Nowadays the world is heading towards an energetic transition phase where the main goal is to reduce the dependency on carbon-based sources of energy and increment the integration of new technologies that can utilize renewable sources as the main fuel to produce usable energy. A high interest is given to solar energy technologies, since it is a power source that is not limited to a few regions around the world. However, solar irradiation available on the Earth surface on solar power premises is not easily manageable due to the unpredictability of weather conditions, such as cloudiness and humidity.
Following this momentum, future plans about increasing solar investment are becoming more popular. Egypt, for instance, established the goal of increasing renewable production up to 20% by 2022. The country also expects a national capacity of solar energy of 3.5 GW by the year 2027. Saudi Arabia devised a plan in which by the year 2023, 9.5 GW of the country’s energy production will come from clean energy. The United Arab Emirates (UAE) set an ambitious goal of 1 GW of solar capacity by 2020 and 5 GW by 2030.
In this regard, the Kingdom of Morocco stated, through the National Solar Plan in 2009, the goal of reaching 2 GW of installed capacity by 2020. Recently, this goal was updated to 4 GW by 2030. According to it, it is expected that up to 52% of electricity production will be based on renewable technology. As a result, Morocco is considered to be an important actor concerning solar energy development in the future of the MENA region and, consequently, it has been chosen as an object of study.
1.1. State of the Art
The scientific literature about solar irradiation is widely extended and covers all the relevant aspects, from those studies focusing on the determination and forecasting of the global horizontal irradiation (GHI) as well as the beam or direct normal irradiation (DNI) on a horizontal surface to those on a tilted surface, at different time frames (see
Figure 2).
The existing literature aimed at forecasting solar irradiation and/or irradiance is quite extensive. Likewise, so are the different methods used to carry it out. As extracted from [
7], which presents an in-depth analysis of the performance of different solar irradiance forecasting models, the predominant forecasted variable in most studies is the GHI, followed by the DNI. Moreover, the most frequent forecasting horizon is the short term, i.e., intra-hour, intra-day or, to a lesser degree, day-ahead predictions. This same trend is observed in [
8], where an overview focused on solar irradiation forecasting methods using machine learning approaches is given.
In this context, some references can be highlighted dealing with short-term solar irradiance forecasting [
9,
10,
11,
12,
13,
14,
15,
16,
17,
18,
19,
20,
21,
22,
23,
24,
25,
26,
27,
28,
29,
30,
31,
32,
33,
34,
35] and, to a lesser extent, some focused on monthly solar predictions [
36,
37,
38,
39,
40,
41].
Intra-hour GHI forecast employing a cloud retrieval technique to develop a physics-based smart persistence model is improved in [
9], and an algorithm using cloud physical properties for intra-day GHI and DNI forecasting with time horizons of 0–4 h at a 15 min temporal resolution is developed in [
10]. A GHI forecasting model based on satellite data from Finland with a forecast horizon of 4 h and a 15 min temporary resolution is developed and validated in [
11], while the error obtained from the Japan meteorological agency mesoscale model in the hourly-averaged GHI forecasts from 2008 to 2012 is assessed in [
12]. The predictions of 15 clear-sky irradiance models by comparison with the RRTMG physical radiative transfer model, acting as a benchmarking reference, for hourly GHI and DNI over a whole year are evaluated in [
13], while the differences induced in the hourly and daily GHI predictions by the mesoscale atmospheric weather research forecasting model in Greece when using different shortwave radiation are assessed in [
14]. Likewise, hourly GHI in the Arabian Peninsula using a three-dimensional meteorology–chemistry model including a state-of-the-art prognostic treatment of aerosols is simulated in [
15]. In another prediction methodology vision, the performance of an exponential smoothing model with decomposition methods to improve its hourly GHI forecasting accuracy and computational efficiency is analyzed in [
16], while prediction intervals for DNI estimates from GHI observations at the minute scale over the Korean Peninsula using the Engerer model under a probabilistic approach are calculated in [
17]. Furthermore, a hybrid convolutional neural network–long short-term memory model with spatiotemporal correlations to improve the accuracy of short-term GHI prediction for ensuring the optimum utilization of photovoltaic power generation sources is proposed in [
18].
