Optimized neural network based thermal and electrical scheduling of virtual power plant in the presence of energy storage

Increasing energy consumption in the world and the presence of distributed generation pose new challenges in the management and operation of electrical grids. Virtual Power Plants (VPPs) and optimization methods assist in the integration of distributed energy resources in the power systems. One way of optimization is the use of Artificial Neural Networks (ANNs). This paper presents a new ANN based method for short-term scheduling of internal resources of a VPP in order to maximize its profit in a period of 24-h. This model is proposed to find an optimized method for connecting to the resources and VPP. The proposed model can be a way to fill the lack of a standard interface in technical problems with VPPs. Because of the different market and regulation conditions, applying an expert system, like a ANN that is independent of these rules, can be attractive. It should be mentioned that the proposed method has been examined on field data that were obtained from the power plant in Tuscany, operating in the range of 6 to 20 kW. As it is shown in the results, in the peak load condition, the Radial Basis Function (RBF) has an average error of 4% and performs better than the Multilayer Perceptron (MLP). However, in the intermittent load condition, the MLP has an average error of 6% and does better than the RBF. Nevertheless, both the MLP and the RBF have similar performance in the off-peak load condition.