This research presents a prediction system using the open-source software Python. We performed, database processing and created LSTM (long short-term memory) recurrent neural networks. We processed data on the energy production of a small power plant in the province of Cotopaxi and the height of the water in the reservoir. Between the years 2017 and 2020, employees collected this data. We divided the data into two groups: 80% for training and 20% for validation. We conducted various scenarios under varied conditions to develop an effective recurrent neural network (RNN) model. We selected 15 input neurons, used 2 hidden layers with a dropout layer to avoid memorization, and included an output layer. We used the RMSprop optimizer with 15 delays. As a result, an MSE of 4.67% was achieved, which validated the use of the algorithm to predict the flow rate. We recommend conducting a similar study using other open-source data processing programs. In order to produce electrical energy, it is imperative to emphasize that flow prediction is an essential component of Ecuador’s National Development Plan “Toda una Vida.” In order to achieve the production goals necessary to promote the country’s sustainable development of the country, maximizing the utilization of water resources is the main objective of this research.