ABSTRACT
Artificial neural network (ANN) modelling of wind energy profile of south eastern Nigeria for power generation was studied in this project. Five locations (Nsukka, Umudike, Owerri, Onitsha and Enugu) were considered whereas two locations (Nsukka and Umudike) were selected for the ANN modelling. The study investigated the most effective parameters in estimating generated power from observed wind speed and other climatological factors in the South eastern region of Nigeria as well as a comparative analysis between the results from ANN model to linear-model fits. The data used for this study was sourced from a wind anemometer reading at 2m hub height for Nsukka, Enugu State and Umudike in Abia State, whereas the data for Enugu, Owerri and Onitsha were extracted from previous works all-in South-Eastern Nigeria. Results showed that the minimum wind speed observed at the various locations were 2.605m/s for Nsukka, and 4.518m/s for Umudike while the maximum wind speeds were 5.501m/s for Nsukka, and 9.807m/s for Umudike respectively at the extrapolated height of 30m. The Weibull shape parameter k is between 13.12 to 32.02 for Nsukka and 13.4 to 28.68 for Umudike, while the scale parameter c is between 2.692m/s to 5.987m/s for Nsukka and 4.682m/s to 9.997m/s for Umudike. The annual mean power density for Nsukka and Umudike are 35.128 kWh/m2 and 147.39kWh/m2 respectively. The result also showed that model 3 (a combination of wind speed, pressure and temperature) returned the least error in prediction for Nsukka while model 2 (a combination of wind speed and pressure.) returned the least prediction error for Umudike. The coefficient of determination (R^2) of the optimal ANN model is 0.999 for both locations under study. The study thus concludes that the optimum ANN architecture for predicting the wind power available in the south east region varies depending on the location and climatological differences within the said location. Hence, a common set of predictors cannot be adopted for all locations. However, the study confirms that ANN has a low error rate in predicting available wind power. Thus, the study recommends that ANN should be employed in wind power prediction in South-Eastern Nigeria. The study thus developed an ANN-Regression equation for the predicting of available wind power density for Nsukka and Umudike.
TABLE OF CONTENTS
Title Page i
Declaration ii
Dedication iii
Certification iv
Acknowledgement v
Table of Contents vi
List of Tables ix
List of Figures x
Nomenclatures xi
Abstract xii
CHAPTER 1: INTRODUCTION
1.1 Background of Study 1
1.2 Statement of Problem 4
1.3 Aim and Objectives of Study 4
1.4 Scope of Study 5
1.5 Justification for the Study 5
CHAPTER 2: LITERATURE REVIEW
2.1 The Nigerian Power Sector 7
2.2 Renewable Energy Resources in Nigeria 8
2.2.1 Biomass 10
2.2.2 Solar energy 11
2.2.3 Hydro energy 12
2.2.4 Wind energy 12
2.3 Challenges for Renewable Energy Solutions in Nigeria 14
2.4 Wind Power Forecast 15
2.4.1 Forecasting models 16
2.5 Artificial Neural Network Fundamentals 21
2.5.1 Artificial neural network structure 21
2.5.2 Activation function 22
2.5.3 Training, validation, and testing datasets 22
2.5.4 Normalization of the data 22
2.5.5 Objective function and optimization process 23
2.5.6 Optimization procedure 24
2.5.7 Selecting the artificial neural network structure 25
2.6 Gap in Literature 26
CHAPTER 3: MATERIALS AND METHODS
3.1 Materials 27
3.2 Methods 27
3.2.1 Data sampling 27
3.2.2 Data pre-processing 27
3.2.3 Weibull wind model analysis 28
3.2.4 Most probable wind speed and wind speed carrying maximum energy 29
3.2.5 Extrapolation of wind speeds at different hub heights 29
3.2.6 Wind power density and estimation of wind energy 29
3.2.7 Designing of artificial neural networks models 30
3.3 ANN Model Development 33
3.3.1 ANN model 1: wind speed 33
3.3.2 ANN model 2: wind speed and pressure 34
3.3.3 ANN model 3: wind speed, pressure and temperature 34
3.3.4 ANN model 4: wind speed, pressure, temperature and humidity 35
3.3.5 ANN model 5: wind speed, temperature, pressure, humidity, air density 36
3.4 Wind Model Performance Accuracy Measures 37
3.5 ANN-Regression Equation Modelling 38
3.5.1 Analysing and interpreting the neural network for Nsukka 39
3.5.2 Analysing and interpreting the neural network for Umudike 39
CHAPTER 4: RESULTS AND DISCUSSION
4.1 Wind Characteristics Results for Study Locations 41
4.1.1 Extrapolating to 30m height 46
4.1.2 Wind power density 47
4.2 ANN Model Results 48
4.2.1 ANN model 1: wind speed 48
