Having a forecast of will behave as the load of an electrical circuit, can be strategic for companies, which may schedule your maintenance for the periods in which a circuit is less over-loaded, therewith not compromising the entire system. This graduate work presents a load forecasting model in the electric power system using artificial neural networks. The study aims to perform load forecasting of a week later taking as basis for learning the last three weeks. For both, it was used the load database in the period of a year, of the Tucuruí circuit - Vila do Conde 1, of Tucurui's substation. During the work, plus a quick detailing of the circuit characteristics in study, will be presented to computational intelligence technique used. There will be a detailed description of the steps in the training, validation and testing of the neural network. The software used for programming the neural network was MATLAB.
|
