Abstract
The establishment of the typical weather conditions of a given locality is of fundamental importance to determine the optimal configurations for solar thermal power plants and to calculate feasibility indicators in the power plant design phase. Therefore, this work proposes a summarization method to statistically represent historical weather data using typical meteorological days (TMDs) based on the cumulative distribution function (CDF) and hourly normalized root mean square difference (nRMSD). The proposed approach is compared with regular Sandia selection in forecasting the electricity produced by a solar thermal power plant in ten different Brazilian cities. Considering the determination of the annual generation of electricity, the results obtained show that when considering an overall average of weather characteristics, commonly used for analyzing solar thermal power plant designs, the normalized mean average error (nMAE) is 20.8 ± 4.8% relative to the use of historical data of 20 years established at hourly intervals. On the other hand, a typical meteorological year (TMY) is the most accurate approach (nMAE = 1.0 ± 1.1%), but the costliest in computational time (CT = 381.6 ± 56.3 s). Some TMD cases, in turn, present a reasonable trade-off between computational time and accuracy. The case using 4 TMD, for example, increased the error by about 11 percentual points while the computational time was reduced by about 81 times, which is quite significant for the simulation and optimization of complex heliothermic systems.