Seasonal Wind Characteristics and Prospects of Wind Energy Conversion Systems for Water Production in the Far North Region of Cameroon

This study aimed at investigating the characteristics of the wind power resource in the Far North Region of Cameroon (FNR), based on modelling of daily long-term satellite-derived data (2005-2020) and in-situ wind measurements data (1987-2020). Five different reliable statistical indicators assessed the accuracy level for the goodness-of-fit tests of satellite-derived data. The two-parameter Weibull distribution function using the energy factor method described the statistical distribution of wind speed and investigated the characteristics of the wind power resource. Six 10-kW pitch-controlled wind turbines (WT) evaluated the power output, energy and water produced. A 50 m pumping head was considered to estimate seasonal variations of volumetric flow rates and costs of water produced. The results revealed that the wind resource in FNR is suitable only for wind pumping applications. Based on the hydraulic requirements for wind pumps, mechanical wind pumping system can be the most cost-effective option of wind pumping technologies in FNR. However, based on the estimated capacity factors of selected WT, wind electric pumping system can be acceptable for only four out of twenty-one sites in FNR.


Introduction
Wind has nowadays become a stable form of power supply and is considered as one of the most cost-effective means for delivering low-carbon energy services, particularly to the most vulnerable segments of the population in numerous developing nations. It's anticipated that by 2050, wind power could contribute to more than 25% of the total emissions reductions needed (approximately 6.3 gigatons of carbon dioxide annually), under the energy goals set out in the United Nations 2030 Agenda and the Paris Agreement. Wind energy (WE) would then generate more than 35% of total electricity needs, becoming the prominent generation source by 2050 [1].
Over the last two decades, the yearly growth rate of global WE has been as high as 38.56% (2001), as low as 9.61% (2018) and on average 22%. At the end of 2019, global WE generation capacity amounted to 622.7 gigawatts (GW), which represented 25% of renewable generation capacity by energy source. Hydropower, the largest share of the global total, accounted for 47% (1190 GW), while the share of solar reached 23% (586 GW) in 2019 [2]. Globally, WE performed particularly well in 2019, expanding by 58.9 GW (10.44%). Asia accounted for 49.47% of new capacity in 2019, increasing its WE generation capacity by 29.13 GW to reach 258.32 GW (41.48% of the global total). WE capacity in Europe and North America expanded by 14.02 GW (+31.46%) and 11.48 GW (+19.85%), respectively [3]. Oceania and the Middle East were the fastest growing regions (+22.18% and +17.75%, respectively), with 2.47% and 0.19%, representing their share of global WE capacity, respectively. Africa accounted for 0.51%, the lowest of new capacity in 2019, increasing its wind energy capacity by only 0.3 GW to reach 5.7 GW (0.93% of the global total). Compared to 2018, capacity growth in Africa and Middle East was somewhat lower than in 2019, but higher in Asia, Europe and North America [3].
Despite being the least growing region in terms of WE generation capacity, Africa has WE resources and potential that can meet its current needs, if properly tapped. Several studies have shown that the wind resource in Africa is greatest around the coasts and in the eastern highlands [4] [5]. However, the WE development in the African continent remains very slow as a result of limited support at the level of the continent, since the vast majority of WE projects necessitate financial support from organizations based out of the continent [6]. By the end of 2019, North Africa and the Republic of South Africa continued to dominate, with 49.44% (2.85 GW) and 36.32% (2.09 GW), representing their share of WE capacity in the African continent.
Sub-Saharan Africa, accounted for 14.24%, representing the lowest share of WE capacity. At roughly 0.82 GW, the entire WE generating capacity of the 47 countries of sub-Saharan Africa (excluding the Republic of South Africa), is less than that of Morocco. As a result, sub-Saharan Africa has the world's lowest WE generation capacity, despite the wind potential that is essentially untapped. Furthermore, transition-related clean energy investments in Sub-Saharan Africa is  [7]. Moreover, sub-Saharan Africa displays the lowest electricity access of only 45%, far lower than the world average of 89%. Furthermore, the vast majority of people (over 99%) deprived of electricity are in developing nations, and four-fifth of them live in rural South Asia and sub-Saharan Africa [8].
Similarly, Cameroon, does not have any installed WE capacity, despite the existing potential. Neighboring countries with comparable wind potential, have taken steps in exploring wind power. By the end of 2019, WE generation capacity in Chad and Nigeria amounted to approximately 1 and 3 megawatts (MW), respectively [3]. Most of the analyses performed to assess the potential of wind power have shown that the whole country lays in low wind resources regime, with very limited high wind sites. The vast majority of sites fall under poor to marginal wind regime. However, detailed information on the potential wind resource, which is of paramount importance when forecasting wind power for the optimal site selection, has yet to be precisely acknowledged. Locally measured wind data are generally available at meteorological stations located at the main airports, while there are no ground station measurements for the vast majority of locations which are far (at least 50 km) from the main airports.
When meteorological measured wind data from masts are missing, wind resource estimation using daily long-term satellite-derived data are considered [9] [10] [11]. Furthermore, for comparison analysis, both meteorological observations and satellite-derived data are used to estimate the local accuracy [12] [13] [14].
All things considered, the proposed work aims at investigating the characteristics of wind power resource from twenty-one locations in FNR, using daily long-term satellite-derived data for the period 2005-2020 and 3-hourly time step observed wind speed data from 02 weather recording locations (Kousseri and Maroua) for the period 1987-2020. The main objective of this study is to provide a reasonable wind power resource assessment in the early phase of wind farm projects using satellite-based wind resource, before higher-accuracy in-situ mea-

