Short-term irradiance forecasting using skycams: Motivation and development

Abstract This paper examines the motivation, applications and development of short-term solar forecasting using ground-based sky imagery for controlling equipment on electrical grids. Historically, there has not been a great deal of interaction between the fields of solar forecasting and electrical grid research. This situation is changing rapidly as solar forecasting is becoming increasingly important for dealing with the mass uptake of photovoltaics (PV) and solar-thermal generation on electrical grids around the world. The interactions in these two fields is examined, along with the opportunities for applying solar forecasting for on-grid and mini-grid applications. We review solar forecasting techniques, summarise their applications and evaluate the links to suitable techniques with applications. A review of current solar forecasting techniques, including Numerical Weather Prediction (NWP), statistical/data-driven approaches, and satellite techniques is presented. We also compare the characteristics of each technique to illustrate the requirements, status and suitability of each for use in short-term solar forecasting applications. The application of sky-camera (‘skycam’) forecasting in electrical grids is discussed in addition to the presentation of a detailed case study demonstrating the use of these techniques to enhance grid operation. The control strategy developed demonstrates an increase the real-time penetration of renewable power by dynamically adapting solar inverter setpoints according to cloud forecast data. A novel short-term solar forecasting system is presented which makes use of inexpensive ground-based sky imaging cameras (or ‘skycams’). This system is able to predict changes in irradiance by forecasting cloud movement up to 20 min ahead, with a 10 s update frequency. Several novel techniques for skycam setup and forecasting are presented. These include: (a) A new, high-performance approach to cloud classification using a novel set of neural network input features. (b) A new method for calibrating a lens distortion model. (c) A novel technique using per-pixel cloud movement vectors to predict the timing and extent of sun shading events. This latter technique is capable of correctly classifying 97% of cloud pixels from a validation database of over 500,000 examples. Finally, we present a new technique for taking features extracted from sky-camera pixel data and building a model for predicting large shading events. We show an example of how this model can be easily adapted for either conservative or aggressive operation of a solar power system with a backup generator. In conservative operation, this model is shown to supply adequate or excess energy up to 99.96% of the time using 4 min-ahead forecasts when used for scheduling a backup electrical generator; meaning the system would require only minimal battery storage, while producing a large reduction in fossil-fuel consumption.

[1]  G. Strbac,et al.  Trading Wind Generation in Short-Term Energy Markets , 2002, IEEE Power Engineering Review.

[2]  J. Shields,et al.  The Whole Sky Imager - A Year of Progress , 1998 .

[3]  J. Kleissl,et al.  Intra-hour forecasting with a total sky imager at the UC San Diego solar energy testbed , 2011 .

[4]  Sylvain Paris,et al.  SimpleFlow: A Non‐iterative, Sublinear Optical Flow Algorithm , 2012, Comput. Graph. Forum.

[5]  F. Rahimi,et al.  Transactive Energy Techniques: Closing the Gap between Wholesale and Retail Markets , 2012 .

[6]  Glenn Platt,et al.  Renewable energy aggregation with intelligent battery controller , 2013 .

[7]  John Pye,et al.  Cloud tracking with optical flow for short-term solar forecasting , 2012 .

[8]  Richard Perez,et al.  Modeling PV fleet output variability , 2012 .

[9]  Teodoro López-Moratalla,et al.  Computing the solar vector , 2001 .

[10]  Gunnar Farnebäck,et al.  Two-Frame Motion Estimation Based on Polynomial Expansion , 2003, SCIA.

[11]  J. Schoene,et al.  Solar forecasting and variability analyses using sky camera cloud detection & motion vectors , 2012, 2012 IEEE Power and Energy Society General Meeting.

[12]  M. J. D. Powell,et al.  Direct search algorithms for optimization calculations , 1998, Acta Numerica.

[13]  Gary R. Bradski,et al.  Motion segmentation and pose recognition with motion history gradients , 2002, Machine Vision and Applications.

[14]  Takeo Kanade,et al.  An Iterative Image Registration Technique with an Application to Stereo Vision , 1981, IJCAI.

[15]  C. Coimbra,et al.  Intra-hour DNI forecasting based on cloud tracking image analysis , 2013 .

[16]  Berthold K. P. Horn,et al.  Determining Optical Flow , 1981, Other Conferences.

[17]  Jan Kleissl,et al.  Towards Intra-Hour Solar Forecasting using Two Sky Imagers at a Large Solar Power Plant , 2012 .

[18]  Luis Marroyo,et al.  Smoothing of PV power fluctuations by geographical dispersion , 2012 .

[19]  Clark W. Gellings,et al.  Integrating Load Management into Utility Planning , 1985, IEEE Power Engineering Review.

[20]  Jan Kleissl,et al.  Solar irradiance forecasting using a ground-based sky imager developed at UC San Diego , 2014 .

[21]  J. Kleissl,et al.  Development of a sky imaging system for short-term solar power forecasting , 2014 .

[22]  Magnus Korpaas,et al.  Operation and sizing of energy storage for wind power plants in a market system , 2003 .

[23]  Martin A. Riedmiller,et al.  A direct adaptive method for faster backpropagation learning: the RPROP algorithm , 1993, IEEE International Conference on Neural Networks.

[24]  Christian Riess,et al.  Representation Learning for Cloud Classification , 2013, GCPR.