Climhydro-2: Canadian hydropower companies' strategies for climate change adaptation

Climate change will alter the flow rates of Canadian rivers. It will therefore be necessary to adapt water resource management practices to take this new climatic and hydrological reality into account. For water system managers, this will mean adapting the operating rules of dams in order to optimize the different, and often conflicting, uses of water, such as hydroelectricity generation and flood protection. This adaptation must involve the development of mathematical approaches and computer tools that are applicable to water systems with several reservoirs and are capable of handling the variability and hydrological change caused by climate change. The main challenge for this research project concerns the creation of these innovative management tools.

To develop approaches and tools to facilitate hydropower generation planning and to adapt operational system management methods to climate variability and changes.

• Development and comparison of several optimization methods that are suitable for non-stationary stochastic processes and applicable to multi-reservoir water systems

• Development of approaches to the forecasting/projection of inflows covering natural variability and climate non-stationarity that are applicable in current and future climates

• Validation of the approaches developed by testing on real management problems

First, the research project enabled the development of high-performance tools to optimize the management of the equipment stocks of the two hydroelectric generation companies involved in the project (Rio Tinto and Hydro Québec). A first tool, based on stochastic dynamic programming (SDP), now makes it possible to solve a reservoir management problem for a system consisting of three or more reservoirs, which classic SDP does not allow, at least with current computing capabilities. Another approach, based on a pre-selected family of rules with optimized parameters rather than on the optimization of management rules, has generated promising results while avoiding the computational requirements imposed by SDP.

Furthermore, the work carried out has made it possible to explore promising methods of seasonal hydrological forecasting. It has been shown that ensemble weather forecasts, produced by forecasting agencies, improve seasonal hydrological conditions (1 to 5 months) under certain conditions compared to the use of climatology. However, a better understanding of the results is required, particularly in relation to the hydrometeorological characteristics of the watersheds concerned. In addition, an original method of seasonal hydrological forecasting based on the use of data mining algorithms and including a combination of climate indicators was developed and shows encouraging results in terms of the prediction of inflows in a context of non-stationarity in northern basins. For example, when tested on the Romaine River watershed, the approach made it possible to calculate seasonal flows up to two seasons in advance (Figure 1). The method must be further developed, in particular by generalizing the results, before its operational implementation.

Figure 1 : Simulated (red) and observed (blue) seasonal inflows of the Romaine River.

Innovative research on the regionalization of hydrological models was carried out and found that a reduction in the complexity of hydrological models by simplifying their structure and thus the number of parameters required to calibrate hydrological models did not result in an overall loss of performance, significantly reducing the equifinality problem. However, it was not possible to conclude that a parsimonious model was easier to regionalize than a more complex model. Despite the progress made in this research, the problem of regionalizing hydrological models, which is essential for hydrological modelling and forecasting on ungauged or poorly gauged basins, remains unresolved.

Lastly, the application of the algorithms developed in this research to real water systems is offering encouraging prospects for technology transfer in view of operational use. For example, one study focused on the management of the Nechako River water system with the help of a computer-based trial that updates management rules at each step of calculation time. The specific objective was to investigate the use of ensemble forecasts to improve the management rules of the Nechako River, established using Rio Tinto’s stochastic dynamic programming optimization tool. The results showed that hydrological modelling has potential for optimizing the real-time management of water systems in northern regions, as it represents different hydrological processes.

• École de technologie supérieure (ETS)

• Hydro-Québec

• Université du Québec à Chicoutimi (UQAC)

500003