Wind Data Description and Source
For this study, in situ measurements (3- [18]. Table 1 provides geographical coordinates of the twenty-one sites considered, as well as satellite and in situ measurements periods.

Wind Speed and Standard Deviation
In this research, the first step in the assessment of seasonal wind characteristics in FNR, is to analyze in situ measurements (3-hourly time step observed wind speed data) from 02 weather recording locations at Kousseri and Maroua and daily long-term satellite-derived data, recorded at a height of 10 m agl, using mean wind speeds and standard deviations. Figure 2 recapitulates monthly, annual and seasonal mean wind speeds and standard deviations using in situ and satellite measurements at Kousseri and Maroua.
It is seen in Figure 2 that the highest in situ wind speeds occur in the dry sea-  (2) where: σ = standard deviation of the mean wind speed [m⁄s];

( )
i v = wind speed [m⁄s]; N = number of wind speed data. Table 2 provides at the twenty-one selected sites, annual, dry and rainy seasons mean wind speeds, standard deviations and ambient temperatures using daily long-term satellite-derived data for the period 2005-2020, recorded at a height of 10 m agl. Mean wind speeds in FNR vary in the ranges of 2.99 -4.32 m/s, 2.12 -3.23 m/s, 3.43 -4.87 m/s for yearly averages, rainy and dry seasons, respectively. It observed that the variance of streamflow occurrence in the rainy season is smaller than that of the yearly average and dry season, which may suggest a more accurate prediction. On the other hand, higher SD in the dry season present streamflow values that are widespread and may be less accurate. Mean ambient temperatures values are between 25.74˚C and 29.67˚C. Lower temperatures are seen in the rainy season, while higher values occur in the dry season.

Weibull Probability Density Function
The Weibull probability density function (PDF) is used to describe the statistical distribution of wind speed. The Weibull PDF is a useful tool to characterize the wind speed and power in a given location, as well as to evaluate mean monthly, yearly and seasonal net energy production and performance wind energy systems [19] [20]. The Weibull PDF can be described by its PDF ( ) f V and cumulative distribution function (CDF), ( ) F V [21] using Equations (3) and (4).

( )
f v = probability of observing wind speed v; v = wind speed [m⁄s]; C = Weibull scale parameter [m⁄s]; k = Weibull shape parameter.
The determination of the two-parameter Weibull PDF requires the knowledge of the shape (k, dimensionless) and scale (C in m/s) parameters. Various well-established estimation methods are used for the purpose of computing Weibull parameters at a given location [22]. In this work, Weibull shape and scale parameters are computed using the energy pattern factor method (EPF). First, the energy pattern factor ( pf E ) [23] [24] [25] is given by Equation (5). Smart Grid and Renewable Energy Then, the shape and scale parameters are computed using Equations (6) and (7).