In turn, the precision of short-term forecasts for GHI and DNI of a global numerical weather prediction (NWP) model in Portugal is evaluated in [
19], while the performance of three NWP models in forecasting daily GHI for Australia is assessed in [
20]. In another forecasting methodology approach, the use of different machine learning techniques for deterministic and probabilistic GHI and DNI forecasts using local irradiance data and sky images with forecasting time horizons ranging from 5 min up to 30 min is evaluated in [
21]. Likewise, the integration of different forecasting models, by means of machine learning techniques, to improve the short-term predictions of GHI and DNI with a forecast horizon of 6 h and a temporary resolution of 15 min in the Iberian Peninsula is studied in [
22], while a benchmarking of different machine learning techniques for intra-day GHI forecasting from 1 h to 6 h ahead in an insular context is proposed in [
23]. An artificial neural network (ANN)-based algorithm to improve GHI forecasts obtained from the NWP model of the European Centre for Medium-range Weather Forecasts (ECMWF) with a time horizon of 72 h and a time-step of 30 min is developed in [
24]. Similarly, ANN models to produce hourly GHI forecasts from 1 to 6 h ahead are designed in [
25,
26], using exogenous (satellite and NWP model of the ECMWF) and ground data, and ground measurement and satellite data, respectively. In the same way, six ANN models to estimate the monthly mean daily GHI in different locations of the UAE are developed in [
27]. A nonparametric method, based on k-means algorithm, for ultra-short-term forecasting of GHI to deliver predictions with a forecast horizon from 500 ms to 5 min under a probabilistic perspective is assessed in [
28]. Similarly, a forecast methodology based on the k-nearest neighbours algorithm for intra-hour GHI and DNI with horizons ranging from 5 min up to 30 min using ground telemetry and sky images is proposed in [
29]. In turn, a Gaussian process regression method for GHI forecast horizons from 30 min to 5 h is modelled in [
30].
With a larger forecasting term, the daily GHI is predicted using ANN models for 25 Moroccan cities in [
31], with empirical and machine learning models for 5 Moroccan cities in [
32] and with hybrid ARIMA–ANN model for 3 cities in Morocco in [
33]. The daily GHI is also forecasted with ANN models for 35 Moroccan, Algerian, Spanish and Mauritian cities in [
34] and the monthly mean daily GHI using time series models in [
35].
Although hourly and sub-hourly solar irradiation data are essential for an accurate techno-economical assessment of a CSP or PV project, a pre-feasibility study using monthly solar irradiation data is a common practice, and it is usually performed for the selection of potential sites where the power plant is expected to be located [
36,
37,
38]. GHI and DNI are the types of data used for evaluating these projects. For the assessment of monthly data, ANN models are used to estimate it in Saudi Arabia [
39,
40] and in Uganda [
41].
When it comes to the design of energy projects that involve GHI and DNI, it is important to test their model for different solar irradiation profiles to study the performance variations that the solar system could have when it is implemented in real life. Using only one set of data has the limitation that the model works completely fine if the studied conditions are met. These conditions often represent the mean values that the solar installations will experience. However, when the solar system is subjected to different input conditions, the outcome results may vary greatly from the ones projected during its planning and design phase. That is, there will be a chance that the projected profitability is not reached. As such, the variability of solar irradiation can be considered a risk since it can jeopardize the project’s profitability. Since solar systems are going to be implemented increasingly in the future, this variability should be properly taken into account.
