Electric cooperative sharpens forecasts, reduces energy costs

NCEMC decreases forecast generation time from six weeks to just a few hours

With population growth of 16 percent in the past decade, North Carolina has become a magnet for families and businesses drawn to its attractive climate, open spaces and business-friendly environment. That growth is increasing demand on the state's electricity infrastructure – demands being felt by the North Carolina Electric Membership Corporation (NCEMC).

NCEMC is comprised of 26 nonprofit, member-owned cooperatives that provide electricity to more than 950,000 households. To achieve the lowest possible costs and ensure a reliable energy supply, NCEMC creates sophisticated forecasts for short- and long-term horizons for its member cooperatives and regulators.

These forecasts are critical because energy in the electrical grid largely cannot be stored. Instead, it must be generated as needed and therefore must closely match the current demand. Imbalances can necessitate keeping expensive capital equipment such as transformers, transmission wires, substations, and even entire generation stations on standby to meet peak consumption periods.

Because nearly half of NCEMC's power is purchased, the organization needed accurate forecasts in order to negotiate the most favorable contractual terms and conditions.

Building reliable forecasts

According to Mike Burnette, Vice President for Wholesale Rates, NCEMC gains that visibility by creating hourly and daily forecasts of the power each of the cooperatives would require. Each forecast encompasses the number of consumers to be served, the energy they will consume and the amount of hourly energy each cooperative will require.

"We relate the energy-usage history to underlying economic conditions on a county-by-county basis," he said. "Then we examine historical relationships between weather and consumption so we can project what we'll need in the future.

"In the past," Burnette continued, "We'd hand off these forecasts to our generation and transmission team, who would convert our forecasts into their own models that simulate what portfolio of resources they need, such as coal and nuclear generation and wholesale purchases to meet the hourly demand curve.”

“This would affect decisions about contracts and plant construction for decades. Unfortunately, there was a forecast gap that forced them to convert our monthly data into 8,760 hours of forecast data for 30 years."

Tom Laing, Director of Market Research, said this gap required significant manual work to reshape monthly data into hourly load curves that the generation and transmission team needed.

"It took as many as six weeks to complete, so we did this about every two years," he said. "This manual process also limited our ability to consider multiple scenarios, which is critical in today's changing environment. It was clumsy and complex … we clearly needed a way to automate the process.”

As a longtime SAS^® customer, NCEMC knew SAS could provide a robust framework for advanced modeling. NCEMC explored additional tools to address the demand-modeling process, implementing SAS/OR – a powerful array of optimization, simulation and project scheduling techniques – to reconcile the hourly load curves with monthly energy forecasts.

SAS Forecast Server and SAS/ETS provide additional econometric, time series and forecasting tools for tighter modeling, forecasting and simulation.

"We worked closely with SAS to explore how these products could positively impact our forecasting process," Burnette said. "Before we were hammering a monthly demand forecast into an hourly shape. Now, we have integrated a load-shape process into the demand forecast.

"No longer do we need a team of individuals creating labor-intensive, time-consuming manual processes," Burnette said. "What once took six weeks can now be done in a few hours."

A significant return for the long term

"We love the ability to forecast hierarchically," said Laing. "We create as many as 100 small area forecasts and combine them for regional and statewide outlooks. To optimize model performance, we specify both hourly and daily demand models, which are combined to provide a robust estimate of hourly and daily peaks. The resulting load shape is then grown from one year to 40 years at all levels of the hierarchy."

The SAS solution has enabled the team to move human resources to other tasks, saving direct expenses. More importantly, armed with a clearer capacity outlook and load demands, NCEMC can secure the lowest-cost power to meet the demand of each member co-op.

"We're making better and faster decisions for the members over the long term," Laing said. "That's translating into a reliable, affordable portfolio of power for North Carolina at lower risk, which is, of course, the heart of our mission.