We present a new model for the calculation of the diffuse fraction of the global solar irradiance for solar system simulations. The importance of an accurate estimation of the horizontal diffuse irradiance is highlighted by findings that an inaccurately calculated diffuse irradiance can lead to significant over- or underestimations in the annual energy yield of a photovoltaic (PV) system by as much as 8%. Our model utilizes a time series of global irradiance in one-minute resolution and geographical information as input. The model is validated by measurement data of 28 geographically and climatologically diverse locations worldwide with one year of one-minute data each, taken from the Baseline Surface Radiation Network (BSRN). We show that on average the mean absolute deviation of the modelled and the measured diffuse irradiance is reduced from about 12% to about 6% compared to three reference models. The maximum deviation is less than 20%. In more than 80% of the test cases, the deviation is smaller 10%. The root mean squared error (RMSE) of the calculated diffuse fractions is reduced by about 18%.

This work assesses the performance of five transposition models that estimate the global and diffuse solar irradiance on tilted planes based on the global horizontal irradiance. The modelled tilted irradiance values are compared to measured one-minute values from pyranometers and silicon sensors tilted at different angles at Hannover (Germany) and NREL (Golden, CO, USA). It can be recognized that the deviations of the model of Liu and Jordan, Klucher and Perez from the measurements increases as the tilt angle increases and as the sensors are oriented away from the south direction, where they receive lower direct radiation than south-oriented surfaces. Accordingly, the vertical E, W and N planes show the highest deviation. Best results are found by the models from Hay and Davies and Reindl, when horizontal pyranometer measurements and a constant albedo value of 0.2 are used. The relative root mean squared difference (rRMSD) of the anisotropic models does not exceed 11% for south orientation and low inclination angles (β = 10–60°), but reaches up to 28.9% at vertical planes. For sunny locations such as Golden, the Perez model provides the best estimates of global tilted irradiance for south-facing surfaces. The relative mean absolute difference (rMAD) of the Perez model at NREL ranges from 4.2% for 40° tilt to 8.7% for 90° tilt angle, when horizontal pyranometer measurements and a measured albedo value are used; the use of measured albedo values instead of a constant value of 0.2 leads to a reduction of the deviation to 3.9% and 6.0%, respectively. The use of higher albedo values leads to a significant increase of rMAD. We also investigated the uncertainty resulting from using horizontal pyranometer measurements, in combination with constant albedo values, to estimate the incident irradiance on tilted photovoltaic (PV) modules. We found that these uncertainties are small or negligible.

We present a method to predict the global horizontal irradiance (GHI) one hour ahead in one-minute resolution using Artificial Neural Networks (ANNs). A feed-forward neural network with Levenberg–Marquardt Backpropagation (LM-BP) was used and was trained with four years of data from all-sky images and measured global irradiance as input. The pictures were recorded by a hemispheric sky imager at the Institute of Meteorology and Climatology (IMuK) of the Leibniz Universität Hannover, Hannover, Germany (52.23° N, 09.42° E, and 50 m above sea level). The time series of the global horizontal irradiance was measured using a thermopile pyranometer at the same site. The new method was validated with a test dataset from the same source. The irradiance is predicted for the first 10–30 min very well; after this time, the length of which is dependent on the weather conditions, the agreement between predicted and observed irradiance is reasonable. Considering the limited range that the camera and the ANN can “see”, this is not surprising. When comparing the results to the persistence model, we observed that the forecast accuracy of the new model reduced both the Root Mean Square Error (RMSE) and the Mean Absolute Error (MAE) of the one-hour prediction by approximately 40% compared to the reference persistence model under various weather conditions, which demonstrates the high capability of the algorithm, especially within the first minutes.

We analyze the output of various state-of-the-art irradiance models for photovoltaic systems. The models include two sun position algorithms, three types of input data time series, nine diffuse fraction models and five transposition models (for tilted surfaces), resulting in 270 different model chains for the photovoltaic (PV) system simulation. These model chains are applied to 30 locations worldwide and three different module tracking types, totaling in 24,300 simulations. We show that the simulated PV yearly energy output varies between −5% and +8% for fixed mounted PV modules and between −26% and +14% for modules with two-axis tracking. Model quality varies strongly between locations; sun position algorithms have negligible influence on the simulation results; diffuse fraction models add a lot of variability; and transposition models feature the strongest influence on the simulation results. To highlight the importance of irradiance with high temporal resolution, we present an analysis of the influence of input temporal resolution and simulation models on the inverter clipping losses at varying PV system sizing factors for Lindenberg, Germany. Irradiance in one-minute resolution is essential for accurately calculating inverter clipping losses.

High resolution global irradiance time series are needed for accurate simulations of photovoltaic (PV) systems, since the typical volatile PV power output induced by fast irradiance changes cannot be simulated properly with commonly available hourly averages of global irradiance. We present a two-step algorithm that is capable of synthesizing one-minute global irradiance time series based on hourly averaged datasets. The algorithm is initialized by deriving characteristic transition probability matrices (TPM) for different weather conditions (cloudless, broken clouds and overcast) from a large number of high resolution measurements. Once initialized, the algorithm is location-independent and capable of synthesizing one-minute values based on hourly averaged global irradiance of any desired location. The one-minute time series are derived by discrete-time Markov chains based on a TPM that matches the weather condition of the input dataset. One-minute time series generated with the presented algorithm are compared with measured high resolution data and show a better agreement compared to two existing synthesizing algorithms in terms of temporal variability and characteristic frequency distributions of global irradiance and clearness index values. A comparison based on measurements performed in Lindenberg, Germany, and Carpentras, France, shows a reduction of the frequency distribution root mean square errors of more than 60% compared to the two existing synthesizing algorithms.