Statistical Indicators for Accuracy Evaluation
To assess the accuracy level for the goodness-of-fit tests of satellite-derived data, five reliable statistical indicators have been used to compare measured 3-hourly time step observed wind speed data and daily long-term satellite-derived data. These statistical indicators are presented using Equations (8) to (12)  ( ) 2) Root mean square error (RMSE) [28] [29]: ( ) 3) Relative root mean square error (RRMSE) [19] [30]: ( ) where:

Extrapolation of Wind Speed
The wind speed data were collected at a height of 10 m agl.  (13) and (14). 10 10 The power law exponent n is given by Equation (15).
where, z and z 10 are in meters, Weibull C 10 and k 10 parameters are determined at 10 m height agl.

Mean Wind Power Density and Energy Density
Expressed in watts per square meter (W/m 2 ), wind power density (P(v)) considers the wind speed frequency distribution of a given location and the power of wind which is proportional to the air density and the cube of the wind speed.
The power of wind (P(v)) can be estimated using Equation (16).
The mean wind power density ( D p ) based on the Weibull probability density function can be calculated using Equation (17).
The mean energy density ( D E ) over a period of time T is expressed as Equation (18).
where: ρ = air density at the site; A = swept area of the rotor blades [m 2 ].
The air density (in kilograms per cubic meter) at a given site is computed as the mass of a quantity of air (in kg) divided by its volume (in cubic meter). It depends on elevation and temperature above sea level and can be computed [35] using Equation (19).  (19) where: Z = elevation (m); T = temperature at the considered site (˚K).

Power Curve Model and Capacity Factor
The typical power curve of a 10-kW pitch-controlled WT is shown in Figure 3(a), while the power curves using the six selected pitch-controlled WT of 10 kW rated capacity are plotted in Figure 3(b). As a result of the pitch regulated systems, the voltage of the electricity at which pitch-controlled WT generate power at WS above their rated levels, does not decrease [36]. Four different zones are observed in this curve (Figure 3(a)). For WS in the range of zero to V I (cut-in WS),  wind speed (v R ), cut-off wind speed (v F ) and rated electrical power (P eR ). All these speeds and power are computed using the parabolic law [37], as a combination of Equation (20). The average power output ( , e ave P ) of the WT, based on the Weibull PDF, can be computed using Equation (21).
The ratio of the average power output ( , e ave P ) to the rated electrical power ( eR P ) of the WT is known as the capacity factor CF. CF can thus be expressed [38] as Equation (22).

Water Pumping Capacity
The water pumping capacity rate ( w F ) is related to the net hydraulic power output ( out P ) and the efficiency of the pump. To determine a volume of water , the net hydraulic power output ( out P ) and volumetric flow rate of water ( w Q ) are computed [39] using Equations (23) and (24). where:

Costs Analysis
The water pumping capacity rate ( w F ) is related to the net hydraulic power output ( out P ) and the efficiency of the pump. To determine a volume of water  With the following assumptions: • I is the investment cost, which includes WT price in addition to 20% for civil works and other connections; • Average specific WT cost per kW is USD 2600, for WT rated power less than 20 kW [41]; • n is the useful lifetime of WT in years (20 years); • i 0 is the nominal interest rate (16%); • S is the scrap value (10% of WT price); • i is the inflation rate (3.6%); • C om is the operation and maintenance costs (7.5% of the investment cost).
The total energy output ( WT E ) over WT lifetime (in kilowatt-hour) is computed using CDF of wind speeds at which WT produce energy (A), rated power of the WT, capacity factor CF and WT lifetime working hours. WT E is computed as Equation (28).
The costs of energy (COE) per unit kWh and costs of water (COW) per unit m 3 are estimated using Equations (29) and (30).

PVC COE
The annual volume of water V w (m 3 /year) produced is determined using Equation (31). Figure 4 and Figure 5 show monthly average PDF at 10 m height agl, respectively at Kousseri and Maroua using both measured and satellite-derived data, while     At the site of Maroua, it is observed under the same conditions, lower probability (around 0.15) using measured data and higher probability (around 0.31) using measured data. Statistical indicators for the accuracy of satellite-derived WS at Kousseri and Maroua are displayed in Table 4. Weibull CDF values provided data for the statistical analysis and comparison between measured and satellite-derived data. Table 4 shows different values obtained using the five statistical indicators for the accuracy of satellite-derived wind speed at Kousseri and Maroua.

