Time Series Analysis for Predicting Hydroelectric Power Production: The Ecuador Case

Abstract

Electrical generation in Ecuador mainly comes from hydroelectric and thermo-fossil sources, with the former amounting to almost half of the national production. Even though hydroelectric power sources are highly stable, there is a threat of droughts and floods affecting Ecuadorian water reservoirs and producing electrical faults, as highlighted by the 2009 Ecuador electricity crisis. Therefore, predicting the behavior of the hydroelectric system is crucial to develop appropriate planning strategies and a good starting point for energy policy decisions. In this paper, we developed a time series predictive model of hydroelectric power production in Ecuador. To this aim, we used production and precipitation data from 2000 to 2015 and compared the Box-Jenkins (ARIMA) and the Box-Tiao (ARIMAX) regression methods. The results showed that the best model is the ARIMAX (1,1,1) (1,0,0)12, which considers an exogenous variable precipitation in the Napo River basin and can accurately predict monthly production values up to a year in advance. This model can provide valuable insights to Ecuadorian energy managers and policymakers.

A Systematic Review of Statistical and Machine Learning Methods for Electrical Power Forecasting with Reported MAPE Score

“…ARIMA models can be applied to non-stationary data, and when the data are seasonal, the SARIMA model can be implemented. The ARIMA and SARIMA models have been used in many studies for forecasting [ 14 , 15 , 16 , 99 , 100 , 103 , 127 , 136 ], reaching forecast accuracies with an average MAPE value of 3.214%. A typical ARIMA model can be expressed by Equation ( 3 ), where the variable is replaced by a new variable obtained by differencing d times [ 25 ]: …”

“…Additionally, other authors [ 10 ] consider hybrid approaches that focus on a series of individual methods, such as noise reduction, seasonal adjustment and clustering, to process the data in advance, whereas combined methods use weight coefficients. With respect to the techniques implemented to forecast energy in recent years, in the international context, we can find a wide diversity; e.g., the application of kernel-based multitask learning methodologies [ 11 ], energy load forecasting methodologies based on deep neural networks as in [ 12 , 13 ], methodologies based on the classic time series approach as in [ 14 , 15 , 16 ], and mathematical representations as in [ 17 , 18 , 19 ]. Developing a model that achieves the highest forecasting precision in the context of electric power has been the object of study in recent years.…”

Hydropower production prediction using artificial neural networks: an Ecuadorian application case

“…Since hydropower strongly depends on meteorological conditions, rainfall data play a crucial role in our study. Previous works identified some limitations of statistical time analysis techniques [7,8]: the difficulties in incorporating multiple exogenous variables and the lack of accuracy for long-term predictions. Following the success of neural networks in many areas [9,10], the departing hypothesis of this paper is that ANN models can better capture the dynamics of hydroelectric production and, therefore, provide better results in our use case-even though the number of observations is not very large.…”

Estandarización de procesos prioritarios en la recuperación del rodete de una turbina tipo Francis

“…Ésta transforma la fuerza que genera una corriente de agua en energía eléctrica. Para poder aprovechar la fuerza del agua se edifican estructuras hidráulicas de gran tamaño que extraen el mayor potencial de esta energía renovable [2]. El proceso consiste en hacer circular una cantidad considerable de agua por un círculo hidráulico que tiene un desnivel [3].…”