A Steam Load Forecasting Technique for Central Heating Plants

• Main Authors: Lin, Mike C, Carnahan, James V
• Format: Report
• Language: English
• Subjects: Air Condition, Heating, Lighting & Ventilating; ARIMA MODELS; ARMY; ATMOSPHERIC TEMPERATURE; BOILERS; BOX JENKINS METHOD; BUILDINGS; CENTRAL HEATING PLANTS; EFFICIENCY; ERRORS; FORECASTING; HEAT TRANSFER; HEATING; HEATING PLANTS; IDENTIFICATION; PE4A161102; PREDICTIONS; SAVINGS; STEAM; TEMPERATURE; TIME SERIES ANALYSIS; TRANSFER FUNCTIONS; VARIATIONS; WEATHER; WUEA-EC0

Because boilers generally are most efficient at full loads, the Army could achieve significant savings by running fewer boilers at high loads rather than more boilers at low loads. A reliable load prediction technique could help ensure that only those boilers required to meet demand are on line. This report presents the results of an investigation into the feasibility of forecasting heat plant steam loads from historical patterns and weather information. Using steam flow data collected at Fort Benjamin Harrison, IN, a Box-Jenkins transfer function model with an acceptably small prediction error was initially identified. Initial investigation of forecast model development appeared successful. Dynamic regression methods using actual ambient temperatures yielded the best results. Box-Jenkins univariate models' results appeared slightly less accurate. Since temperature information was not needed for model building and forecasting, however, it is recommended that Box-Jenkins models be considered prime candidates for load forecasting due to their simpler mathematics.