Predicting customer demands is critical for accurate simulation of gas pipeline networks. A long-standing challenge for accurate demand forecasting has been to reveal the customer's behavior within monthly bills correlated with monthly average temperatures. Access to usage data on a daily or even hourly basis can provide significantly improved forecasts of daily demands and hourly usage patterns of customers, particularly at colder temperatures.
The installation of Advanced Metering Infrastructure (AMI) at PG&E has created the opportunity to evaluate the hourly usage patterns of gas customers. This presentation and paper highlight the challenges in gathering and data mining this voluminous usage data and the development of mathematical models to forecast 24-hour demand profiles of residential customers through the use of multiple linear regression analysis.
At PG&E the recent availability of hourly customer gas usage information has initiated a transformation of the load projection methods supporting hydraulic gas system modeling. This virtual metering of gas usage throughout PG&E's service territory has highlighted the limitations of the current model loading process and has opened up opportunities to build significantly better modeling tools to accurately predict customer gas demands for a variety of design conditions.
The initial challenge facing the gas planning department began over 10 years ago at the onset of the Advanced Metering Initiative (AMI) project at PG&E. The $1.7 billion dollar (USD) SmartMeter™ 1 project was cutting edge due to the ambitious goal to install hourly meters for 4.5 million gas customers and over 6 million electric customers by the end of 2012. Due to the massive size of the project, despite Gas System Planning's early involvement with the project team, our data needs soon took a back seat to the significantly more important role of the new SmartMeter™ system - to accurately gather meter reads to ensure accurate billing.
