No abstract

No abstract

No abstract

Scientific load forecasting methods and accurate forecasting results are an important basis for the power system planning department work, as well as the basis and guarantee for correct decision-making on investment and construction. This paper has thoroughly studied the basic principles of recurrent neural networks and their shortcomings that they cannot solve long-term dependence problems, and a power system load forecasting method based on LSTM networks is proposed. Through the built-in two memory state lines of long term and short term, the problems that it is difficult to capture the long-term development trend and short-term fluctuation characteristics of data in time series prediction were solved. The results of case analysis show that this method can effectively improve the accuracy of power system load prediction.

This study presented a penetrating insight into the basic principle of ensemble learning and the ensemble technique Boosting, and deduced the theoretical model and learning principle of the adaptive ensemble learning. Besides, it proposed a Adaboost-based power system load forecasting method, and validated the effectiveness of this method through the empirical forecasting of a provincial medium and long-term load. The calculation example in this paper proves that high-accuracy of medium and long-term load forecasting can be achieved by using Adaboost-based power system load forecasting method.