Economic assessment of WECS for water pumping systems in the North Region of Cameroon

This paper analyses the potential utilization of wind electric pumping for water distribution in offgrid locations of the North Region of Cameroon (NRoC), using ground measured data as well as long-term satellite-derived data. Furthermore, this paper puts emphasis on statistical indexes of accuracy for the comparison of measured (2007–2012) and long-term satellite-derived (2005–2020) data. The outcome of this study clearly indicates that long-term satellite-derived data, obtained through the Prediction of Worldwide Renewable Energy Resources, can be considered as a viable alternative to missing site-specific data from ground stations, mainly in developing countries. The accuracy of satellite-based wind resource is deemed sufficient to provide a reasonable assessment in the initial phase of wind project planning, before in situ measurements with high accuracy are available. To model wind speeds characteristics, the energy pattern factor method (EPF) has been utilized as one of the reliable Weibull distribution methods for the assessment of wind energy potential at desired turbines heights in the selected locations. Five 20-kW pitch-controlled wind turbines (WT) with a hub height of 30m, are considered to evaluate the power output and energy produced. Volumetric flow rates and costs of water produced are estimated. The results showed that, out of the eight chosen locations, Figuil shows the best combination of capacity factors (CF) and costs of energy (COE), while the site of Poli displays the worst values of CF and COE, no matter which of the five WT is employed. Thus, selecting WT for low wind speeds regimes, should require to combine location wind resource and WT characteristics such as lower cut-in wind speed (1.5m/s) as well as lower rated wind speed (8.0m/s) in order to successfully and affordably implement the wind technology in the NRoC.


Introduction
The sustainable development of a country is closely linked to the utilization of renewable energy resources (RES) at its disposal [1]. RES provide not only an important source of employment, wealth and profits to the country, but also play a critical role in building a sustainable and modern society [2][3][4]. Cameroon is endowed with conventional sources of energy such as natural gas and oil fuels, however associated with the emissions of both air pollutants, greenhouse gases and global warming. These conventional sources of energy share the fuel sources for electrical energy production (natural gas and oil fuels) in the country and other energy needs of industries and populace [5]. Moreover, fossil fuel sources, which are gradually depleting, need to be preserved and utilized efficiently [6]. As the global energy demand increases and fossil fuels become a gradually expensive commodity to humankind, the focus has been shifted towards renewable energy for a sustainable development [7].
RES, such as biomass, geothermal, hydro, solar, wind are inexhaustible, clean, free and offer many benefits such as non-polluting and almost inexhaustible characteristics in contrast to fossil fuel [8]. To further add to RES advantages, they can be harnessed on a local basis for applications, such as decentralized production of electricity in remote or rural areas, with low density of population [9].
Amongst RES, wind energy is one of the world's most abundant and permanent energy source, with a high-level of utilization. In addition, the potential of wind energy is credited, in the best of cases, with forty times the global electricity demand [10]. Over the last twenty years, the yearly growth rate of global wind energy has achieved an average growth rate of 22% [11]. Wind energy conversion systems (WECS) are considered as a mature and costcompetitive technology for electricity production [12], which is produced locally [13], but its dissemination remains marginal in most African countries. WECS necessitate to correctly characterize the wind resource over a location in addition to its geographic distribution [14]. As result, wind power is poised to play a key role in meeting global growing demand for electricity, especially in developing countries such as Cameroon, where it can be harnessed at an affordable cost on a local basis for water pumping applications.
The NRoC lays in a low wind speed regime and as a result the region is faced with numerous challenges in developing wind energy [15]. Measured wind data are mostly available at the ground measurements station at the Garoua main airport, while for the seven other considered sites located beyond a radius of at least 50 km of the main measuring wind station, there are no ground measurements stations. Therefore, long-term satellitederived data [16], obtained through the POWER (Prediction of Worldwide Renewable Energy Resources), are deemed suitable and a viable alternative to missing site-specific data from ground stations [17][18][19]. These satellite-derived data are at daily level, at a 0.5 Â 0.5degree resolution. Additional scientific literature related to the use of long-term satellite-derived data and their accuracy in comparison with in situ measurements are provided as referenced [20][21][22][23].
The aim of this paper is to acknowledge the use of satellite-based wind resource as appropriate data to provide a reasonable assessment, before higher-accuracy in situ measurements are accessible. Furthermore, this work explores wind power production using electric systems for water pumping in the NRoC. Measured wind speeds at 10 m height above ground level (agl), in hourly time-series format, from January 2007 to January 2012 were collected through a cup-generator anemometer for Garoua. When measured ground level data remained nonexistent, satellite-derived data for the sites of Bashéo, Beka, Figuil, Pitoa, Poli, Rey-Bouba and Touboro, were obtained from the NASA Langley Research Center (LaRC) POWER Project funded through the NASA Earth Science/Applied Science Program [24]. The accuracy of satellite-derived data for Garoua (the only site with ground measurements) was assessed using statistical indexes of accuracy, namely coefficient of determination (R 2 ), index of agreement (IOA), mean bias error (MBE), root mean square error (RMSE) and relative root mean square error (RRMSE).
Kidmo et al. [5] summarized most recent studies performed across Cameroon to investigate wind speeds and power characteristics, wind energy and costs of WECS. Recent and relevant research studies related to Garoua are detailed as followed. Kidmo et al. [25] performed a statistical analysis of wind speed distribution using six Weibull Methods for wind power assessment. Kidmo et al. [26] evaluated wind energy potential for small scale water pumping. Less attention was so far paid to the combination of location wind resource and wind turbines (WT) characteristics such as cut-in wind speed as well as rated wind speed. This paper highlights the use of low wind speeds WT to take full advantages of local wind resource and WT characteristics. Capacity factors (CF), costs of energy generated (XAF/kWh) and water produced (XAF/ m 3 ) are estimated using the Present Value Cost (PVC) method of energy produced per year, considering the frequency at which WT produce power.
To model wind speeds characteristics, the energy pattern factor method (EPF) [27,28] has been utilized as one of the reliable Weibull distribution methods for the assessment of wind energy potential at desired turbines heights in the selected locations. Five 20-kW pitchcontrolled WT with a hub height of 30 m, are considered to evaluate the power output and energy produced. Volumetric flow rates and costs of water produced are estimated using a 50 m pumping head, for the sake of simplicity. The results showed that, out of the eight chosen locations, Figuil shows the best combination of capacity factors (CF) and costs of energy (COE), while the site of Poli displays the worst values of CF and COE, no matter which of the five WT is employed. Thus, selecting WECS for low wind speeds regimes, should require to combine location wind resource and WT characteristics such as lower cut-in wind speed (1.5 m/s) as well as lower rated wind speed (8.0 m/s) in order to successfully and affordably implement the wind technology in the NRoC.

