Forecasting solar power generation with 'deep photovoltaic nowcasting'

• The 'deep photovoltaic nowcasting' project established by Chile's Institute of Engineering Sciences of the University of O'Higgins, Canada's Laval University and Japan's Kyoto University, seeks to make short-term, high-resolution forecasts of solar power generation.

Anticipating energy production on a minute by minute or second by second scale –-- short-term or instant forecasting –-- is crucial for handling solar plant operations in a smart energy network as well as for guaranteeing power supply and enhancing system efficiency.

Forecasting systems based on climate designs; satellite data; or predictive designs based upon historic generation data become insufficient due to their low spatial and temporal resolution –-- weather stations are scarce so the closest to a solar panel could be too far away to be trusted, geostationary satellites have fairly restricted image resolution and predictive models based upon photovoltaic panel information are not dependable enough.

That is why Rodrigo Verschae, an academic at Chile’& rsquo; s Institute of Engineering Sciences at the University of O’& rsquo; Higgins; and researchers Jinsong Zhang and Jean-Fran & ccedil; ois Lalonde, from Laval University, Canada; and Shohei Nobuhara, of Kyoto University, Japan established ‘& lsquo; deep photovoltaic nowcasting & rsquo;. The job aims to make short-term, regional, accurate, high-resolution solar project output forecasts.

Nowcasting

“& ldquo; In the future [grid operators] will have to decide in real time, since of the renewable energy that is readily available –-- and making use of other sources of generation and intake –-- who will consume [electrical power] and when, collaborating all the generation units in an automatic way,” & rdquo; said Verschae. & ldquo; In a neighborhood, several homes and industries might be arranged based upon schedule & hellip; that is a fantastic inspiration to perform this work.”

‘& rdquo; & lsquo; Nowcasting & rsquo; on a per-minute or per-second basis uses pictures of the sky recorded by video near solar panels. Nevertheless, estimating weather conditions from such images, including sun intensity and the look and motion of clouds, is an uphill struggle with standard computer system visual techniques.

In their new technique, the worldwide researchers built a system which instantly correlates the relationship between the look of the sky and photovoltaic panel output utilizing ‘& lsquo; deep learning & rsquo; machine learning techniques.

Dataset

“& ldquo; What we propose is to install a camera looking at the sky on the roofing system of a home or building, near solar panels, which takes four images every second to produce a picture of fantastic dynamic range and then a sequence of these images,” & rdquo; Verschae stated. & ldquo; This enables us to see where the sun is, how the clouds move and when they will occlude it, preventing the generation of solar energy.” & rdquo; Based on that details, the scientists were able to predict when there would be less energy output with extremely exact resolution.

The job, in development because 2017, has six months’ & rsquo; worth of information from cams at the University of Kyoto.

“& ldquo; This type of system, in addition to helping to predict solar energy, could be useful for some climate models: forecasting weather in the short-term, or to manage variables in clever structures,” & rdquo; said Verschae. & ldquo; And another thing that intrigues us relates to working with solar plants, not just with panels, and if you install more than one electronic camera, you could, for example, predict which part of the panel is going to be concealed.”