Among the different types of natural disasters we know, there are those which derive from severe droughts or torrential rains, affecting a whole country. A case in point would be the extreme drought that affected Colombia during 1992-93, when reservoirs reached 14% of their capacity, leading to a profound reform process of the national electricity system, as the power sector was based on hydroelectric generation.
The flow prediction models are important in this regard, as they help companies generating hydroelectric power to take decisions (electricity generated using the energy of moving water), and to all those which depend on them, so that they are able to cover the supply and prevent natural disasters.
In countries with no severe weather conditions these models, though uncertain, do not present too much complexity and are based on the estimation of the incoming water flow (precipitation) and the outgoing water flow (irrigation, soil permeability, evaporation or supply).
However, hydrological processes in countries affected by extreme weather effects exhibit a high nonlinearity, and hence prediction of the relevant variables in a basin requires continued search and study of predictive models that work well.
Prediction models of this type of behavior are complex, among other reasons, because of the presence of explanatory variables which in turn are difficult to predict, such as the number of sunspots, the number of hurricanes, the average water temperature or the barometric pressure.
Once these preliminary predictions have been fine-tuned, we can create chained predictive systems, allowing power generating companies to make accurate medium term predictions, very useful for decision-making about opening of floodgates, purchase or generation of thermal and/or wind energy, or the maximum power to cover in the market.