Physical and Economic Determinants on Forecast Horizon for Long-Term Reservoir Operation

The forecast horizon (FH) reflects the effective length of future inflow inputs for real-time reservoir operation. Its determination currently still relies on numerical experiments. Using a conceptual model of multistage water supply operation, this paper presents an analytical basis for determining FH, in which physical and economic determinants are explicitly incorporated. The criteria for FH determination are first derived employing the comparative relationship between the marginal benefits among stages. The effects of discount rate, inflow forecast variance, and reservoir storage capacity on FH are then analyzed both theoretically and numerically. The results show that FH responds nonlinearly and sometimes nonmonotonically to these determinants. A higher discount rate tends to consume more water in advance and shortens FH in most cases; however, a prolonged FH can also be observed when an extreme drought is expected and the resulting potential economic loss outweighs the immediate release benefit brought by high discount rate. In contrast, a larger forecast variance results in more water stored for future use and a generally prolonged FH; but a shortened FH also exists. A larger storage capacity enables higher flexibility in flow regulation, and therefore consistently prolongs FH if it changes. The presented analytical framework provides a promising access to analyzing FH in flood control, hydropower generation, and other analogy operations.