Sizing Curve for Isolated Photovoltaic-Battery Systems using Artificial Neural Networks

Isolated power systems are designed to generate electric power in a decentralized manner. They are helpful in the electrification of inaccessible locations where grid extension becomes infeasible. Mathematical modeling helps in the system sizing and optimization of isolated systems. Design space approach provides a methodology for the sizing of isolated power systems incorporating the uncertainty of the electrical demand and the energy resource at the design stage. The design space can be represented on generator rating vs. storage capacity coordinates showing the feasible and infeasible design configurations by plotting a sizing curve. The data set of system configurations obtained based on design space approach may be utilized for building an Artificial Neural Network (ANN) based system sizing tool. A two-stage ANN approach is proposed in this paper. In the first stage network, the geographical coordinates of the location are the inputs which provide the minimum normalized array rating and the corresponding normalized battery capacity as the outputs for the specified location. In the second-stage network, with the geographical coordinates and the normalized array ratings as the inputs, the corresponding normalized storage capacities are obtained as the outputs. Thus, the developed ANN based model can be used for generating the sizing curve without detailed simulations for the location of interest. As a representative case study, the design space approach is used for generating isolated photovoltaic-battery system sizing curves for fifty representative locations in southern India (latitudes ranging from 8°N to 25°N). Two-stage ANN model is developed using this data and the generalization capabilities of the developed networks are illustrated based on this case study.