Improving Solar Forecasting by 33% Using Artificial Intelligence

The National Grid ESO has developed a program which incorporates artificial intelligence in order to improve its solar forecasting technology by 33%. The new forecasting technology was developed in collaboration with the Alan Turing Institute. The forecasting program was built using aspects of machine learning and data science. The project was funded by Ofgem's Network Innovation Allowance.

Initially ESO’s solar forecast used to consider two inputs/variables – solar capacity and solar irradiance. Using this data the old program then produced a forecast for solar generation output by establishing a simple relationship from the data inputted by the two variables.

The new, redesigned program, known as the ‘random forest’ model, looks at previous data and takes into account approximately 80 input variables some of which include temperature and more granular solar irradiation data. Using this historical data, the AI based algorithm then trains itself by attempting to finding others of hundreds of different mathematical pathways, known as decision trees. These decision trees allow inputs to be taken in and then provide an output figure. An average of these 80 variables are then taken in using this tree decision algorithm which then produces a new solar generation forecast. As you can see, this new approach is significantly more complex, advanced and much more reliable than ESO’s previous approach to solar forecasting as it mixes historical data and a variety of different variables to produce a much more accurate forecast.

ESO’s new forecasting program now uses 80 input variables as opposed to 2, initially.

ESO’s new forecasting program now uses 80 input variables as opposed to 2, initially.

This solar forecasting technology is all thanks to a combination of a variety of machine learning algorithms. With new variables, machine learning techniques and technologies being implemented, ESO has managed to improve its solar forecasting accuracy by 33%.

With climate change becoming a greater threat now than ever, accurate forecasting is becoming increasingly important as we must be able to maximise our efficiency with renewable resources. Eventually, we must reach a stage at which we are generating more energy from renewable energy than combusting fossil fuels. In fact, I’m the second quarter of 2019, solar generation increased by 18% compared to 2018.

Zacharia Sharif