DC-Microgrid Price Prediction using Long Short Term Memory with Double Activation Function

Due to growth in renewable energy sources and distributed generation (DG), the importance of short term forecasting is increasing day by day and has been a matter of interest in power system operation and control. Short term forecasting is vital for industries, buildings and microgrids that are using distributed energy resources for optimal operation of their resources to meet their daily energy balance. Hence in this paper, a multistep short term forecasting method has been suggested for microgrids and smart grids for optimal power generation and consumption. The proposed method can select the best combinations of inputs and the best techniques for Short Term Load Forecasting (STLF) by evaluating several combinations of inputs and various types of loads. Better forecasting results can be obtained even if there is insufficient information at the input. In this paper, a robust forecasting model has been designed which compares a set of conventional rule based base line predictions and various machine learning techniques such as ANN and Random forest. So the suggested model has rule based base line model which gives input to the machine learning model and a comparison has been made between the two ANN models. The proposed model has been achieved a 62% improvement in the prediction accuracy as compared to the conventional techniques.