The term "[[Nowcasting]]"{{dn|date=June 2019}} in the context of solar power forecasting, generally refers to same spatial and temporal scales as [[Nowcasting (meteorology)|meteorological Nowcasting]], which focuses on forecast horizons from a few minutes ahead, out to 4-6 hours ahead. In general, the 'Nowcasting' forecast horizon refers to those not well-served by global numerical weather prediction models, which produce outputs at up to hourly resolutions and only update every 6 hours. [[Solar power]] nowcasting then refers to the prediction of solar power output (or energy generation) over time horizons of tens to hundreds of minutes ahead of time with up to 90% predictability.<ref>{{cite web |last1=Vorrath |first1=Sophie |title=New APVI solar tool shows daily, time-based forecast for each state |url=https://reneweconomy.com.au/new-apvi-solar-tool-shows-daily-time-based-forecast-for-each-state-52445/ |website=RenewEconomy |language=en-AU |date=31 May 2019}}</ref> It has historically been very important for [[Electrical grid|electrical grid operators]] in order to guarantee the matching of supply and demand on energy markets. Such solar power nowcasting services are usually related to temporal resolutions of 5 to 15 minutes, with updates as frequent as every 5 minutes. The regular updates and relatively high resolutions required from these methods require automatic weather data acquisition and processing techniques, which are chiefly accomplished by two primary means<ref>{{Cite web|url= https://www.elsevier.com/books/solar-energy-forecasting-and-resource-assessment/kleissl/978-0-12-397177-7 |title=Solar Energy Forecasting and Resource Assessment - 1st Edition|website=www.elsevier.com |access-date=2019-05-08}}</ref>:
#'''Statistical techniques.''' These are usually based on [[time series]] processing of measurement data, including [[Surface weather observation|meteorological observations]] and power output measurements from a solar power facility. What then follows is the creation of a [[Training, validation, and test sets|training dataset]] to tune the parameters of a model (I. Espino eta al, 2011), before evaluation of model performance against a separate testing dataset. This class of techniques includes the use of any kind of statistical approach, such as [[Autoregressive–moving-average model|autoregressive moving averages]] (ARMA, ARIMA, etc), as well as machine learning techniques such as [[Artificial neural network|neural networks]], [[support vector machine]]s (etc). These approaches are usually benchmarked to a persistence approach in order to evaluate their improvements. This persistence approach just assumes that any variable at time step t is the value it took in a previous time.
