Home Technology Use of Forecasting Techniques to Estimate Energy Generation in Catalan Reservoirs
Use of Forecasting Techniques to Estimate Energy Generation in Catalan Reservoirs

Use of Forecasting Techniques to Estimate Energy Generation in Catalan Reservoirs

Researchers from The Open University of Catalonia assessed state-of-the-art methods to estimate the water level in Catalan reservoirs

Reservoirs are enlarged natural or artificial lakes used as a source of water supply and electricity through hydropower plants. Reservoir water level plays a major role in energy production. Although reservoirs may contain a large quantity of water, meteorological variations may have a significant impact on water availability throughout the time. This in turn increases the need for real-time water-level measurement and estimation. Now, a team of researchers from Open University of Catalonia used state-of-the-art forecasting techniques to predict reservoir water levels in Catalan reservoirs. The team focused on predicting reservoir water levels in high accuracy.

The team collected data from public online sites regarding volume of water and flow of entry of the reservoir and meteorological data such as temperature, rainfall, and wind speed. The team compared the most used methods: neural networks, SVM, and random forest along with neuronal network with radial base function (RBF) and the classifiers Kstar, k-nearest-neighbor (KNN), and random tree. Support-vector machine or SVM is a supervised learning model with associated learning algorithms that analyze data used for classification and regression analysis. Random Forest is a flexible, easy to use machine learning algorithm that produces, even without hyper-parameter tuning, a great result most of the time.

SVM offer the most optimum results in both in the multi-variant and in the uni-variant analysis. The results of the neural networks were closely identical to SMV. The team recorded longer prediction with higher precision compared to state-of-the-art techniques. Moreover, uni-variant random forest was superior to other works. In further research, the team plans to model a reservoir, which can generalize (in each case), include various natural events, and assess sophisticated techniques. The research was published in the journal MDPI Energies on May 14, 2019.


Abhijit Ranjane
Abhijit Ranjane,

Abhijit Ranjane
All Posts

YOU MAY ALSO LIKE