Statistical Indicators for the Accuracy of Satellite-Derived WS at Kousseri and Maroua
The   shows an accuracy level in the range of excellent to good, to test the goodness-of-fit of satellite-derived data. Therefore, satellite WS are found to be a good fit with high correlation at both locations. Figure 8 presents seasonal average PDF at 10 m height agl for the twenty-one selected locations. Dry season average PDF (Figure 8(a)) displays lower percentage probability, with a larger range of speeds, while rainy season average PDF ( Figure 8(b)) shows higher percentage probability, with a narrower range of WS. Table 5 Figure 9 illustrates seasonal average PDF at 30 m height agl for the twenty-one selected locations. As previously described, dry season average PDF ( Figure   9(a)) displays lower percentage probability, with a larger range of speeds, whereas rainy season average PDF (Figure 9(b)) shows higher percentage probability, with a narrower range of WS.             [43]. Wind electric pumping system can be implemented at Hilé-Alifa, Blangoua, Kousseri and Goulfey, using WT characteristics similar to WT 1 . Based on hydraulic requirements for wind pumps, the use of Mechanical wind pumping system is highly suggested as the most cost-effective option of wind pumping technologies in FNR. Figure 11 shows mean monthly flow rate capacity (m 4 /h) histograms using WT 1 , at (a) Blangoua, (b) Goulfey, (c) Hilé-Alifa and (d) Kousseri. Higher flow rate capacity are observed in dry season, whereas lower values are seen in rainy season. The lowest flow rate capacity are observed in September followed by August, whereas the highest values are shown in March followed by February. Table 11 illustrates mean seasonal volumetric flow rate of water (m 3 /day) at 50 m dynamic head using the six selected WT at the twenty-one selected locations. Volumetric flow rate (Q w ) and flow rate capacity (F w ) are lineary related to each other, hence they follow the same trend when ranking WT performance. WT 1 achieves the highest volumetric flow rate, whereas WT 3 , WT 2 and WT 4 rank, respectively 2 nd , 3 rd and 4 th . WT 4 reveals the same performance as WT 5 Table 12 illustrates mean seasonal costs of water (XAF/m 3 ) at 50 m dynamic head using the six selected WT at the twenty-one selected locations. COW and flow rate capacity are lineary related to each other, hence they follow the same tendency when ranking WT performance. WT 1 achieves the highest volumetric flow rate, whereas WT 3 , WT 2 and WT 4 rank, respectively 2 nd , 3 rd and 4 th . WT 4 reveals the same performance as WT 5 . The least efficient is WT 6 . With consideration to WT 1 , the most performing of considered WT, dry season COWare 9.06,    Figure 12 displays mean monthly COW and volumetric flow rate using WT 1 , at (a) Blangoua, (b) Goulfey, (c) Hilé-Alifa and (d) Kousseri. With respect to the PVC method, COW are inversely proportional to volumetric flow rate. It is observed that the higher the volumetric flow rate, the lower the COW. Lower COW are observed in dry season, whereas higher COW are experienced in rainy season. COW are highest in September and August, while March and February display the lowest COW.

Conclusion
In this work, seasonal wind characteristics, net energy production and performance of selected 10-kW pitch-controlled WT in twenty-one selected locations in FNR have been evaluated using measured wind and satellite-derived wind data at 10 m height agl. Five reliable statistical indicators have been employed to assess the accuracy level of satellite-derived data. The 2-parameter Weibull PDF using the energy factor method provided the required tool to investigate seasonal wind characteristics, net energy production and performance of selected WT. The outcomes of this study show that mean wind speeds at 10 m height agl in FNR vary in the ranges of 2.99 -4.32 m/s, 2.12 -3.23 m/s, 3.43 -4.87 m/s, respectively for yearly average, rainy and dry seasons. Satellite-based wind resource can be appropriate to assess the potential of wind energy in the early phase of wind farm projects, before higher-accuracy in-situ measurements are available. The wind resource in FNR is deemed suitable for wind pumping applications. Based on the hydraulic requirements for wind pumps, mechanical wind pumping system can be the most cost-effective option of wind pumping technologies in FNR. Wind electric pumping systems using WT, with cut-in WS (less than 2 m/s) and rated WS (less than 10 m/s) can be a cost-effective option for water pumping for four locations only, namely, Blangoua, Goulfey, Hilé-Alifa and Kousseri.