4.2.2 ANN model 2: wind speed and pressure 50
4.2.3 ANN model 3: wind speed, pressure and temperature 52
4.2.4 ANN model 4: wind speed, pressure, temperature and humidity 54
4.2.5 ANN model 5: wind speed, temperature, pressure, humidity and air density 56
4.3 Model Performance Evaluation 58
4.4 Determination of Optimal Model 59
4.5 ANN-Regression Equation Results 59
4.5.1 ANN-regression equation result for Nsukka 59
4.5.2 ANN-regression equation result for Umudike 62
CHAPTER 5: CONCLUSION AND RECOMMENDATIONS
5.1 Conclusion 66
5.2 Contributions to Knowledge 67
5.3 Recommendations 67
References 69
Appendices 75
LIST OF TABLES
2.1 Renewable energy potentials in Nigeria 9
4.1 Characteristic speeds and mean power density in Nsukka 41
4.2 Characteristic speeds and mean power density in Umudike 42
4.3 Characteristic speeds and mean power density in Enugu 43
4.4 Characteristic speeds and mean power density in Owerri 44
4.5 Characteristic speeds and mean power density in Onitsha 45
4.6 Extrapolated wind energy density and Weibull parameters at 30m for the locations 47
4.7 Model 1 summary 48
4.8 Model 1 Weibull and ANN predicted WPD values for the locations 49
4.9 Model 2 summary 50
4.10 Model 2 Weibull and ANN predicted WPD values for the locations 51
4.11 Model 3 summary 52
4.12 Model 3 Weibull and ANN predicted WPD values for the locations 53
4.13 Model 4 summary 54
4.14 Model 4 Weibull and ANN predicted WPD values for the locations 55
4.15 Model 5 summary 56
4.16 Model 5 Weibull and ANN predicted WPD values for the locations 57
4.17 Performance evaluation of the different ANN models 58
4.18 Input weights and input layer biases for Nsukka 60
4.19 Hidden layer weights and hidden layer biases for Nsukka 60
4.20 Training dataset result for Nsukka 60
4.21 Testing dataset result for Nsukka 61
4.22 Input weights and input layer biases for Umudike 63
4.23 Hidden layer weights and hidden layer biases for Umudike 63
4.24 Training dataset result for Umudike 64
4.25 Testing dataset result for Umudike 65
LIST OF FIGURES
Page
2.1 Energy mix in Nigeria 10
2.2 Schematic of a single-hidden-layer ANN model 21
3.1 Schematic of the designed artificial neural network 31
3.2 ANN diagram for model 1 33
3.3 ANN diagram for model 2 34
3.4 ANN diagram for model 3 35
3.5 ANN diagram for model 4 36
3.6 ANN diagram for model 5 37
3.7 Schematic of Nsukka ANN layout 39
3.8 Schematic of Umudike ANN layout 39
4.1 Mean monthly wind speed for the locations 46
4.2 Monthly WPD for the locations 47
4.3 Model 1 independence variable importance 49
4.4 Model 2 independence variable importance 51
4.5 Model 3 independence variable importance 53
4.6 Model 4 independence variable importance 55
4.7 Model 5 independence variable importance 57
4.8 Regression Plot of the training, testing and the full data for Nsukka 62
4.9 Regression Plot of the training, testing and the full data for Umudike 65
NOMENCLATURES
c Weibull scale parameter
ED Mean wind energy density
GWh GigaWatt hour: The Gigawatt is equal to one billion (109) watts or
1 Gigawatt = 1000 Megawatts
J Joule
k Dimensionless Weibull shape parameter
kV kiloVolt
kWh kiloWatt hour
MAD The mean absolute deviation
MAPE Mean absolute percentage error
MSE Mean square error
Mtoe Million tonnes of oil
MV MegaVolt
MW MegaWatt
MWe MegaWatt electrical
P Pressure
PV Photovoltaic
R^2 Coefficient of determination
RH Relative humidity
RMSE Root mean square error
T Temperature
V_E Wind speed carrying maximum energy
V_m Mean wind speed
V_mp Most probable wind speed
WPD Wind power density
ym Forecast data
yt Target data
F(v) Cumulative distribution function
Η Efficiency
σ Variance
PNL Pacific Northwest laboratory
JT Jacobian transformation
CHAPTER 1
INTRODUCTION
1.1 BACKGROUND OF STUDY
Energy commodities facilitate economic development by increasing productivity and income as well as creating employment (Ajayi et al., 2014). It is a generally accepted fact that there exists a strong correlation between socio-economic development and the availability of electricity. The electricity demand in Nigeria far outweighs the supply even as the supply is epileptic in nature. Nigeria is challenged with maintaining stable electricity supply even with the vast conventional and renewable energy resources available. Energy is an important tool that drives the economic growth of any country. An efficient energy market aims to provide energy commodities to power the industrial, transport, household and service sectors of the economy. Hence, energy remains the lubricant of sustainable economic growth (Chukwueyem et al., 2015).
