No abstract

Abstract:The objective of this study is to examine the conformity of different simulation tools in analyzing the performance of photovoltaic (PV) systems in countries with high solar radiation. Primarily an installed system was evaluated and the results were compared with the simulation results of 3 globally known PV software tools: pvPlanner, PVsyst, and Homer. The parameters evaluated in this study are energy production, specific yield, performance ratio, and capacity factor. Detailed explanations are presented for monthly, seasonal, and annual variation of installed system data and simulation results. Northern Cyprus is selected as a case study due to high solar radiation and duration values.The total annual energy production of the installed 5.76 kW system amounts to 12,216 kWh for the year studied. All the simulation tools appear to underestimate the installed system's energy production and the variances observed are 5.3%, 9.3%, and 7.5% for pvPlanner, PVsyst, and Homer, respectively. Energy production in summer was observed to be about twice the production in winter. The percentage shares with respect to energy production are 34%, 28%, 22%, and 16% in summer, spring, autumn, and winter, respectively. The performance ratio of the system is 80.8%. However, the average performance ratio of the 3 simulators was found to be 78.6%. PVsyst modeled a performance ratio with the least deviation from the system with 79.2%. The specific yield and capacity factor of the installed system are 2121 kWh/kW p and 25.06%, respectively. The average specific yield value and average capacity factor of the 3 simulators are nearly 7% lower than the measured data of the installed system. Different factors led to the difference between real-world application and simulation results. These are discussed in this study in detail.

No abstract

No abstract

The objective of this study is to develop a model that can predict the specific yield of PV plant and statistically examine the significance of independent variables during the process. The methods presented will help investors make quick and reliable decisions about installing PV plant at a particular site. To determine the best‐fit model for energy prediction, data from all variables were collected and a relevance check was conducted by examining the relationship between independent variables and dependent variable. The forecasting method makes use of Multiple Linear Regression model. Seven models were developed and two models with the best accuracy are proposed in this article. Variance Inflation Factor, which assesses potential multicollinearity among independent variables, was used to optimize the models that have similar characteristics. The first proposed model has coefficient of multiple determination of goodness of fit (R2) of 96.8%. The second model has R2 of 96%. The models were applied to data series selected from same site; prediction accuracies of 97% and 95% were obtained, respectively. They were also validated with Middle East Technical University North Cyprus Campus PV plant, yielding prediction accuracies of PV specific yield of 92% and 90%, respectively. It could be concluded that the results presented in this study will guide the energy experts about the reliability of the methods used to analyze the technical feasibility of medium or large scale PV projects. © 2018 American Institute of Chemical Engineers Environ Prog, 38:e13098, 2019

The paper presents an entropy generation minimization study for a solar parabolic trough collector (PTC) operating with SiO2–water nanofluid using a genetic algorithm (GA) and artificial neural network (ANN). The characteristic variables of nanoparticle volumetric concentration (0.01 ≤ φ ≤ 0.05), mass flow rate (0.1 ≤ ṁ ≤ 1.1 kg/s), and inlet temperatures (350–550 K) are used to analyze the rate of entropy generated in the PTC. GA is used in optimizing the entropy generation rate for the specified parameters, while ANN is used for predicting and observing the behavior of these parameters on the rate of entropy generation in the collector. The optimum ANN model is derived with one hidden layer of 18 neurons when training the input variables for the entropy generation predictions. The optimal mean square error used as a performance validation of the model is 0.02288 for training and 0.0282 for testing with an R2 value of 0.9999. The impact of the defined parameters on the entropy generation rate is presented in Sec. 5. It is concluded that machine learning techniques can be an efficient tool for predicting the rate of entropy generation in a collector within the constraint of the defined parameters.

No abstract

No abstract