Volumetric Inflow Estimation in a Hydropower Dam Using Autoregressive Integrated Moving Average (ARIMA) Modelling and Altimetric Lake Levels

Abstract

Background: Several models have been developed for inflow forecasting in reservoirs based on local parameters which may not include an implicit system characteristic like seasonality. Autoregressive integrated moving average (ARIMA) models can be developed to cater for the presence of seasonal and non-seasonal behavior of natural water systems.  Aims: The present study aims to estimate Volumetric Inflow in a Hydropower Dam using Autoregressive Integrated Moving Average (ARIMA) Modelling and Altimetric Lake Levels. Study Design: The study was conducted in the Kainji reservoir, West Africa located along the Niger River. This study combines satellite-altimetry-based rating curves with reservoir inflow models that capture the seasonality of upstream characteristics. Results and Discussion: Seasonal multiplicative ARIMA models were developed based on 27-year inflow records and used to forecast seven subsequent years. Reservoir levels measured by satellite radar altimeters were matched with actual inflows to generate rating curves from which future inflows may then be estimated. The model with the best forecasts relative to actual inflow - a seasonal multiplicative ARIMA (2,1,1) x (2,1,2)12 model – was adopted. Conclusion: Strong agreements between all three inflow series (actual, model-based, and satellite-based rating curve) suggest that reservoir inflow models can be combined with satellite altimetric for reservoir inflow estimation.