Over the years, energy crisis has been a source of concern in Nigeria. The Federal government through its Renewable Energy Master Plan (REMP, 2009) set a target to increase the installed capacity from 5000MW generating capacity to 16,000MW by 2015 (Iloeje, 2001). However, this could not be achieved due to diverse reasons, but this target has to be achieved to ensure stability in electricity supply. It is worthy to note that if this target is to be met, the exploration of renewable energy resource with wind power as a prime element must be well integrated into this master plan. The utilization of renewable energy in the national energy consumption is on the low side while at present, natural gas, hydroelectricity, fuelwood and petroleum product constitute 5.2%, 3.1%, 50.5% and 41.3% share respectively. This justifies the fact that presently, Nigeria is dependent on hydroelectricity and traditional fuel wood as renewable energy sources (Akinbami, 2001). Notwithstanding these enormous wind potentials in the country, there is no existence of wind power plant connected to the national grid (Pam, 2007). There is but only a small number of small scale, stand-alone wind power plants installed and dates back to the 1960s in some northern states mainly for water pumps and grinding mills (ECN, 2014).
The energy situation in Nigeria is critical and is a key constraint for economic development. About 55% of the population has no access to electricity (Nigeria Power Baseline Report) and out of the total energy consumption, traditional biomass (firewood and charcoal) accounts for 86%. Population growth and economic development contribute to an increased need for electricity. The gap between production capacity and demand in combination with poorly maintained generation installa tions and a poor national and regional electricity grid, results in unstable and unreliable electricity supply for both households and companies. Firms from around Nigeria report average power outages of 8 hours per day. As a consequence, many companies and households rely on diesel generators for their electricity, with larger companies relying on gas or diesel for both electricity and power. However, also gas and diesel supply has proven unreliable (FMENV, 2014).
Wind speed profiles are known to vary in space and time. Wind energy analysis and modelling is an important pre-requisite in the design and citing of wind power plants. Therefore, accurate and detailed knowledge of the wind characteristics of the location is required for proper and efficient conversion/utilization of wind energy resource (Genc et al., 2005). To efficiently study wind energy resource, meteorological wind data which includes the distribution of wind speed is monitored on continuous basis through ground and satellite stations. The acquired data form a database useful as input parameters in the exploration of renewable energy resource. The study of wind characteristics and its exploration of available energy potentials has been an increasing concern to researchers in many developed countries (Riahy et al., 2008; Bekele and Palm, 2009 and Weigt, 2009).
In recent times, studies have been conducted at various locations in Nigeria (Fadare, 2008). The variability of wind speed modelled using different analytical tools such as: statistical models including Weibull and Rayleigh distribution functions (Igbokwe and Omekara, 2002); stochastic simulation; seasonal autoregressive integrated moving average; linear and multiple regression models (Oriaku et al., 2007). Several methods have been proposed for wind power forecast. These methods can be categorized into five (Wen-Yeau, 2013); these include persistence method, physical method, statistical method, spatial correlation method and artificial intelligence method. The artificial intelligence method includes artificial neural network (ANN), particle swarm optimization, adaptive neuro-fuzzy inference system among others. Amongst these methods of artificial intelligence, ANN gives the best approximation capability for any continuous non-linear function (Chaouachi and Nagasaka, 2012; Kelouwani and Agbossu, 2004).
ANN is a modelling and prediction tool widely accepted as a technique offering an alternative way to tackle complex and ill-defined problems (Kalogiru, 2001). They learn from examples and can handle noisy and incomplete data to deal with non-linear problems and once trained, can perform prediction and generalization at high speed. ANN has been employed in modelling of complex meteorological parameters (Kalogiru, 2001). ANN models also are efficient and less time consuming in modelling of complex systems compared to other mathematical models such as regression (Lin et al., 2003).
