Our primary focus at the moment is forecasting of solar photovoltaic (PV) power production, a difficult task.  As clouds move over PV panels, the power output moves up and down rapidly. To keep the energy grid in balance, operators need to have readily available generation reserves which usually come from fossil fuel sources.

If we have more accurate predictions of how much electricity a PV installation will produce over the next few hours then we can reduce the amount of fossil fuel reserve required. By making solar energy more predictable, we will make it easier for the grid to absorb more PV generation and for investors to reduce the risk of solar investments. Here is a detailed write-up of what we are doing in Wired magazine: Bad weather forecasts are a climate crisis disaster. Supported by the European Space Agency, National Grid Electricity System Operator, Google.org and Innovate UK, we are improving how Machine Learning can use satellite images to improve forecasts of PV power generation.

If you would like to trial our forecasting service, Quartz Solar, or know more about it, please get in touch.

When it comes to the energy market, data is needed for everything: to balance the grid, to decide where to install the next wind farm, to train machine learning models and many more. However, data, especially in the energy sector, is hard to access. In addition to being scarce, once found, energy data is often difficult to use.

Together with the Open Data Institute, Icebreaker One, Passiv Systems, along with others, and based on the recommendations of the Energy Data Taskforce in 2019, we worked to create the first version of Open Energy to enable sharing of energy data to improve the efficiency of the grid.

To accurately forecast solar power generation, we need to know where all PV panels are located. In the UK today, this information is still largely unknown. We are supporting the OpenStreetMap community to map the location of the world's PV panels. OpenStreetMap is the Wikipedia of maps: anyone can edit the database.

We use a combination of machine learning and crowd-sourced wisdom to locate PV panels and add them to OpenStreetMap. We are also working on ways to combine existing asset registers and then remove duplicate information.