High-Resolution Load Forecasting Approach for Micro-grid with Renewable Energy Sources

摘要

Power system management is a complex project and accurate load forecasting is necessary for energy production because predicted loads can determine the balance of future energy supply and demand. We present in this paper a multilayer perceptron neural network model for short-time load forecasting with 15-min time intervals under the microgrid scenario. Specifically, a feature selection approach is proposed to improve the forecast results which considering lagged power value information. A case study of a real microgrid is performed, the proposed model is applied to load power prediction taking into account the penetration of renewable energy sources, and the obtained average prediction error was evaluated. By analyzing the experiment results, it shows that the proposed method provides accurate predictions, with more efficient and high generalization ability.