A Comprehensive Deep Learning Method for Short-Term Load Forecasting

Load forecasting is an essential issue in future smart grids where inaccurate forecasting causes energy waste, power shortages, or cross-blackouts. Therefore, increasing forecasting accuracy is crucial due to the expansion of the type of loads and the amount of consumption and parameters that affect the load changes. Machine learning is a powerful tool for achieving artificial intelligence, and it is used for load forecasting as one of its applications. In this paper, short-term load forecasting is performed using an ensemble supervised learning based on random forest method named Deep Forest Regression. This method is also derived from deep learning and deep neural network theory. This forecast has been done using the data of residential consumption of an Iranian city for five months, including from half of May to half of September. The data is gathered every 30 minutes and stored in the system. By comparing the proposed method with some common methods, it can be seen that the proposed method has higher accuracy than those.