Study area
The North Region as shown in Figure 1, is one of the ten regions of Cameroon, Located between 7.30 À 10.00 N latitude and 12.30 À 15.00 E longitude. It covers an area of 66 090 km 2 and is bordered by two regions of the country, the Far North Region to the north and the Adamawa Region to the south, Chad to the east, Central African Republic to the southeast and Nigeria to the west. The North region accounts for one of the driest regions of Cameroon under the influence of the harmattan winds in the dry season which lasts eight months, temperatures rise at their highest (40-45°C) and there is no rainfall. During the rainy season which lasts four months, average rainfall is between 900 and 1500 mm and torrential rains are observed as well as lower temperatures compared to dry season. The Benoue depression constitutes the primary land feature of the North region, with an elevation in the range of zero to 200 m. Land elevations are in the range of 200-500 and 500-1000 m, respectively in the north and south of the region [29]. The selected sites, namely, Bashéo, Beka, Figuil, Garoua, Pitoa, Poli, Rey-Bouba and Touboro, as shown in Figure 1, have been chosen based on relevant landscape information they provide in order to fully assess the potential of wind energy in the NRoC.

Wind data description and source
Two sources of data have been utilized in this work, measurements from a ground station for the site of Garoua and long-term satellite-derived data for the seven other sites. Wind speed measurements for the locality of Garoua, at 10 m height agl, in hourly time-series format, from January 2007 to January 2012 have been collected through a cup-generator anemometer at the Garoua, main meteorological station [25]. For the seven other selected sites of the north region, measured ground level data are non-existent. Therefore, long-term daily satellite-derived data, from January 2005 to January 2020, obtained through the POWER, have been utilized. These data were obtained from the NASA Langley Research Center (LaRC) POWER Project funded through the NASA Earth Science/Applied Science Program [24]. Table 1 provides geographical coordinates of the eight sites, as well as mean ambient temperature and measurement period.

Mean wind speed and standard deviation
Monthly mean wind speed v m and standard deviation s of wind speed data are calculated as Equations (1) and (2) [25]: After the statistical analysis of wind data, monthly mean wind speeds and standard deviations for Garoua using measured and satellite-derived data, are summarized in Table 2, while Figures 2 and 3 show, respectively monthly mean wind speeds and standard deviations using satellite-derived data for for Bashéo, Beka, Figuil, Garoua, Pitoa, Poli, Rey-Bouba and Touboro. The magnitude of monthly mean measured wind speeds and standard deviations for Garoua at 10 m height agl, lay respectively within the range of 1.24 to 2.71 m/s and 0.73 to 1.39 m/s, while corresponding values using satellite-derived data range from 1.67 to 3.38 m/s and from 0.47 to 1.00 m/s, in that order.

