From understanding where solar generation capacity like solar panels is located, opening up energy data to predicting how much energy will be produced – our projects cover a variety of topics.
Forecasting solar photovoltaic (PV) power production is hard: 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 power 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. Supported by the European Space Agency and many other partners we are investigating how to use Machine Learning and satellite images to improve forecasts of PV power generation.
2. Open Energy Data
Data is needed for everything: To balance the grid, to decide where to install the next wind farm, to train machine learning models and more. However, data is hard to access, especially in the energy sector. It’s hard to find data and it’s 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 are working on enabling sharing of energy data - focussed initially on PV data - to improve the efficiency of the grid.
3. Solar Panel Mapping
To accurately forecast solar power generation, we need to know where all the PV panels are. That is still largely unknown today in the UK. 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 the wisdom of the crowd to locate PV panels and add them to OpenStreetMap. We are also working on ways to combine and then de-duplicate existing asset registers.