163 10.08 333.986 0.41
2002 387.037 412.580 6.60 388.396 0.35
2003 443.074 461.103 4.07 441.715 0.31
2004 493.624 515.332 4.40 494.983 0.28
2005 579.922 575.940 0.69 578.563 0.23

18
Jos.hueuni.edu.vn Vol. 128, No. 5B, 2019

Actual Forecasted value Forecasted value

Years % error % error
value by GM(1,1) by FRMGM(1,1)
2006 655.326 643.675 1.78 656.685 0.21
2007 738.472 719.377 2.59 737.113 0.18
2008 813.159 803.982 1.13 814.518 0.17
2009 917.396 898.537 2.06 916.037 0.15
2010 1034.607 1004.212 2.94 1035.966 0.13
2011 1115.498 1122.316 0.61 1114.139 0.12
2012 1219.795 1254.310 2.83 1221.154 0.11
2013 1290.983 1401.828 8.59 1289.624 0.11
2014 1439.156 1566.695 8.86 1440.515 0.09
MAPE (%) 10.13% 0.87%
  100  MAPE (%) 99.13% 99.13%
Forecasting power Qualified Excellent

Table 2 clearly shows that the FRMGM (1, 1) model is better than the GM (1, 1) model in
this case with the MAPE and is 0.87% and 99.13%, respectively. Therefore, this study suggests
that the FRMGM (1, 1) model should be used for the estimation of the electricity consumption
demand in the future. The forecasted value in 2018 to 2020 is shown in Table 3.

Table 3. Forecasted value by FRMGM (1, 1)

Forecasted value Electricity consumption (kWh per capita)

2018  2456.85
2019  2738.37
2020  3058.10

Table 3 shows that the forecasting values in 2019 and 2020 will be over 2738 and 3058
kWh per capita, respectively. This figure indicates that the demand for electricity consumption
in Vietnam will grow significantly in the future. This is a reference for the managers in the
power sector to make a good decision in planning and development.

4 Conclusion

The electricity sector is an important industry in the socio-economic development. In

Vietnam, the rapid development of the economy leads to the increasing demand for electric
consumption. Through simulation, this study found that FRMGM (1, 1) is the fitting model in
order to forecast the electricity consumption demand in Vietnam with an accuracy of 99.13%.
On the basis of this result, this study strongly suggests that FRMGM (1, 1) is an effective tool to

19
Phan Van Thanh Vol. 128, No. 5B, 2019

estimate the electricity consumption demand in the future. Further, the results of this study can
be a good reference for the policymakers to make a good decision in the planning and
development of the power sector. Due to the data limitations, this study just compares the
forecasted and actual values during the period from 1980 to 2014. Future research could also
utilize different models of grey forecasting models such as Grey Verhuslt model, the GM (2, 1)
model to compare with the proposed model in the current study.

References
1. J. L. Deng (1982), Control problems of grey systems, Systems and Control Letters, 1, (5), 288–
294, doi.org/10.1016/S0167-6911(82)80025-X.
2. S. Emil and D. Camelia (2011), Complete analysis of bankruptcy syndrome using grey
systems theory, Grey Systems: Theory and Application, 1, 19–32, doi:
10.1108/20439371111106704.
3. W. Hong and C. Fuzhong (2011), The application of grey system theory to exchange rate
prediction in the post-crisis era, International Journal of Innovative Management, 2, (2), 83–89,
link online: www.ismeip.org/IJIMIP/contents/imip1122/10.pdf.
4. H. J. Chen (2011), Application of grey system theory in telecare, Computers in Biology and
Medicine, 41 (5), 302–306, doi: 10.1016/j.compbiomed.2011.03.007.
5. J. J. Guo, J. Y. Wu, and R. Z. Wang (2011), A new approach to energy consumption
prediction of domestic heat pump water heater based on grey system theory, Energy and
Buildings, 43 (6),1273–1279, doi: 10.1016/j.enbuild.2011.01.001.
6. L. Yi and L. Sifeng (2004), A historical introduction to grey system theory, IEEE international
conference on system, man and cybernetics, 2403–2408, doi: 10.1109/ICSMC.2004.1400689.
7. L. Sifeng, J. Forrest and Y. Yingjie (2011), A brief introduction to grey system theory, IEEE Grey
Systems and Intelligent Services (GSIS) IEEE International Conference on, doi:
10.1109/GSIS.2011.6044018.
8. Y. L Huang and Y. H Lee (2011), Accurately forecasting model for the stochastic volatility
data in tourism demand, Modern economy, 2 (5), 823–829, doi: 10.4236/me.2011.25091.
9. F. L Chu (1998), Forecasting Tourism Demand in Asian-Pacific Countries, Annual of Tourism
Research, 25 (3), 597–615, doi: 10.1016/S0160-7383(98)00012-7.
10. C. C Hsu and C. Y Chen (2003), Application of improved grey prediction model for power
demand forecasting, Energy Conversion and management, 44, 2241–2249,
doi: 10.1016/S0196-8904(02)00248-0.
11. J. Kang and H. Zhao (2012), Application of Improved Grey Model in Long-term Load
Forecasting of Power Engineering, Systems Engineering Procedia, 3, 85–91,
doi: 10.1016/j.sepro.2011.11.012.
12. E. Kayacan, B. Ulutas, and O. Kaynak(2010), Grey system theory-based models in time
series prediction, Expert Systems with Applications, 37,1784–1789,
doi: 10.1016/j.eswa.2009.07.064.

20
Jos.hueuni.edu.vn Vol. 128, No. 5B, 2019

13. M. Askari and H. Askari (2011), Time Series Grey System Prediction-based Models: Gold
Price Forecasting, Trends in Applied Sciences Research, 6, 1287–1292,
doi: 10.3923/tasr.2011.1287.1292.
14. G. Gemmill (1986), The forecasting performance of stock options on the London Traded
Option Markets, Journal of Business Finance and Accounting, 13 (4), 535–546,
doi: 10.1111/j.1468-5957.1986.tb00516.x.
15. L. C. Tsai and Y. S. Yu (2004), Forecast of the output value of Taiwan's IC industry using the
Grey forecasting model, International Journal of Computer Applications in Technology, 19,
(1), 23 – 27, doi: 10.1504/IJCAT.2004.003657.
16. S. Makridakis (1993), Accuracy measures: Theoretical and practical concerns, International
Journal of Forecasting, 9, (4), 527–529, doi.org/10.1016/0169-2070(93)90079-3.
17. C. D Lewis (1983), Industrial and business forecasting methods, Journal of Forecasting, 2 (2),
194–196, doi.org/10.1002/for.3980020210.
18. Website of world bank (2017), Retrieved from
https://data.worldbank.org/indicator/EG.USE.ELEC.KH.PC?locations=VN. Accessed date:
12th, Dec, 2016.

21