Weibull probability density function
The Weibull probability density function (PDF), which is a special case of a generalized two-parameter Gamma distribution, has been extensively utilized in scientific literature for wind speed forecasting and wind energy potential assessment [30][31][32]. Weibull PDF can be characterized by its PDF f (V) using Equation (3) and cumulative distribution function (CDF), F (V) using Equation (4) [33]: Where: Weibull parameters k and C are typically obtained using well-established estimation methods [25,34]. In this paper, Weibull parameters are computed using the energy pattern factor method (EPF). The first step is to define the power density expressed using Equation (5) [35].
Where: r = air density at the site; The air density (in kilograms per cubic meter) at a given location is calculated 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 [36] using Equation (6).
The second step of the EPF method is to determine the energy pattern factor (E pf ), as Equation (7) [37]: Where: v m = mean wind speed [m/s]; ðv 3 Þ m = mean of wind speed cubes.
Once the energy pattern factor is calculated by using Equation (7), the Weibull shape parameter is estimated from Equation (8) [38]: where k is the dimensionless shape parameter illustrating how peaked the wind distribution is, while the scale parameter C indicates how 'windy' the wind location under consideration is [39]. The scale parameter is obtained using Equation (9): Where the standard gamma function is given by [40]: The gamma function can be approximated by [25,41]:

Statistical indexes of accuracy
In order to evaluate the performance of the considered satellite-derived data of Garoua, the following statistical indexes of accuracy are utilized: 1. Mean Bias Error (MBE) of satellite-derived values using Equation (12) [42,43]: RMSE provides the deviation between satellite-derived and measured values. Successful forecasts correspond to low values of RMSE, while higher indicate deviations [43]. RMSE should be as close to zero as possible.
3. Relative root mean square error (RRMSE) using Equation (14) [30,46]: R 2 determines the linear relationship between CFD of average satellite-derived and average measured values. A higher R 2 represents a better fit using satellite-derived data and the highest value it can get is 1.
Where: Gr i is the i th CFD of ground (measured) wind speeds, Sat i is the i th CFD of satellite-derived wind speeds, Sat i is the mean value of Sat i , N the number of non-zero wind speed data points.
5. Index of Agreement (IOA) using Equation (16) [30,49]: Where: Gr i denotes the average of Gr i , Sat i is the mean value of Sat i , N the number of non-zero wind speed data points. IOA estimates the accuracy of satellite-derived wind speeds in predicting ground (measured) wind speeds. IOA values range from zero to one. IOA values above 0.5 imply efficiency in using satellite-derived data [43] 2.6 Extrapolation of wind speeds at different hub heights In this study, five WT, using a 30 m hub height have been considered. Wind speeds data obtained at 10 m height agl, are therefore adjusted to the relevant WT hub height. The vertical extrapolation of scale and shape parameters at an elevation (z), is defined by the relationships in Equations (17) and (18) [50,51]: Where z and z 10 are in meters and the power law exponent n is given by Equation (19): Where, the scale C 10 and shape k 10 parameters are determined at 10 m height agl.

Mean wind power density and energy density
Wind power density (P (v)), in watts per square meter, is estimated using Equation (20) [52].
The mean wind power density (p D ) based on the Weibull probability density function can be calculated using Equation (21) [33].
The mean energy density (E D ) over a period of time T is expressed as Equation (22).
Where: r = air density at the site; A = swept area of the rotor blades [m 2 ].

Wind turbine and electric pumping systems
Five WT, from different manufacturers are considered, as illustrated by wind power density regime, for locations falling into Class 2 or lesser, based on PNL wind power classification scheme [50]. Although a 50 m pumping head is considered to calculate volumetric flow rates, a different pumping head can be considered, since the volumetric flow rate of water is inversely proportional to the pumping head.

Power curve model and capacity factor
The typical power curve of a 20-kW pitch-controlled WT is shown in Figure 4. Four different zones are observed as illustrated. For WS in the range of zero to cut-in wind speed (V C ), there is no output yielded. From the cut-in to rated WS (V C to V R ), the power increases with the wind speed (WS). The power output curve is assumed to show a quadratic power shape, based on the selected power curve model in this work. From the rated WS (V R ) to the cut-out WS (V F ), the WT yields a constant output at the rated power (P R ), regardless of WS variations. For WS higher than V F , there is no output yielded.

Power curve model and capacity factor
The power curve model of a WT can be modelled by four parameters: the cut-in wind speed (v c ), the rated wind speed (v R ), the cut-off wind speed (v F ) and the rated electrical power (P eR ). In this study, the power output curve model of a WT can be approximated with the parabolic law, using the combination of Equation (23) [54]: The average power output (P e,ave ) from the turbine, which is related to the total energy production can be determined from Weibull distribution as Equation (24): The capacity factor CF is defined as the ratio of the average power output (P e,ave ) to the rated electrical power (P eR ) of the WT. The capacity factor CF can thus be

Water pumping capacity
The net hydraulic power output (P out ) required to deliver a volume of water V w (m 3 ) can be expressed using Equation (26) [56].
Where: The volumetric flow rate of water is therefore evaluated using Equation (27).
When taking into account the efficiency of the pump (h PUMP = 62% for the submersible electric pump model) the water pumping capacity rate (F w ) can be expressed as Equation (28).

