Renewable energy generation like wind and solar have more inherent uncertainty than others, simply because the strength of the sun or wind can fluctuate. But even before renewables add uncertainty to the grid, human behaviour made electricity demand difficult to predict.
When human behaviour no longer follows a pattern
In 2020, the pandemic flipped a switch on human behaviour. The usual routines that created semi-predictable surges, peaks and slumps suddenly disappeared. People were no longer getting up at 5am for a commute, nightlife routines were redundant, many commercial spaces ceased to use power for weeks and months at a time. Shops were left unheated, offices unlit.
In the UK, energy consumption was reduced as a whole in 2020. This lower demand led to long stretches of time when the GB electricity transmission grid was running without any coal, and a zero carbon future seemed to inch a little closer.
How can energy demand be effectively forecasted and managed in this period of unprecedented change? Not by the standard means of relying on historical data. Nor by standard point forecasts that give a single figure for predicted energy consumption. The energy sector needs not just data, not only historical projections into the future, but a way of intelligently calculating the best course of action in light of great uncertainty.
When the uncertainty will only grow
So what is ahead? The energy sector has had a little time to adapt to these new patterns of human behaviour, and some are saying that once normality is restored we can ignore the last year as an anomaly. But, there’s lots of talk of the ‘new normal’, and we can’t assume that people will return to their old habits. For a start, behaviour patterns we form will be shaped by our new reliance on online workplaces and a lesser need for travel.
More than that, new waves of disruption are on the horizon: extreme weather conditions heightened by climate change, technological advances that demand more energy, and the shift to an energy sector run solely on renewables – which bring their own volatility.
Point forecasts vs. probabilistic forecasts
When we rely on point forecasts we are depending on very narrow predictions. To look at a single estimate is to close our eyes to the inherent uncertainty there is in any prediction of the future. Like Michael Fish predicting a light breeze, we will ignore the hurricane when it comes.
At the Smith Institute, we advocate an alternative. Bespoke statistical and machine learning models that can quantify the uncertainty in their predictions. This gives us a probabilistic forecast: a likely range of outcomes, quantifying the uncertainty that exists. And once we understand it, we can work with it.
At the risk of making light of a complex process, let’s consider an example of a probabilistic forecast that we all know: weather forecasting. A point forecast might tell you ‘It’s going to rain in four hours’ but what does that mean? How accurate is that prediction? You need a probabilistic forecast, like the percentage risk of rain weather apps give, so you know how likely you are to need that umbrella.
While the energy sector works with point forecasts, they end up wasting energy and resources. There might also be some cases where, unknowingly, we haven’t properly mitigated against risks such as a generator tripping or sharp changes in energy prices. This can be a serious issue for system security if load shedding needs to take place as a result and essential infrastructure is left unpowered.
A way to make better business decisions
We want to enable the energy sector to make more robust decisions. To not only calculate a daily accurate probabilistic forecast, but to anticipate the impact the uncertainty will have, and to assess what is the most secure and cost-effective course of action.
It’s not an easy thing to achieve, which is why our mathematical consultants are on the task. They draw on an advanced scientific skillset to exploit the full potential of data and enable energy companies to map out the future, mitigating risk and seizing opportunities. Get in touch to explore what we can achieve for your own organisation.
More in Energy
SI Insight: A renewable energy sector: how to weather uncertainty to make robust decisions
Case Study, EDF: How the Smith Institute is helping EDF safely deliver low-carbon energy to the UK
Case Study, National Grid ESO: Modernising optimisation algorithms for the control room