Impact of the quality of hydrological forecasts on the management and revenue of hydroelectric reservoirs – a conceptual approach

Abstract

Abstract. The improvement of a forecasting system and the continuous evaluation of its quality are recurrent steps in operational practice. However, the systematic evaluation of forecast value or usefulness for better decision-making is less frequent, even if it is also essential to guide strategic planning and investments. In the hydropower sector, several operational systems use medium-range hydrometeorological forecasts (up to 7–10 d ahead) and energy price predictions as input to models that optimize hydropower production. The operation of hydropower systems, including the management of water stored in reservoirs, is thus partially impacted by weather and hydrological conditions. Forecast value can be quantified by the economic gains obtained with the optimization of operations informed by the forecasts. In order to assess how much improving the quality of hydrometeorological forecasts will improve their economic value, it is essential to understand how the system and its optimization model are sensitive to sequences of input forecasts of different quality. This paper investigates the impact of 7 d streamflow forecasts of different quality on the management of hydroelectric reservoirs and the economic gains generated from a linear programming optimization model. The study is based on a conceptual approach. Flows from 10 catchments in France are synthetically generated over a 4-year period to obtain forecasts of different quality in terms of accuracy and reliability. These forecasts define the inflows to 10 hydroelectric reservoirs, which are conceptually parameterized. Relationships between forecast quality and economic value (hydropower revenue) show that forecasts with a recurrent positive bias (overestimation) and low accuracy generate the highest economic losses when compared to the reference management system where forecasts are equal to observed inflows. The smallest losses are observed for forecast systems with underdispersion reliability bias, while forecast systems with negative bias (underestimation) show intermediate losses. Overall, the losses (which amount to millions of Euros) represent approximately 1 % to 3 % of the revenue over the study period. Besides revenue, the quality of the forecasts also impacts spillage, stock evolution, production hours and production rates, with systematic over- and underestimations being able to generate some extreme reservoir management situations.

Choosing between post-processing precipitation forecasts or chaining several uncertainty quantification tools in hydrological forecasting systems

“…The question then shifts from using a postprocessor or not in each system to which sources of uncertainties should be prioritized and quantified. Furthermore, such evaluation would allow us to know whether implementing each system (with raw and corrected precipitation forecasts) would result (or not) in a different decision and how the decision would (or not) be influenced by the quality (bias, reliability, accuracy, and sharpness) of the forecast (see, for instance, Thiboult et al, 2017;Cassagnole et al, 2021). In the end, the "perfect" system is not only the one that can represent the dominant hydrological processes and variability, but also the one that allows us to make the right decision at the right time and situation.…”

“…Reliable and accurate hydrological forecasts are critical to several applications such as preparedness against floodrelated casualties and damages, water resources management, and hydropower operations (Alfieri et al, 2014;Bogner et al, 2018;Boucher et al, 2012;Cassagnole et al, 2021). Accordingly, different methods have been developed and implemented to represent the errors propagated throughout the hydrometeorological forecasting chain and improve operational forecasting systems (Zappa et al, 2010;Pagano et al, 2014;Emerton et al, 2016).…”

“…Nevertheless, remote sensing of the atmosphere and surface by satellites has revolutionized forecasting and has provided valuable information around the globe with increased accuracy and frequency to forecasting systems. Therefore, satellite observations have been used in many hydrological applications as alternative data (Choi et al, 2021;Zeng et al, 2020;Kwon et al, 2020).…”

Day-ahead energy production in small hydropower plants: uncertainty-aware forecasts through effective coupling of knowledge and data

“…Predictive uncertainty is defined as the probability of occurrence of a predictand's value (in the particular case, energy production) conditional upon prior observations and knowledge, as well as on all the information we have obtained on that specific value from model forecasts (Coccia and Todini, 2011). A typical means to quantify the predictive uncertainty of a deterministic simulation model, is to add a random component (noise), w t , to its output, y t , where the random process w t should be consistent with the statistical and stochastic regime of the associated residuals (Efstratiadis et al, 2015).…”

“…The first refers to the short-term energy production forecasting by solar and wind power systems, typically on the basis of Numerical Weather Prediction (NWP) models, providing deterministic point forecasts. The second field of interest deals with the long-term energy production by large hydropower reservoirs, based on projections of their inflows (e.g., Cassagnole et al, 2021). Nowadays, emphasis is given to datadriven approaches (e.g., machine learning), also combined with stochastic-probabilistic schemes for representing uncertainties that are ignored by NWP models (Felder et al, 2018;Talari et al, 2018;Croonenbroeck and Stadtmann, 2019).…”