Forecasting Winter Daily Natural Gas Demand at Vermont Gas Systems
Flock, Mike, The Journal of Business Forecasting Methods & Systems
On a cold Vermont winter day, the customers of Vermont Gas Systems need a dependable source of fuel to heat their homes and businesses.
Vermont Gas Systems is a natural gas utility with approximately 26,000 residential, business and industrial customers in 13 towns and cities in northwestern Vermont. Vermont Gas Systems' Gas Control Department forecasts the gas demand and arranges the gas supply and transportation from suppliers in western Canada and storage facilities along the Trans-Canada Pipeline which delivers the gas to our pipeline. The quantities of gas must be specified to the suppliers at least 24 hours in advance. The Gas Control Department must request enough natural gas to meet the needs of the customers but must not over-request gas which will needlessly and expensively tax Trans-Canada Pipelines' facilities. Because Vermont Gas Systems has the storage capacity for only one hour use of gas as a buffer between supply and demand, an accurate forecast of daily natural gas demand is critical.
FACTORS AFFECTING DEMAND
Vermont Gas Systems daily gas demand forecast models are developed using least squares regression analysis with weather information as independent variables and natural gas demand history as the dependent variable. Other factors considered in forecasting winter natural gas demand are industrial customer demand, and the changing end-use demand of gas customers.
When the weather is cold, customers use more gas to heat their homes and businesses. During the winter months, space heating accounts for a major use of natural gas, and an accurate weather forecast is the most important factor for forecasting winter natural gas demand. Detailed weather forecasts are delivered five times daily from a weather forecasting service.
The weather service provides temperature, wind and sun heat-value forecasts for the current and next three days allowing the Gas Control Department to plan three days in advance.
Natural gas demand is forecasted individually for 24 large-use industrial and municipal customers (factories, hospitals, schools, etc.,) who use the gas for space heating and industrial processes. A detailed daily demand history is kept for each of these customers and individual forecast regression models are developed for each.
A factor that changes little from day to day but must be considered in the long run is the changing end-use demand of gas customers. …