Flow Prediction Models for a Mini Hydroelectric Power Station

In this paper, a comparison between two systems implemented in Jupyter Notebook with the Python computation kernel is presented. They use machine learning to predict the flow rate. The models were trained and evaluated using data from the year 2022. The first system uses the ARIMA method and shows an error of 4.55% in flow prediction. A high similarity is observed between the actual and predicted flow curves for each month. The second system uses the RNA-GRU method and shows a MAPE error of 4.62% in the flow prediction. As with ARIMA, there is a significant resemblance between the actual and predicted flow curves for each month. These graphs allow us to evaluate the average errors of the forecasts. In this case, it is evident that the ARIMA method has a lower error compared to the GRU method. This paper demonstrates the ability of both systems implemented in Jupyter Notebook to predict the flow rate.