|Myers, Daryl -|
|Boyson, William -|
Submitted to: Solar Energy
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: March 12, 2012
Publication Date: April 1, 2012
Citation: Vick, B.D., Myers, D., Boyson, W. 2012. Using direct normal irradiance models and utility electrical loading to assess benefit of a concentrating solar power plant. Solar Energy. 86:3519-3530. Interpretive Summary: Power plants using natural gas and coal generate 70% of the electricity in the United States. Natural gas is expected to last 70 more years, and coal is predicted to last 100 to 200 more years. Renewable energy resources like wind and solar energy will be available as long as the wind blows and the sun shines. Concentrated solar power (CSP) plants had the lowest cost of energy of renewable energies for utility scale systems at sunny, low wind, locations until a few years ago. The price of solar photovoltaic (PV) modules has decreased about 50% in the past few years which has led to competition with CSP for MW size systems. Accurate global solar radiation data needed to predict solar PV output is available for most locations in the U.S. However, accurate solar resource data required for CSP systems is only available at a few locations in the U.S. Direct normal irradiance (DNI) data are needed in order to predict the power output of a CSP plant. The instruments and sun tracking equipment needed for measuring DNI data is expensive and require daily maintenance (e.g. more expense). Therefore, estimating DNI with DNI models is desirable. Three DNI models were evaluated: DISC, DIRINT, and DIRINDEX. The DNI prediction of the models was compared to measured DNI data at three labs. The labs were located at: Bushland (Texas), Albuquerque (New Mexico), and Golden (Colorado). Utility electrical loading data was also gathered at these three labs. The utility loading data was needed to determine the accuracy of the DNI models when the utility loading was greatest. Annually, two of the three DNI models (DIRINT and DIRINDEX) predicted the direct beam solar energy within 5% of the measured values at all three locations. During months of greatest utility loading, the DIRINT and DIRINDEX models' predictions differed by less than 5% from the measured values at the Texas lab. During the high utility load months, the accuracy of these DNI models for the New Mexico and Colorado labs varied from 6 to 11%. On both clear and partly cloudy peak utility loading days, the DNI was modeled well with the DIRINT model at the Texas lab. On partly cloudy peak utility loading days, the DNI was not modeled well at the other two labs. On an annual, monthly, and hourly basis, the DIRINT model could be used to assess the benefit of a CSP plant at the Texas lab location. However, the DNI models need to be improved at the other two lab locations (New Mexico and Colorado) in order to assess the benefit of a CSP plant. Ways to possibly improve the DNI models were also discussed. If the DNI models can be improved, their use in other locations in the U.S. will allow electrical utilities to assess the benefit of CSP plants, and more of these plants may be constructed. Therefore, improving the DNI models could help transition the U.S. from using finite energy resources like natural gas and coal to the use of a renewable energy resource (e.g. solar).
Technical Abstract: Direct normal irradiance (DNI) is required to evaluate performance of concentrating solar energy systems. The objective of this paper is to analyze the effect of time interval (e.g. year, month, hour) on the accuracy of three different DNI models. The DNI data were measured at three different laboratories in the United States and compared with that calculated by three different DNI models. The DNI models evaluated were: the Direct Insolation Simulation Code (DISC), DIRINT, and DIRINDEX (both DIRINT and DIRINDEX are modified versions of DISC that require additional input variables and other models/computer programs). On an annual solar insolation (e.g. kWh/m**2) basis, the three DNI model predictions were within 7% of measured DNI at all three locations. Since the DISC model is the easiest to use, DISC would likely be the best model to use for a preliminary annual DNI solar resource assessment. The annual root mean square errors (RMSE) for DIRINT and DIRINDEX were about 50 W/m**2 less than those of the DISC method at all three locations, but the annual mean bias errors (MBE) differed for the different DNI models at the three locations. On a monthly basis, the direct normal insolation prediction could range from 10 to 20% different than measured for all three models, but the DIRINT and DIRINDEX models were usually more accurate than DISC in modeling DNI diurnally. Also, the monthly mean insolation of the DISC method can be more accurate than that of DIRINT or DIRINDEX models due to compensating errors in the DISC model. The DISC model often inaccurately modeled the first hour after sunrise and the first hour before sunset, but correcting this would likely result in a more inaccurate annual, monthly, and daily insolation prediction. The monthly RMSE of DNI varied from: 111 to 205 W/m2 for the DISC model, 65 to 116 W/m2 for the DIRINT model, and 66 to 117 W/m2 for the DIRINDEX model. Although DIRINDEX is the model used predominantly in DNI solar resource assessment, the DIRINT model for the three locations evaluated was as good or better than the DIRINDEX model, and the DIRINT model was easier to use.