A two-stage energy management framework for optimal scheduling of multi-microgrids with generation and demand forecasting

With the increasing impact of renewable energy sources in distribution power systems, the short-term scheduling of microgrids is facing many uncertainties. The supply of microgrids' consumers, on the one hand, and the optimal participation of microgrids in the market the next day, on the other hand, are faced with many challenges for microgrid operators. Therefore, this paper presents a two-level framework for day-ahead energy management of microgrids in the presence of wind turbines, solar panels, and electrical energy storage (EES) systems. In the former level, each microgrid predicts the load demand of its consumers, wind speed, and solar irradiance using historical data. In the latter one, the day-ahead scheduling of the microgrids and the power exchange rate are accomplished based on a game strategy. To predict the uncertain parameters, a stochastic method and three different artificial neural networks (ANN) based techniques are used and compared, which ANN-based methods are as follows; 1- conventional ANN (CANN) with single hidden layer, 2- long short-term memory based on deep learning neural network (DLNN), and 3- a hybrid DLNN combined by the water cycle metaheuristic algorithm (HDLNN-WCMA). The simulation results implemented on a modified 33-bus distribution power grid show that the HDLNN-WCMA compared to the CANN method have a more accurate prediction, resulting in a 2.67% reduction in operating costs. The results also show that applying the game theory method are led to a fairly dividing market power. In addition, the study of the impact of EES system on the operation of microgrids showed that the absence of EES system causes a loss of about 5970 kW of wind turbine output power and also a 10.41% increase in operating costs.