Why Models Don’t Matter During an Eclipse

Grid operators and energy traders refined their predictive models for months in preparation for the total solar eclipse on August 21. Sophisticated algorithms calculated how it would affect energy output from solar panels across the US at every point in time. Traders adjusted futures prices, and utility operators altered plans for power generation and transmission. But they both failed to account for another critical element of the event: how Americans energy-use behavior might also change as the eclipse unfolded outside their windows.

As more and more people stepped outside their homes and offices to view the eclipse — shutting off televisions and air-conditioners in the process – energy demand dipped well below expectations. As reported in Bloomberg, traders were caught flat-footed as spot prices for electricity took a nose-dive in many parts of the country. Thankfully, power utilities were able to balance supply with the unexpected dip in demand and avoid any disruptions in service.

While far from a catastrophe, the surprising drop in demand last Monday serves as a powerful reminder of the limits of predictive models for managing our energy grid. Electricity demand is ultimately a reflection of human choices – when to do the laundry, or whether to cook dinner on the stove versus the microwave – which are challenging for a power utility to predict. Add to that a growing number of rooftop solar panels — equally unpredictable in their own right, as the weather changes – and grid operators are left with an increasingly complex and volatile supply-demand equation which must be balanced at all times.

The unpredictability of local solar generation on a normal day dwarfs that of utility-scale solar plants during an eclipse. While the macro-effect of an eclipse on high-voltage transmission can be modeled precisely, as was done on the 21st, the local effect of scattered clouds over scattered rooftops on the distribution system is impossible to predict with accuracy. Per the DOE’s ‘grid study’ published last week, we can expect the number of these distributed, variable energy resources in the US to continue its exponential growth.

The solution to unpredictability in both supply and demand of energy at the edge of the distribution grid is intuitive: stop trying to predict it. Advances in sensing, communications, and analytics – which constitute what’s often called the ‘internet of things’ – now present energy distribution utilities with an alternative to the predictive models widely used today. For the first time, grid operators can see and respond to changing power conditions in real time. At Utilidata, we help utilities leverage this real-time behavioral insight to make the grid more efficient, stable, and secure.

The next total solar eclipse viewable in the US will be in 2024. Instead of updating our models to better predict its effect on the electric grid, we’ll be better off gaining the ability to manage any surprises through continuous operational visibility. For watching and responding to the grid’s real-time behavior, no special glasses are required.