Solar forecasting solutions provide utility companies with predictions of power output from large-scale solar installations or from distributed solar generation with a time scale ranging from the next few minutes up to several days ahead. These predictions decrease the risk associated with bidding renewable electricity to the regional grid. Increasing solar photovoltaic efficiency and decreasing manufacturing costs have driven solar electricity generation to become the fastest growing form of renewable electricity production. Adding solar generation in large quantities to the aging electricity grids of the world poses a problem due to the variability and intermittency of solar irradiance. The current state-of-the-art in solar forecasting is focused on the hour-ahead and day-ahead time horizons using publicly available satellite imagery or numerical weather prediction models. Conventional intra-hour forecasting methods are based on sky imagery and basic image processing and computer vision techniques. This thesis discusses the architecture of an intra-hour forecasting tool and outlines the steps involved in taking a sky image and outputting a value of irradiance at specified intra-hour intervals. The thesis includes technical discussions on obstruction masking, geometric transformation, cloud-motion detection and ray tracing for irradiance estimation. The goal is to improve and enhance conventional techniques with innovative approaches to intra-hour solar forecasting. The forecasting tool provides predictions of irradiance and the associated uncertainty through the use of a novel irradiance estimation algorithm and a Monte Carlo simulation. The ray tracing procedure allows for multiple irradiance estimations to be made at spatially distributed points, providing a high-fidelity irradiance mapping of the area within the range of the sky imager. This map can be used to accurately estimate power output from large scale solar power plants or distributed solar generation sites.

Very short-term solar forecasts based on sky images of ground-based cameras introduce a new forecasting methodology for solar energy applications which covers forecast horizons up to 30 minutes. In this thesis, a newly developed image-based forecasting model for the usage in different applications is presented. The core components of the model are cloud detection, cloud motion tracking, cloud shadow projection and irradiance modelling. The model takes raw camera images, local solar irradiance measurements and cloud base height estimations as input data and provides estimations of near-future surface solar irradiance distribution. Large data sets comprising sky images, pyranometer measurements and in some cases cloud base height and PV power measurements are the basis for an in-depth analysis of its forecast performance. Model forecasts are compared with persistence forecasts to evaluate its skill. engl.

Short-term solar forecasting models based solely on global horizontal irradiance (GHI) measurements are often unable to discriminate the forecasting of the factors affecting GHI from those that can be precisely computed by atmospheric models. Our previous study introduced a Physics-based Smart Persistence model for Intra-hour forecasting of solar radiation (PSPI) that decomposed the forecasting of GHI into the computation of extraterrestrial solar radiation and solar zenith angle and the forecasting of cloud albedo and cloud fraction. The extraterrestrial solar radiation and solar zenith angle were accurately computed by the Solar Position Algorithm (SPA) developed at the National Renewable Energy Laboratory (NREL). A cloud retrieval technique was used to estimate cloud albedo and cloud fraction from surface-based observations of GHI. With the assumption of persistent cloud structures, the cloud albedo and cloud fraction were predicted for future time steps using a two-stream approximation and a 5-minute exponential weighted moving average, respectively. The model evaluation indicated the estimation and forecast of cloud fraction mostly contributed to the uncertainty of the PSPI though it overcame the persistence and smart persistence models in all forecast time horizons between 5 and 60 minutes. This study aims to enhance the PSPI by ingesting surface-based observations of cloud fraction from a total sky imager (TSI). The estimation and forecast of cloud albedo is correspondingly improved by utilizing the cloud fraction observations and thus leads to more accurate GHI forecast. Various time-series analysis methods are also investigated on the forecasting of cloud fraction and cloud albedo for further improving the GHI forecast. These improvements are valuable for many applications, such as forecasting energy use for buildings, grid operations, and ultimately bringing down the cost of solar energy.

A very short term solar power forecasting technology which uses ground-based visible wavelength imagery is presented. A sky camera system suitable for use as a solar power forecasting tool is described. Relevant imaging considerations are discussed, including the need for high dynamic range imaging of the daytime sky and an associated stray light assessment. To photogrammetrically calibrate this sky camera system, a general camera model applicable to a fixed focal length photo objective lens with significant radially symmetric distortion is developed, and an accurate calibration technique for a stationary, skyward pointing daytime camera using the sun's position is given. Remote sensing algorithms used in the solar forecasting process are detailed, including clear sky characterization, cloud detection, cloud velocity estimation, and cloud height estimation using stereography. A cloud stereo photogrammetry method which provides dense 3D cloud position is presented. Correspondence is automatically determined using intra-scanline dynamic programming applied to a normalized cross correlation matching metric; an ordering constraint is implicit in the approach used. Using the described remote sensing tools and methods, a complete solar power forecasting framework is detailed. The method is based on the estimation of cloud shadow position via ray tracing, and the forecast position of the cloud shadows relative to solar collectors. A ray tracing procedure that works with a planar mapping of cloud position is used to compute shadow position. Cloud transmissivity is characterized using past observations and applied to forecast cloud positions. The application of the procedure to two case studies: the UCSD DEMROES weather station network, and a 48MW solar photovoltaic power plant is presented. A comparison of the forecasting performance using a common Total Sky Imager is compared to the UCSD Sky Imager, where it is shown that the UCSD Sky Imager performs better overall.

