Kansas City Power and Light Uses Eureqa® to Precisely Match Electric Power Supply to Demand
Founded in 1882, Kansas City Power and Light (KCP&L) is an investor-owned, regulated electric utility serving more than 800,000 customers throughout northwest Missouri and eastern Kansas.
It relies on a diverse portfolio of power generation technologies - coal, nuclear, natural gas, fuel oil, wind, and biogas - to provide reliable and affordable power while striving to minimize environmental impact.
The most important question KCP&L faces daily is that of how much power to produce. Producing more than what customers demand means wasted resources and unnecessary emissions. Under producing can mean having to buy power from other producers at high prices. Good predictions of total demand for electric power are therefore essential to efficient operation, and constantly increasing the accuracy of such predictions is a top priority at KCP&L.
Predicting usage is a complex task, but one thing is clear: the weather is the most important factor. In a nutshell, making predictions involves leveraging historical weather and electric-load data to generate models relating weather data to electric load. Forecasted weather conditions are then fed into those models to generate usage predictions.
Hundreds of Columns, Thousands of Rows, Several Frustrations
Gary Halbert, Principal Quantitative Analyst at KCP&L, leads the usage-prediction effort, and he and his department suffer from no shortage of relevant data. "We commonly work with data sets containing 15 to 25 variables including temperature, wind, barometric pressure, etc.," he says, "but the regional data set involves well over 300, and we have hourly measurements on each of those, so if we're looking at a whole year, that's 8,760 values for each of those 300 variables. Picking out the true drivers amid all that data can be difficult."
Up until recently, Excel was the tool of choice for KCP&L analysts. While it worked well for most linear regressions, it was slow when working with the larger data sets. Also, the analysts knew that a lot of the weather-to-usage relationships were nonlinear, and time-consuming transforms were often required before those relationships could be modeled within Excel.
These frustrations led Halbert to try out a popular data mining application. He and his team found that it offered various useful tools, but familiarity with R, a statistical computing language, was required before they could try it out on more complex modeling tasks.
"Scientists understand the basic forces of nature, but making precise predictions about what will happen in the real world, with its staggering complexity, is overwhelming to the human mind," Morgan Freeman intoned in an episode of "Through the Wormhole." He was introducing a segment showcasing Eureqa's ability to derive physical laws from complex data. The segment convinced Halbert that Eureqa was worth a try, so he and his team downloaded free trial versions. They saw Eureqa's value immediately
Faster Results, Better Fits, and Whole New Avenues
Since then, Eureqa has become more and more integral to KCP&L's analytic and predictive operations. For linear relationships, the analysts are getting better fits to the data than they ever got with Excel, and they're getting them significantly faster. Nonlinear relationships are handled effortlessly, and in about the time Excel would take to handle simple linear relationships.
Eureqa's ability to generate differential equations was an unexpected bonus. "It has provided the path to a solution for us in a couple of instances," says Halbert. “For example, we were able to generate models relating temperature to electric load that made use of the derivative of the parabola. Unless I did it by hand, that wasn't an avenue I previously had open to me, so that's a whole new area we can work with."
Halbert also likes Eureqa's output of not just one equation but a list of them, sortable by complexity, goodness of fit, or a combination of the two. The chosen equation will often wind up in a VBA macro, so he sometimes opts for one of the simpler equations, at the expense of an small reduction in accuracy, in order to keep those macros easy to understand and maintain.
In short, Eureqa is allowing KCP&L to more quickly and accurately do what they had been doing with Excel and to delve into other modes of analysis that hadn't previously been possible. The result has been a new level of insight into the effects of weather on power usage. "Eureqa has opened our eyes to the fact that some variables we thought were highly influential appear not to have much influence at all, and others have much higher influence than we had believed," says Halbert. "All in all, Eureqa is a great tool for quickly regressing data to find simple and unique solutions that you might not otherwise be able to find."Contact us to learn more »