Frequently, however, the availability of enough measured data is limited. For this reason, satellite-based data are used to assess the solar resource when the measured data are scarce. References [
42,
43] compared the performance of several satellite data sets with ground measurements data in Morocco and North Africa, respectively. Statistical methods for measuring errors were used in various papers as a way to validate the solar irradiation data. For instance, Aguiar et al. [
44] compared satellite data and ground measurements for various sites in the Canary Islands. Urraca et al. [
45] employed the same validation method for various sites in Europe, detecting operational errors for some Baseline Surface Radiation Network (BSRN) stations. Schumann et al. [
46] undertook data validation through statistical methods for Tamanrasset, Algeria, and Meyer et al. [
47] applied it in Spain. Ineichen et al. [
48] concluded that for European and Mediterranean sites, the irradiance data retrieved from various satellite databases had low uncertainty with a negligible bias when compared to ground measurements for the same locations. For African sites however, the lack of meteorological stations in the region could prove fatal for solar projects evaluation. Because of that, satellite data are one option that most projects decide on. References [
49,
50,
51] showed that the SARAH database is good enough for monitoring and analysis of solar conditions in many sites, especially for Africa. Additionally, Huld et al. [
52] concluded that the PVGIS database is of high quality for PV performance estimates for both Europe and Africa. Finally, in references [
53,
54] it was stated that solar irradiance values have high variability in areas with variable landforms, such as those with mountains and coastal areas.
The main characteristics of interest of the several references considered here are summarized in
Table 1. The references have been listed either according to the long- or short-term character of the solar irradiance forecasting or according to their use of satellite-based data.
1.2. Justification and Objectives
As mentioned earlier, Morocco is currently a key country for solar development, and as a consequence some studies dealing with irradiation prediction and characterization focused on that country can be found in the literature. For instance, El Mghouchi et al. [
55] carried out an assessment on the different solar irradiation prediction models for the northern Moroccan city of Tetouan. In reference [
56], two empirical models were analyzed for 24 different cities across Morocco. Marchand et al. [
42] conducted a study on the variability of solar irradiation data on different locations in the country. Proving the above-mentioned trend of employing satellite data for the solar resource validation, Wahab et al. [
43] compared satellite and ground measured data for northern African countries, including Morocco. Likewise, references [
31,
32,
33,
34,
35] have also been focused on the irradiation forecasting in several cities of Morocco, employing different methodologies.
Given the high potential for solar energy development in the MENA region and, particularly in Morocco, this paper will focus on this country to ease the deployment of solar energy projects.
As concluded in the above paragraphs, the prediction and characterization of solar irradiation using satellite data provides accurate and satisfactory results. Moreover, as highlighted in
Section 1.1, state of the art, the prediction and characterization of solar irradiation rely mostly on either the use of complex models (i.e., Collares-Pereira) or on complex mathematical techniques (such as ANN-based algorithms). This mathematical complexity might hamper their use by businesses and project developers when assessing the solar resource.
In this regard, the first objective of the paper is to introduce a comprehensive and simple methodology to assess the solar resource for a project, based on the determination of its probability distribution function (PDF) for a specific location, assuming that the knowledge of statistical techniques may be more widely extended than other more complex mathematical techniques.
A second objective of the paper is to illustrate the selection of the best satellite database available for a determined location, focusing on the case of several cities across Morocco.
To the best of the authors’ knowledge, no studies have been found where a new database for solar irradiation values that depends on historical data is provided for Morocco. Thus, it is considered to be the first study covering this gap in the scientific literature. This methodology is easily replicable to other locations, and it might be a useful tool to provide reliable solar irradiation data for future solar project assessments.
The paper is organized as follows. First, a general overview of the employed methodology is given in
Section 2. Then, the selected case study is defined, and the data collection process is explained, as well as the validation and quality control phase, in order to use the best possible set of satellite data. Once the definitive set of data is prepared, distribution -fitting tests are used and the frequency of the irradiation values for every hour in a typical day of each month is presented in
Section 3. In
Section 4, the discussion on the results obtained is undertaken followed by a discussion on the potential applications that this methodology has. Finally, conclusions are raised in
Section 4.