Department of Electrical Engineering, Faculty of Engineering Final Year Project External Defense
Kano University of Science and Technology Session 20202021
ANN AND GIS-ASSISTED METHODOLOGY FOR WIND RESOURCE
ASSESSMENT (WRA) IN SARAWAK
Ijeoma Ademola*, Salisu Muhammad Lawan
*UG35/ELEE/1021
Abstract: Wind energy is reliable, clean sources and has emerged TNN wind speed prediction model and GIS- based methodology is
as the safest and the fastest developing renewable energy in the proposed in Sarawak. In order to study the complex, dependent of
world. It has an insignificant environmental impact compared to wind speed in the non-monitored locations.
other energy sources. In Sarawak, Malaysia, wind resource differs
depending on the location. Inadequate numbers of wind stations II. MATERIALS AND METHODS
are the major hindrances that decelerate the growing of green A parametric feedforward neural network model was designed,
energy in the region. Site selection is a crucial issue for potential by considering topographical effects (terrain and roughness class)
investors and policy makers. This paper examines the spatial and meteorological data. The TNN model has nine input
distribution and the amount of potential wind power and energy parameters and one target input. The number of neurons in the
densities for wind energy production and suitable locations in hidden layer varies from 7-202 with a step of 1. Logsigmoid and
Sarawak. A geographical Information System (GIS) assisted Purelin activation functions are used in order to obtain a
methodology, which includes wind speed, power and energy differential function between the inputs and output. Summary of
densities using the existing wind station and based on the the design is shown in Table 1,
developed topographical neural network (TNN) prediction models, Prior to the training, all the data were scaled in a range of [-1, 1].
was used. Kriging interpolation is used for simple interpolation of A supervised learning was used to train the model by means of
data between locations. The results show that the northeast, Levenberg-Marquadt back propagation algorithm. The training
northwest and coastal regions have better prospects of wind energy. was carried out continuously by updating the weights for a single
The studied GIS methodology can be applied for identification of epoch, until the error difference between the inputs and targets
the most suitable locations for wind energy harvesting. The data is neglble. Once the training is completed, the weights and
developed maps can further be used in micro-siting and economic biases are obtained, which leads to the development of
evaluation analysis. mathematical prediction models. The suitability of the model was
judged based on two gods of fit, mean absolute percentage error
I. INTRODUCTION and correlation coefficient.
Exponential growth, increase in the utilization of renewable For the wind energy mapping, a base map of Sarawak was
energy source such as wind may offer several environmental and digitized so as to obtain the shape files. The coordinates of each
economical advantages comparable with conventional energy station were converted to degree decimal unit World Geodtetic
sources polluting and generating green house gases Wind energy Sytem (WGS) 1984 was used to define the core dinate point and
as other forms of alternative energy may offer the promise of it was employed in generating the contour lines. This approach has
meeting energy demand in the direct, grid connected modes as been verified in many reported studies [9, 10].
well as standalone and remote applications.
Before a wind turbine is installed, wind energy resource Table 1. T-FFNN Design Summary
assessment (WERA) must be conducted. Different WRRE has Network Feedforward back propagation
been carried out around the world [1-4]. The studies can be Layer 3-layesr with Single hidden layer
categorized into, suitability of fitting wind speed data, analysis of
Transfer function Log Sigmoid and Purelin
wind power potential based on empirical and statistical modeling,
and geographical information system (GIS) based methodology, Neurons 7-202 with a step of 5
micro-siting performances and feed-in-tariff (FIT). In Malaysia, Epochs 0-1000 with a step of 1
the case is not left out. Several studies to explore different ideas Training algorithm LM
were carried out [5, 6]. In the case of Sarawak, little literature has
been identified. Development of the wind map for Sabah and
Sarawak and evaluation of wind and solar potential at five
III. RESULTS AND DISCUSSION
locations of Sarawak [7, 8], limited data points underscores the
The training for all the networks was performed until the
usability of wind map constructed by [7,], while statistical
modeling and wind power evaluation was not covered in [8]. optimum model was selected based on the MSE and computation
There is different between the current long-term data and time. It was found that 9-152-1 has lower MSE and faster
short-term data. Furthermore, extensive WERA must be conducted processing time of 4 minutes twenty-two second. Figure 1 shows
before micro siting is performed. Because wind speed varies, the error model function digram. The best validation performance
depend on the time, space and locations. Based on these reasons, a was 0.001007 at 997 epochs. The regression values varied from
0.9851-0.99541. These values fall within acceptable range a much
－1－
lower than this has been reported in [4]. Journal of Energy and Environment, vol. 4, 2013.
[9] K. Abdeladim, R. Romeo, and S. Magri, "Wind mapping of a region in
the north-east of Algeria," Renewable Energy, vol. 9, pp. 789-793, 1996.
[10] A. Sözen, E. Arcaklioǧlu, M. Özalp, and E. G. Kanit, "Use of artificial
neural networks for mapping of solar potential in Turkey," Applied
Energy, vol. 77, pp. 273-286, 2004.
Figure 1. Traning error function
A geographical, meteorological and topographical parameters
(latitude, longitude, altitude), and month of the year, temperature,
atmospheric pressure, relative humidity and developed terrain
elevation and roughness class were used as input for the developed
model to predict the monthly long-term wind speed from January
to December for 27 locations within Sarwak consisting of 8
ground stations and additional 19 locations where wind station is
not available. The model predicted monthly wind speed were
averaged to for a period of ten and five years respectively. GIS
method was applied to generate the wind atlas map. Based on the
developed maps, the results showed that the northeast, southwest
and coastal regions have better prospect of wind power, even
though other regions seems to posses low wind speed, a small
wind turbine of cut-in wind speed 1-1.5 m/s can be run for
small-scale application in rural and remote areas.
REFERENCES
[1] G. Bekele and B. Palm, "Wind energy potential assessment at four
typical locations in Ethiopia," Applied Energy, vol. 86, pp. 388-396,
2009.
[2] A. Keyhani, M. Ghasemi-Varnamkhasti, M. Khanali, and R.
Abbaszadeh, "An assessment of wind energy potential as a power
generation source in the capital of Iran, Tehran," Energy, vol. 35, pp.
188-201, 2010.
[3] T. Ramachandra and B. Shruthi, "Wind energy potential mapping in
Karnataka, India, using GIS," Energy conversion and management, vol.
46, pp. 1561-1578, 2005.
[4] M. Abbes and J. Belhadj, "Development of a methodology for wind
energy estimation and wind park design," Journal of Renewable and
Sustainable Energy, vol. 6, p. 053103, 2014.
[5] A. Albani and M. Ibrahim, "An Assessment of Wind Energy Potential
for Selected Sites in Malaysia Using Feed-in Tariff Criteria," Wind
Engineering, vol. 38, pp. 249-260, 2014.
[6] S. L. Muhammad, W. A. W. Z. Abidin, W. Y. Chai, A. Baharun, and T.
Masri, "Development of Wind Mapping Based on Artificial Neural
Network (ANN) for Energy Exploration in Sarawak," International
Journal of Renewable Energy Research (IJRER), vol. 4, pp. 618-627,
2014.
[7] A. Ab. Wahab, "Establishing Wind Map of Sabah and Sarawak,"
unpublishedUTM eprints 2002|.
[8] A. Jakhrani, A. Othman, A. Rigit, and S. Samo, "Assessment of Solar
and Wind Energy Resources at Five Typical Locations in Sarawak,"