Very short-term bus reactive load forecasting models based on KDD approach

The very short-term bus reactive load forecasting allows the electrical system operator to determine the optimal amount of energy to supply the demand with quality, safety and reliability. With this premise, this paper used a knowledge data discovery approach to handle the forecast, from raw data to results analysis, using the neural network for data mining. Two forecasting models were developed: the individual model using only your historical data; and the clustered model using data from other similar buses. The models were applied to Brazilian power system bus data. The results were analyzed according to the error and the confidence level of the forecast.