ANN with different topologies has been developed for spatial prediction of wind speed in different parts of the world (Cellura et al., 2008). The application of ANN in renewable energy systems and energy based application has been comprehensively reviewed by (Kalogiru, 2001). Hence, this study tends to use ANN in modelling and analysis of wind power in south-east Nigeria using available climatological parameters and data.
1.2 STATEMENT OF PROBLEM
Electricity is the bedrock to industrialization and economic development. Its availability for useful work is eminent hence the need to source for cheap and available energy source. Nigeria lags behind in the community of Nations even amongst developing nations in terms of grid based electricity consumption.
Only about 40% of the nation’s over 140 million has access to grid electricity and those who are connected to the grid face incessant power outages and interruptions, forcing many Nigerians to either look for alternative sources of energy or rely on self-power generation (Sambo, 2006). Renewable energy consumption is vital so as to reduce dependence on petroleum products, curtail greenhouse gas emissions, promote sustainable economic development and mitigate environmental pollution. Increasing energy efficiency is a primary strategy for achieving these goals. All these and many more could be achieved by efficient utilization of wind energy through the application of wind turbine.
1.3 AIM AND OBJECTIVES OF STUDY
The aim of this study is to analyse the wind energy profile of Umudike and Nsukka in eastern Nigeria for power generation using artificial neural network. The specific objectives include to:
i. determine the wind characteristics of Umudike and Nsukka in eastern region of Nigeria
ii. analyse the wind power potential of the region.
iii. investigate the most effective parameters in estimating generated power from observed wind speed in Umudike and Nsukka.
iii. establish an optimal ANN model for wind power prediction in Umudike and Nsukka in eastern Nigeria.
iv. make comparative analysis between the results from ANN model to Weibull technique.
v. development of predictive ANN-Regression equation for Umudike and Nsukka.
1.4 SCOPE OF STUDY
This study involves collection and analysis of data from Nsukka and Umudike. The results were compared to that of Owerri, Onitsha and Enugu (Oyedepo et al., 2012) all in the eastern region of Nigeria and finding the optimal model for the artificial neural network. It also involves investigation of the most effective parameters necessary for estimating available wind power using wind speed and other climatological parameters and data. The study tends to establish an ANN optimal model for wind power prediction, make a comparative analysis between the results from ANN models to Weibull model as well as development of ANN-Regression equation for Nsukka and Umudike.
1.5 JUSTIFICATION FOR THE STUDY
The erratic and epileptic state of electric power in Nigeria and the concern about global warming is of great concern for all. Harnessing wind power for electricity generation has become necessary owing to the fact that there is an increasing demand for electricity for industrial and household consumption in Nigeria. Also, the fact that fossil fuel emits and produces greenhouse gases that aids the depletion of the ozone layer, environmental pollution and degradation also gives a good reason for research into renewable energy sources for electricity generation.
Renewable energy uses energy sources that are continually replenished by nature (the sun, the wind, water, the Earth’s heat, and plants). Renewable energy technologies turn these fuels into usable forms of energy—most often electricity, but also heat, chemicals, or mechanical power.
Today fossil fuels are primarily used to heat and power our homes and fuel our cars. It’s convenient to use coal, oil, and natural gas for meeting energy needs, but because these fossil fuels are limited in supply, safety concerns and waste disposal problems, hence renewable energy can help fill the gap. Wind energy is the form of renewable energy obtained from wind movement. The use wind energy for electricity generation involves the application of wind turbine. Wind turbine converts wind energy into electrical energy, which is fed into electricity supply system. However, to do this, a proper study on the wind regimes and other characteristics must be studied.
ANN is an efficient parallel computing system whose central theme is borrowed from the analogy of biological neural networks. ANN acquires a large collection of units that are interconnected in some pattern to allow communication between the units. These units, also referred to as nodes or neurons, are simple processors which operate in parallel. These elements try to mimic the biological nervous system. They can be used in any application where a relationship exists between input and output variables; even if the relationship is complex and hard to articulate using traditional methods (Zhang, 2007).
Harnessing of wind power for electricity generation has become necessary owing to the fact that there is an increasing demand for electricity for industrial and household consumption in Nigeria. The fact that fossil fuel emits and produces greenhouse gases that aids the depletion of the ozone layer, environmental pollution and degradation also gives a good point for research into renewable energy sources for electricity generation.
Hence, this study applied the feed forward back propagation neural network to predict the available wind power in the south east region of Nigeria using Climatological data such as wind speed, pressure, temperature, relative humidity and air density were used for the prediction.
Click “DOWNLOAD NOW” below to get the complete Projects
FOR QUICK HELP CHAT WITH US NOW!
+(234) 0814 780 1594
Login To Comment