Like other commodities, energy prices are tied to supply and demand, making it critical to forecast energy requirements to better quantify future needs for purchased power. Greater insight into customer usage delivers more accurate demand and load forecasting, which supports planning, pricing, financial reporting and business unit performance reporting.
Advanced metering infrastructures are providing increasing amounts of valuable data that progressive utility providers are using to understand customer usage patterns, detect outages and theft, and perform load research. But larger volumes of data require improved analytic approaches.
Advanced analytics can give companies the competitive advantage to optimize demand forecasts and maximize profitability. In a recent webcast from SAS, Mary Russell, senior manager of performance analytics for Texas utility TXU, explains how her company is using advanced analytic tools to predict energy requirements.
With the typical electricity meters used to monitor power consumption for most residential and small to medium business customers, readings as usually only taken monthly, Russell says. So to accurately predict future energy requirements, utilities turn to profiling.
Portfolio management is very dynamic, Russell says. Because customers have changing needs, product offering must be flexible. But different product offerings also have different risks, so data aggregation also must be dynamic to provide accurate information. And that requires usage load analytics to be efficient and repeatable, she says.
Load profiles, an estimate of average energy patterns for a group of customers based on load research sample data, are used to shape a cumulative meter reading to 15 minutes intervals, Russell says. The profiles are regressions made up of calendar, weather and daylight variables.
Profile decision trees, which take into account profile codes, weather sensitivity, weather zone assignments, meter type and whether customers are classified as “time of use”, are used to assign profiles to individual customers, she explains.
Because meters are read in monthly cycles, Russell says the first step in shaping meter data is to calculate the usage factor, which is the meter reading divided by the sum of the assigned profile during the same period of time. The usage factor can then be used to determine consumption based on normal weather patterns.
Calendarized consumption can then be determined based on when the meter is read each month and data is aggregated by customer groupings. Next, Russell says, companies must forecast customer counts for each category.
Retail customer count forecast models are driven by several variables, including; switch ins and outs, drops, net losses for disconnects, organic growth, competitive market activity, internal marketing initiatives, price movements and historical customer counts.
With customer counts determined, utilities must forecast average monthly consumption.
“We know historically what that category has used, but we know the portfolio is changing, so we must incorporate those changes into forecasted average monthly consumption,” Russell explains.
Forecasted monthly consumption is then multiplied by forecasted customer counts to allow predictions of retail volumes by category, based on historical weather normal consumption and other price differentiating variables.
Russell says retail volume is forecasted hourly because prices vary hourly and follow load. “Hourly load and prices shapes provide insights into the cost of serving different load profiles,” she says.
Advanced analytics provides better planning, Russell says. TXU’s retail forecast, she explains, is a key element in the operational and financial forecasts presented externally.
It also facilitates risk management assessment. “Determining hourly retail volume position and supply positions allows us to schedule load and anticipate impact of market price movements on supply and demand,” she says.
New technology, like automated meter reading, she predicts will make more granular data available and allow utility providers to:
• Conduct more in-depth statistical analysis on individual customers and customer groups
• Complete more frequent updates on profiles or even more profile groups
• Utilize actual hourly historical data as opposed to “shaped” data
• Achieve more accurate profiles to reduce unaccounted for energy
• Distribute costs to users more precisely
To succeed, energy retailers must acquire customers in the most cost effective way, increase sales per customer, reduce the cost of service, increase customer retention, avoid commoditization, and effectively manage supply contract risk. And advanced analytics and new technologies, like AMR, can help utility providers achieve these goals.
More data will facilitate more refined forecasts. Advanced analytics, SAS senior business solutions specialist Craig Carothers says, provides a mechanism to not only forewarn the organization and market of change, but to be able to leverage it to reduce costs and quality of service.
The bottom line: better forecasting will increase profitability and allow utility providers to deliver better service to their customers.
