Load estimation based on self-organizing maps and Bayesian networks for microgrids design in rural zones

Abstract

Microgrids are suitable electrical solutions for providing energy in rural zones. However, it is challenging to propose in advance a good design of the microgrid because the electrical load is difficult to estimate due to its highly dependence of the residential consumption. In this paper, a novel estimation methodology for the residential load profiles is proposed. Socio-demographic data and electrical power consumption are used to generate significant knowledge about the load behavior. Socio-demographic data are used as input for a neural network called Self-Organizing Maps (SOM). The SOM proposes a way to group dwelling according to their different features. Moreover, a probabilistic model based on Bayesian networks incorporates daily variations of the electrical load, simulating the behavior of the electrical appliances. The methodology, as a whole, is applied to a case study in a rural community located in Chile. The methodology is easily adaptable to other rural communities.