However, in a market like India, with such impressive and ambitious solar integration targets, traditional (non-modern) solar forecasting technologies are poorly suited. Over the past decade, day-ahead forecasting for unit commitment and hourly forecasting for load balancing across a given day in regions like Europe and the United States have been the status quo. For these purposes, modelling approaches based on numerical weather models and machine learning techniques have been suitable.
However, despite the credentials of forecasting providers from these regions, such traditional approaches are far from suitable for making the short-term predictions of solar farm power output on the timescale of minutes to 1-2 hours ahead that a high-penetration market like India will soon require.
This is driven by one fundamental reason: such methods do not actually know where the clouds actually are at any given time. Contrary to popular belief, weather models do not actually model cloud conditions directly, they instead rely on parameterisations and other tricks to handle them at a broad scale. Even with the best machine learning methods available, if the forecasting approach doesn’t actually use knowledge of actual local cloud cover conditions, it will fail to make accurate short-term forecasts.
Satellite-base forecasting is essential, but not all approaches are suitable
At the forecasting horizons of greatest interest to the Indian market (minutes to hours ahead), satellite based imagery and forecasting can offer significant improvements over the aforementioned ‘traditional’ methods. Weather satellites are placed in geostationary orbits where they record images of the Earth, which includes the surface and cloud cover. These images update every 10-15 minutes, and are available at resolutions as fine as 1km2, which is exactly the kind of data input short-term solar forecasting methods should be utilising.
Yet, not all satellite based forecasting services are the same. Just because one can pull in this imagery data, does not actually mean they are able to identify clouds or track their motion with precision, let alone model the amount of solar radiation arriving at the Earth’s surface or the eventual power output from a solar facility. Furthermore, nearly every satellite based forecast methodology in practice oversimplifies the processes of cloud motion, extrapolating a motion vector from previous cloud movement and projecting the cloud positions forward all as one mass, in the same direction, to make a forecast. This is simply not how clouds behave, as each layer of the lower atmosphere, where clouds reside, often move in different directions due to a phenomena known as ‘wind shear’ (the change in cloud speed and direction with height).
Modern short-term solar forecasting approaches
To properly identify a suitable *modern* solar forecasting approach for short-term predictions of utility scale solar farm power output, there are three key requirements.
First, the solar forecast should be a rapid update, satellite based service built with meteorological expertise. By rapid update, I mean that solar forecasts are continually being regenerated based on the latest information from the satellite imagery (where the clouds are) and the actual generation statistics from the solar facility. For example, at my company Solcast, we track the actual locations and estimate the characteristics of cloud cover in real-time, over every continent except Antarctica. At any given time, we know where all of the clouds near your solar facility are, how thick they are with respect to sunlight, and then forecast of where they’ll move next, by leveraging wind velocity data from high-resolution weather models. With each satellite scan, we regenerate an entirely new set of forecasts. That’s rapid update. You’d be surprised how few solar forecasting providers actually do this.
Secondly the forecasts should be delivered via API. An API is an Application Programming Interface, which is the modern protocol for securely and quickly transmitting data between cloud based computing platforms and devices/facilities operating across the world. Information packages are encrypted, and completely secure against cyber-intrusion, so long as the user keeps their API keys confidential. At Solcast, we deliver all of our forecasting services via a REST API running on the latest generation of Microsoft’s. NET Core technology, making the data instantly available via HTML, JSON, XML, csv and jsv formats. This cloud based approach also comes with the added capabilities of guaranteed uptime, including redundancy across multiple service regions. With an API, you get all the benefits of the cloud, including speed, safety, reliability and easy integration. Anyone using FTP servers or email downloads to deliver your forecasts, is simply not operating a modern architecture, nor implementing best practice with regard to the safety of your data.
Third, the forecast needs to be using the latest generation data from your utility scale solar farm site, in combination with satellite imagery. Satellite data is absolutely required for the identification of cloud features before they arrive at your solar facility. However, even the best satellite based forecasting approaches aren’t perfect at timing cloud cover, or capturing the development of small convective cloud cover, or estimating the thickness of the local cloud cover. For this reason, the actual measurement data from the utility scale solar farm site needs to be a part of the forecast. Fortunately, any modernised service which utilises APIs, will offer two-way communications (unlike FTP based solutions), which allow the forecast customer to POST (send to the API) the SCADA measurements from the solar farm site to the API.
At Solcast, we use this information to ‘tune’ the PV power forecast according to the individual characteristics of the site. Our PV Tuning technology learns the real characteristics of your utility scale PV plant, by using your measured power output data. Using either a historical extract, or a real-time feed of your SCADA data, we match up the performance of your site with our solar irradiance and weather data, to produce a ‘tuned’ forecast that represents your specific facility. This is not only important because of the limitations of satellite based approaches, but also due to the complexity of utility scale solar installations! Even with perfect execution against a construction plan, once commissioned, solar facilities encounter a host of real-world challenges that can impact performance in many ways.Tuning captures these complexities!
In closing, India has shown great leadership with its aggressive solar installation targets. With its state distribution companies now due to enforce penalties for inaccurate solar forecasts, it would be quite wise to also show leadership in its implementation of its solar forecasting technologies and network/market operations by choosing modern approaches.
About the Author: Dr. Nick Engerer