188宝金博页面版

Contents lists available at ScienceDirectSolar Energyjournal homepage: www.elsevier.com/locate/solenerIncorporating ground-measured pollution observations to improvetemporally downscaled solar irradiance simulationsXiaofeng Shi a , Brendan Acord b , Peng Wang a, ?a Department of Mathematics and System Sciences, Beihang University, Chinab Syene Clean Energy, Hong KongA R T I C L E I N F OKeywords:Pollution ef f ectSmall-scale PVTemporal downscalingPM 2.5A B S T R A C THigh temporal resolution modeling of solar irradiance is of great interest to stakeholders in the power sector andhelps understand the impact of additional solar photovoltaic (PV) power on the grid. Recent developments insolar irradiance downscaling modeling (SID), such as the Bright model, have facilitated the generation of syn-thetic minutely solar irradiance and shown promising results in U.K. and the U.S. and Australia. However, in thepresence of heavy airborne pollution, the SID performances may be hindered by a data shortage of high-tem-poral-resolution aerosol optical depth (AOD). In this paper, we present a novel approach to incorporate pollutionef f ects into a solar irradiance downscaling model, i.e. the Bright model, by utilizing the local ground-measuredpollution data that is publicly available. The proposed framework is evaluated through a case study in Beijing,China, and shows an improved estimation of high-frequency solar irradiance for areas experiencing heavypollution.1. IntroductionMinutely information on solar irradiance at f i ne temporal scales is ofgreat importance in the solar energy industry. It matters not only in thesimulations of storage, photovoltaic (PV) supply and integrated elec-tricity demand at the grid level (Marcos et al., 2014; Widen et al., 2015;Sayeef et al., 2012; Hansen et al., 2012; Hummon et al., 2012; Cao andSirën, 2014), but also in the management of distributed PV projects,whose number have doubled globally over the last few years (Massonand Brunisholz, 2016).Due to the scarcity of calibrated minutely irradiance data in locationand duration and associated inconsistency in measurements betweensites (Bright et al., 2015c), temporal downscaling of hourly data pro-vides a viable means to synthesize typical meteorological year irra-diance datasets for use in system design and optimization. Recent de-velopments (Bright et al., 2015c, 2017) have shown success in utilizinghourly weather data to generate sub-hourly solar output estimates.Categorized as a “sun obscured” type approach, the Bright model(Bright et al., 2015c, 2017) employs Markov chains to stochasticallydetermine the future states of meteorological condition; and henceobtain the minutely time series of cloud cover (okta number), clear-skyindex and eventually, the global all-sky irradiance. Currently beingadapted to include spatial correlation, such an approach provides asuitable model for small-scale PV systems whose size and connectionscheme to distribution feeders require less geographic smoothing thanutility-scale PV projects, and whose economics do not permit extensivesite resource measurement campaigns during development. However,when assessing the Bright model against ground measured irradiance inChina, the model accuracy was found to be less than at sites used tovalidate the model in England and the United States.One particular dif f erence between the Bright model validation sitesand China is the high level of airborne pollution in the latter. In general,airborne pollutants reduces solar output by soiling the PV panels (Mejiaand Kleissl, 2013), absorbing, scattering and blocking incoming solarradiation (Lopez and Batlles, 2004; Vindel and Polo, 2014). The in-creasing formation of cloud condensation nuclei would reduce clouddroplet ef f ective radius and add up the optical thickness of the clouds,through which light would be scattered, leading to less transmission(Facchini et al., 1999; Scott et al., 2014). Pollutants’ temporal varia-bility could further complicate the estimation of solar irradiance. Asnoted in recent works (Cebecauer et al., 2011; Kleissl, 2013), modelsdriven by daily pollution data (aerosol optical depth) outperform thosedriven by monthly data, since such models can better capture the dy-namic changes in atmospheric transmissivity associated with weatherfronts, pollution, and dust-transport events. We therefore hypothesizedthat accessible ground-measured pollution data could be incorporatedinto the simulation to increase the accuracy of temporal downscalingirradiance using the Bright model.https://doi.org/10.1016/j.solener.2018.06.076Received 17 October 2017; Received in revised form 6 June 2018; Accepted 20 June 2018? Corresponding author. at: Beijing Advanced Innovation Center for Big Data and Brain Computing, Beihang University, China.E-mail address: wang.peng@buaa.edu.cn (P. Wang).Solar Energy 171 (2018) 293–301Available online 30 June 20180038-092X/ © 2018 Elsevier Ltd. All rights reserved.T