Forecasting an Energy Choice gas supplier delivery requirement follows a basic business tenet—the amount of gas delivered by a supplier each day should match the amount of gas used by its customers. When the two pieces of this transaction match, supplier costs for energy correlate closely with the revenue received from its customers. However, when delivery volumes do not match usage billed on a consistent basis—requiring future transactions to true-up the mismatches—then a level of risk exists in both the commodity side and cost side of the supplier operations.
At Baltimore Gas and Electric, developing plans for leveraging gas smart meter technology was a way to revisit prior forecasting, variance identification and true-up processes, which had suffered from several inherent problems. First, the forecasting process was dependent upon a small number of sample meters, representing general gas usage for each of five customer classes. The population of about 700 sample meters did not sufficiently capture specific customer demographics unique to each of the 38 participating gas suppliers.
Examples of differing residential customer demographics among suppliers that would result in vastly different actual gas usage include: Supplier A serving residential gas customers who mostly live in condos in the city; Supplier B serving residential gas customers residing in larger homes in the extreme northern part of the service territory; and Supplier C serving townhome communities located in the western part of the service territory. While the various residential customer types all belong to the same customer class and would share an identical regression analysis graph, the actual usage volumes of these separate residential customer types vary significantly from each other and the forecasted requirements.
Similar demographic differences arise in the commercial classes, where individual suppliers might primarily be serving customers in one of the following categories—churches, dry cleaners, restaurants or supermarkets—and the regression analysis for the same commercial class may be populated by sample meters mostly from commercial offices. Overarching both residential and commercial energy demographics is customer location, where on a given day or extended time period, winter temperatures can vary widely among different regions within the BGE service territory, affecting gas use by customers for heating purposes.
Other problems in identifying supplier delivery/customer usage variances included the amount of time needed to capture variances and the inaccurate proration of variances into calendar months when usage is obtained from non-calendar month meter reads. The final unwieldy piece of this puzzle is that variance “true-up” periods would sometimes not commence until several months after the variance occurred. These delays could result in potentially large winter variances “truing-up” out of season during the summer months.
As a solution, BGE Gas Choice developed and proposed a process that would collect daily data from each customer’s smart meter data. As part of the process, the usage data—available in hourly intervals—would be grouped into the industry standard gas day (10 a.m. ET to 10 a.m. ET). This grouping alone was something that had not been considered previously in the design of gas smart meter applications data capture and required systems programming changes.
All of the customer demographics issues were now resolved using this new forecasting process. The Supplier A regression analysis and associated forecast only included usage data from its actual customers residing in downtown condos, while the Supplier B forecast was driven from its customer usage data in larger homes in the extreme north of the territory, and so on. For commercial customers, the Supplier X regression analysis and associated forecast only included usage data from actual customers that were primarily restaurants, while the Supplier Y forecast was driven by its customer data in the office buildings that had enrolled, and so on. In the new grouping of gas smart meter data, we would be running about 100 separate regression analysis forecasts to capture unique customer demographics versus the five that were utilized during the legacy meter forecasting process.
Leveraging gas smart meter technology to support gas supplier delivery forecasts also enabled the delivery/usage variance capture process to evolve. Variance identification that had previously taken up to two months could now be reduced to just two days. A daily routine comparing customer usage against the same-day supplier delivery produced a daily variance. The gas usage/delivery variances captured two days after each supplier delivery were subsequently trued up within 24 hours by rolling the variance (positive or negative) into the next day’s delivery requirement. With the old legacy meters, the final day of true-up could take up to 17 months after the first variance occurred in the delivery cycle.
The consensus among both internal and external stakeholders is that BGE has taken a significant step forward in its efforts to match supplier delivery and customer usage with the implementation of this project and along the way has taken steps toward setting a new industry standard specifically for Gas Choice supplier delivery forecasting.
Michael K. McShane is the manager for energy supplier services at BGE.
