Corrigendum: “A novel estimation approach for the solar radiation potential with its complex spatial pattern via machine-learning techniques” (Renewable Energy (2019) 133 (575–592), (S0960148118312503), (10.1016/j.renene.2018.10.066))

Choongwan Koo, Wenzhuo Li, Seung Hyun Cha, Shaojie Zhang

Research output: Contribution to journalComment/debate

Abstract

The authors regret that some information about the previous studies in the section of literature review has mistakenly been mentioned or missed; and thus, the authors would like to correct. Please note that none of these changes impact on the conclusions made by the paper [1] but are necessary to correct for accuracy. The Supporting Information has also been updated to reflect these corrections. In the Section “1.2. Literature review” of the paper [1], the authors presented several studies that have proposed various methods to estimate the global solar radiation on horizontal surface in places without any observations and classified these methods into deterministic and data-driven models. To make it clear, the authors modify the classification of the methods into three types as follows: In the Section “1.2. Literature review” of the paper [1] Previous studies proposed various methods to estimate the global solar radiation on the horizontal surface in places without no observations. The proposed methods can be classified into three types: (i) deterministic models; (ii) data-driven models; and (iii) both of them (i.e. deterministic and data-driven models) [2,8–24] (refer to Tables 1–3). - Deterministic models (refer to Table 1): Deterministic models are based on physical process and mathematical formulas, explaining the mechanisms of the solar radiation. Some studies improved deterministic models with some modifications of the relevant parameters and clarified the applicability of the developed models in different places [2,8–13]. Zhao et al. [10] improved an Angström–Prescott (A–P) equation by introducing the air pollution index (API) in estimating the daily solar radiation. For nine stations in China, the study defined sunshine hours, geographical position, altitude and API as independent variables that were used to determine the regression parameters of three forms of models (i.e. linear, exponential and logarithmic models). The improvement in the prediction performance of the developed model with API was proven to be more significant in highly-polluted locations. Fan et al. [13] proposed the four types of the sunshine-based estimation models by considering the combinations of input parameters (e.g., sunshine duration, air temperature, precipitation, vapor pressure, and relative humidity). The proposed models combined with sunshine duration were expected to accurately estimate the daily global solar radiation in South China.- Data-driven models (refer to Table 2): Other studies developed data-driven models using various algorithms, as a black-box method, so as to estimate the solar radiation in different regions [14–21]. Chen et al. [14] used the temperature-based support vector machine (SVM) methods to estimate the monthly solar radiation in Chongqing, China. The study attempted to evaluate the various kernel functions and the combinations of input variables. The SVM model with polynomial kernel function using maximum temperature (Tmax) and minimum temperature (Tmin) as input variables was proven to be the most promising method. Wang et al. [19] used three kinds of artificial neural network (ANN) methods (i.e. multi-layer perceptron (MLP), generalized regression neural network (GRNN) and radial basis neural network (RBNN)), in which the routinely-measured meteorological datasets were considered to predict the daily global solar radiation at 12 stations in different climate zones of China. The MLP and RBNN models were finally considered with higher applicability in different climate zones of China.- Deterministic and Data-driven models (refer to Table 3): The other studies compared the prediction performance between deterministic models and data-driven models [22–24]. Zou et al. [23] proposed an adaptive neuro-fuzzy inference systems (ANFIS) as a data-driven model and compared with expanded-improved bristow-campbell model (E-IBCM) and improved yang hybrid model (IYHM) as deterministic models to predict daily global solar irradiance in China. The ANFIS model performed better than the E-IBCM and IYHM in predicting global solar radiation in Hunan province with subtropical monsoon climate. Qin et al. [24] estimated the daily solar radiation at different climate and terrain in China by using the deterministic models (e.g. parameterization radiation model, hourly solar radiation model) and the data-driven model (i.e. ANN method). The prediction accuracy of the deterministic models was determined to be superior to that of the data-driven model.In particular, some information on the following three papers has been modified or supplemented: - First, Zang et al. [2] developed a new empirical model (deterministic model) based on day of the year and compared with other two existing models (based on day of the year), which could estimate the daily global solar radiation on horizontal surface at 35 stations across China using 10 years data-driven from typical meteorological years (TMYs) for the 35 cities which were generated using the Finkelstein–Schafer statistical method. The model based on day of the year is a deterministic one that is independent model where it uses only the number of days of the year to predict solar radiation and used by several researchers [3–7]. Thus, the summarized data about the work of Zang et al. [2], which was listed in Table 2 of the study by Koo et al. [1], was moved from Table 2 to Table 1 as listed in this corrigendum.- Second, Liu et al. [8] used the k-means clustering and support vector machine-genetic algorithm (SVM-GA) methods for solar radiation zoning, but not for estimating the global solar radiation. They estimated the global solar radiation by four existing sunshine-based models and combined the site-specific models with the geographical parameters in general model which previously developed in literature. That is, the developed models were deterministic models (but not data-driven models). Thus, the summarized data about the work of Liu et al. [8], which was listed in Table 3 of the study by Koo et al. [1], was corrected and moved from Table 3 to Table 1 as listed in this corrigendum.- Third, Zou et al. [23] proposed an adaptive neuro-fuzzy inference systems (ANFIS) (data-driven models) to predict daily global solar irradiance in China and compared the developed models with the expanded-improved bristow-campbell model (E-IBCM) and the improved yang hybrid model (IYHM) (deterministic models). However, the information in Table 3 of the study by Koo et al. [1] has only described the details of the deterministic model, where the details of the data-driven model has been missed. Thus, the summarized data about the work of Zou et al. [23], which was listed in Table 3 of the study by Koo et al. [1], was corrected and updated in Table 2 as listed in this corrigendum.Remarks for researchers and readers. All comments and corrections that listed in this corrigendum do not influence the obtained results of Koo et al. [1] and the main purpose of this work is to illustrate the missed and mistakenly mentioned information of some literature, which makes the scientific data and literature of the mentioned work clear and reliable. Additionally, from the previous discussion, it would be helpful for the potential readers and researchers to follow the literature and be up to date with the publication in the subject of interest. Table 1. Literature review on solar radiation prediction models (Deterministic models)

Original languageEnglish
JournalRenewable Energy
DOIs
StateAccepted/In press - 2020 Jan 1

Fingerprint Dive into the research topics of 'Corrigendum: “A novel estimation approach for the solar radiation potential with its complex spatial pattern via machine-learning techniques” (Renewable Energy (2019) 133 (575–592), (S0960148118312503), (10.1016/j.renene.2018.10.066))'. Together they form a unique fingerprint.

  • Cite this