Use of predicted rainfall in forecasting inflows for reservoir operations

The purpose of this study is to determine if an improved and more accurate schedule of a reservoir system is obtained by incorporating an increase in hydrologic inflows from short-term (10 days) predicted rainfall on tributary reservoir watersheds. The TVA reservoir system is used to illustrate the analysis. The TVA reservoir system is operated and scheduled on a daily basis with the intent on maximizing the benefits for which the reservoir system was designed. The ability to forecast the inflows to the reservoir system for a 10 to 14 day period is a very challenging process. Although incorporating the increase in hydrologic locals from predicted rainfall may not be entirely accurate due to the reliability of weather forecasts, it is the purpose of this study to prove there are significant benefits, particularly with regards to predicting hydroelectric generation, in using the increase in hydrologic inflow^; when scheduling the TVA reservoir system.

Hydroelectric generation totals, for 6 to 10 days into the future. were compared from reservoir schedules based on predicted rainfall and those with hydrologic inflows based on observed rainfall. The generation totals from the observed rain schedules were underestimated by an average of 30 to 50 percent. The schedules using predicted rainfall were only underestimated by an average of 10 percent. Hypothesis testing, by using the rank-sum test, indicate the data are significantly different, with confidence levels of at least 95 percent. The central tendency of the means of the predicted rainfall schedules were also closer to actual observed reservoir conditions. All the data clearly indicate a significant improvement in the reservoir system schedule by incorporating predicted rainfall.

As a direct result of this study initiative the TVA reservoir system recently began scheduling the tributary reservoirs using short-term predicted rainfall.