Costs analysis
In this work, the adopted method to evaluate the costs of energy (COE) and costs of water (COW) produced is the present value of costs (PVC) of energy produced per year using Equation (29) [57].
Where the following assumptions are made to estimate the cost of energy produced by the considered WT: -I is the investment cost, which includes the WT price in addition to 20% for civil works and other connections; -Average specific WT cost per kW is USD 1175, for WT rated power between 20 and 200 kW [53] n is the useful lifetime of WT in years (20 yrs); i o is the nominal interest rate (16%); -S is the scrap value (10% of the WT price); i is the inflation rate (3.6%); -C om is the operation and maintenance costs (7.5% of the investment cost).
The discount rate (r) is determined using Equation (30) [58]: The Wind availability (A) is determined using Weibull CDF of wind speeds at which WT produce energy. The total energy output (E WT ) over the WT lifetime (in kilowatt-hour) is computed as Equation (31).
The cost of energy (COE) per unit kWh using the PVC method can be estimated using Equation (32).
The annual volume of water V w (m 3 /year) produced by the WT is determined from Equation (33).
The cost of water (COW) per unit m 3 using the PVC method can be estimated using Equation (34).
3 Results and discussion

Measured vs satellite data
Average monthly Weibull CDF values helped to perform the statistical comparison between measured and satellitederived data. satellite-derived data can accurately represent measured data and when measured ground level data are missing, satellite-derived data can be suitable to assess the potential of wind energy. Similar studies on the use of satellitederived data are found in the literature. Schmidt et al. [23] performed a comparison between in situ measurements of wind speed from Wave Glider (WG) and satellite/ reanalysis products, including the wind speed error per wind speed category, RMSE, bias, and correlation coefficients. Accuracy values are within the same ranges as the present work. Barthelmie and Pryor [21] worked on satellite sampling of offshore wind speeds to represent wind speed distributions. Their study have stressed the difficulties to obtain through in situ methods wind speeds over the oceans. Remmers et al. [22] observed a very strong positive linear relationship between satellite-derived wind speed data and the in situ measurements. Ayompe and Duffy [59] assessed the energy generation potential of photovoltaic systems in Cameroon using satellite-derived solar radiation datasets due to the lack of a reliable network of surface observation stations for collecting weather data in the country. These studies concluded that the accuracy of satellite-derived data varies according to empirical model functions utilized. Besides, the accuracy of satellite-based wind resource is sufficient to provide a reasonable assessment in the initial phase of wind project planning, before higher-accuracy in situ measurements are available.

Wind characteristics
It is observed that the probability of observing higher wind speeds is highest for Figuil. In addition, the probability of observing lower wind speeds is highest for Poli.  [60]. Class 1 areas are considered unsuitable for large scale wind power development. It can be further observed that, for each of the eight sites, the peak of the PDFs, which indicates the most frequent wind speed, skews towards the higher values of wind speed. Table 5 proposes statistical Weibull analysis of  Table 6 presents for WT 1 , WT 2 , WT 3 , WT 4 and WT 5 , the annual average frequency for wind speeds greater or equal to WT's cut-in wind speed for each of the eight selectd sites. It's observed that the probability of WT to produce power is the highest using WT 1 , while this availability is the lowest using WT 5 . Figure 5 shows a typical wind electric pumping system. In order to highlight WECS characteristics needed for low wind speed locations, the five WT, with a rated capacity of  20 kW and a 30 m tower hub height, were considered for uniformity in the comparison. These WT are represented by WT 1 , WT 2 , WT 3 , WT 4 and WT 5, to avoid the use of trade names. Cut-in and rated wind speeds are different for each of the five WT. WT 1 has the lowest cut-in and rated wind speeds, while WT 5 has the highest cut-in and rated wind speeds as illustrated by Table 3. Five 20-kW pitchcontrolled WT, whose power curves are illustrated by Figure 6, combine Figuil wind characteristics with WT characteristics. The WECS are equipped with a submersible electric pump model to produce energy from wind and pump water using the produced energy for each of the eight selected sites. For the purpose of this work, a total dynamic head of 50 m was adopted as the average in the NRoC. Any other pumping head could be used since the volumetric flow rate of water is inversely proportional to pumping head.

Conclusion
The economic assessment of wind energy conversions systems, for electric water pumping systems in the NRoC, has been explored, using measured and satellite-derived wind data.