Solar generation is an increasing part of the energy portfolio in the United States. An accurate forecast of the available solar resource and power is essential to managing the electric grid, market operations, and reducing the cost of solar energy. High-frequency forecasts of solar radiation in intra-hour horizons is important for real-time electric power system energy management, especially at the distribution level. Conventional Numerical Weather Prediction models perform poorly in intra-hour, high-frequency forecasts because of the limits on real-time computing, spatial resolution, and infrequent availability of observations. Although a number of alternative technologies, e.g., time-series analysis and machine learning, have been used to fill this gap, the smart persistence model is among the top-performing models in short-term forecasting and therefore often serves as the baseline to evaluate other forecasting models. Although the smart persistence model often serves as the baseline model in these intra-hour forecasts, obvious uncertainties exist in the current smart persistence model: (1) clear-sky index does not respond to the variation of the solar incident angle when cloud conditions are persistent within the forecast horizon, and (2) cloud coverage is inherently persistent though it is constrained by cloud advection. In this study, we developed a Physics-Based Smart Persistence Model for Intra-Hour Solar Forecasting (PSPI) that integrates cloudy property estimation, a radiative transfer model, and cloud fraction forecasts to improve the performance of the smart persistence model. Compared to the smart persistence model, PSPI does not require additional observations of various atmospheric parameters past global horizontal irradiance, but it is customizable because additional observations, if available, can be ingested to further improve the forecast. Our results show that the PSPI outperforms the persistence and smart persistence model on 5-minute, 15-minute, and 30-minute forecast horizons. The software package of PSPI is flexible to users' needs and provides low computational time to run at site-specific locations across the continental United States.

This research was performed to increase the optimization of hour-ahead hybrid solar irradiance prediction methods through the design and implementation of a new hybrid system and model that utilized sky images as a replacement variable for cloud types identified by overhead satellites. Improving current solar radiation prediction methods will benefit electricity producers that need to better understand the availability of solar irradiance and how it may impact their forecasting. Phase one outlines a comparison between current supervised learning methods and deep learning methods. Recurrent Neural Networks produced lower RMSE and higher R2 values and outperformed supervised learning methods. Phase two outlines building and validating a new one-hour ahead hybrid prediction model by combining a deep learning approach with a replacement feature derived from real-time image collection and location specific numerical weather features. This replacement feature was the percentage of sky cover from an observation point on the ground and was called Sky Types. Sky Types are less expensive to obtain and can be collected at any location, which also makes the prediction of solar irradiance specific to the same location. Deep learning models validated that the use of Sky Types was not only a valid substitution for cloud types, but were more optimal for training as model performance improved with reduced network topology and while still being optimal for hour-ahead predictions. To use Sky Types in the new hybrid prediction model, a system was designed to collect the sky condition information from the National Weather Service and relevant images representing the sky condition; both were captured at the same time intervals. Phase three outlines the creation of a system used for collecting images and weather data, preparing images for use by the new hybrid prediction model and building a classification model using a Convolutional Neural Network. The system's GHI predictions were validated using hour-ahead ground truth solar irradiance amounts from ten locations and averaged an RMSE of 41.26 W/m2 and outperformed GFS forecasted GHI by 32% on highly variable weather days. This new hybrid system can be used anywhere numerical weather data and sky images can be captured.

Accurate solar radiation knowledge and its characterization on the Earth’s surface are of high interest in many aspects of environmental and engineering sciences. Modeling of solar irradiance from satellite imagery has become the most widely used method for retrieving solar irradiance information under total sky conditions, particularly in the solar energy community. Solar radiation modeling, forecasting, and characterization continue to be broad areas of study, research, and development in the scientific community. This Special Issue contains a small sample of the current activities in this field. Both the environmental and climatology community, as the solar energy world, share a great interest in improving modeling tools and capabilities for obtaining more reliable and accurate knowledge of solar irradiance components worldwide. The work presented in this Special Issue also remarks on the significant role that remote sensing technologies play in retrieving and forecasting solar radiation information.