AWWA ACE65282

This powerpoint presentation focuses on short-term consumption forecasting (STCF) techniques for energy management (AwwaRF Project #3066 Water Consumption Forecasting to Improve Energy Efficiency of Pumping Operations). The first section of the presentation outlines short-term consumption forecasting - a key strategy to minimize increasing energy costs and includes these topics: operating costs for 19 sampled water utilities in the 2003 AwwaRF Project; Electric Costs Rising Dramatically Due to Increasing Fuel Costs; Other Factors Pushing Up Electric Energy Cost; Water Utility Responses to Rising Costs; Short Term Consumption Forecasting (STCF) Initiates Optimized Operational Pump Scheduling Through an Energy & Water Quality Management System (EWQMS); Forecasting Consumption Moves Water Utility from Reactive to Proactive Operations Using an EWQMS; Pumping Operations are Typically Consumption-Following; STCF's Objective: Predict Hourly Consumption for Multiple Areas in the System and Move Water from Source to Customer at Lowest Cost; STCF Enables Time Value Purchasing and Selling Energy - A Supply and Demand Strategy; STCF Window (Daily and Hourly) Depends on Application; Time-of-Use Pumping Strategy Based on a Forecast; Example of Time-of-Use, Demand, and Efficiency Strategies; STCF Enables Significant Water Supply Opportunities; and, Scheduling Treatment Plant and Import Water Production Schedules Based on Short Term Forecasting has Saved San Diego $Millions. The second part of the presentation provides a description and the results of the AwwaRF Project #3066, "Water Consumption Forecasting to Improve Energy Efficiency of Pumping Operations" and provides an overview of the project and participating facilities. The four steps of the forecasting project include: Research Existing Forecasting Techniques Used at Water, Gas, and Electric Utilities; Analyze Operational Results at Four Water Utilities Using STCFs; Develop and Test Prototype STCFs Under Operating Conditions at Five Water Utilities; and, Analyze Data and Publish Results. The third part of the presentation provides conclusions and lessons learned: accuracy is highly dependent on quality of historical data; other than temperature and rainfall, weather may not be as important as one would think; retraining an ANN may degrade the model; ANN models hold up well over time; Training ANN models with a full year of data provides better results than training with only a single season of data; daylight savings is problematic for all forecast models; large number of input parameters can make models less responsive and more difficult to maintain; Forecast is limited by accuracy of measurement equipment; and, a short-term consumption forecaster "enables" proactive system operations to save substantial energy and other operating costs.