State-of-the-art solar PV generation forecast for individual PV systems

Summary: impacts and findings

The project aimed to develop a state-of-the-art open-source method for forecasting power generation from individual solar photovoltaic (PV) systems. This will help solar farms, solar storage, and smart homes to optimise and manage their systems, reducing both costs and CO2 emissions.

Project aims and approach

Most of the forecasts currently available rely solely on numerical weather predictions, do not make use of all the data available and struggle to forecast short time horizons ahead, which are important when making operational decisions on assets.

The project aimed to develop a state-of-the-art open-source method for forecasting power generation from individual solar photovoltaic (PV) systems. The main output is an operational, open-source solar electricity forecasting system capable of producing forecasts for thousands of individual solar PV systems. The solution uses machine learning, incorporates real-time satellite imagery, and takes into account the unique properties of each individual PV system. The outputs are provided to users via an easy to use interface.

You can watch a short ‘show and tell’ presentation about this project below, recorded in September 2022. Start from timestamp 08:47.

Partners

Open Climate Fix 

Dates

July 2022 to March 2023

Achievements and barriers

The main technical challenge was to build a state-of-the-art forecasting algorithm using cutting-edge machine learning techniques and huge quantities of data – more than ten years’ PV data from thousands of sites as well as satellite data.

With this project Open Climate Fix has been able to deploy the first version of its site-specific solar forecast, to create a desktop and mobile version of the user application, and to start trials with multiple users.

Lessons learnt

Lessons from the project included:

• User engagement in solar forecasting is most effective when output is higher, outside the winter period
• Investment is needed to take open-source community projects to the next level
• Deploying products early, even if with minimum features, is helpful as it allows useful user feedback as early as possible
• In such projects, using real data as early as possible can prevent the need for late changes to the user interface.

Next steps

The next steps are to continue to improve the accuracy of the forecasting system, to build an open source community around the product, and to move towards generating revenue.