Energy Management of a Microgrid based on LSTM Deep Learning Prediction Model

Accurate and stable forecasting of total demand in micro-grid is essential for the proper operation of the energy management system. On the other hand, forecasting total renewable energy generation in micro-grid also plays a major role in scheduling the other conventional distributed generators in MG. Future power shortages in micro-grid and future excess energy generations from renewable energy sources are easily calculated when load and total renewable energy generation are forecasted. In this paper, the LSTM model is designed to forecast the total electric demand, total solar power and wind power production in micro-grid separately. To increases the performance accuracy and to reduce the complexity of the model due to a large amount of data, all three datasets are clustered in terms of seasons. So total demand and power productions are forecasted separately for all four seasons. These datasets are arranged in time series format and 2-steps predictions are employed in this study. MAE, MSE, RMSE, RSE and RAE are the performance metrics used to evaluate the model accuracy. This model of predicting both demand and fluctuating renewable energy resources enhances the energy management of microgrid.