Using a bayesian network to evaluate the social, economic and environmental impacts of community deployed renewable energy

Social, economic and environmental (SEE) impacts resulting from the adoption of solar PV have been modelled at a community scale for the first time using a probabilistic graphical model in the form of a Bayesian Network (BN). Model parameters required to conceptualise this multi-disciplinary problem domain are characterised by uncertainty due to stochastic variability, measurement and modelled data errors, or missing or incomplete information. A BN conveniently represents the model parameters and the associations between them and endogenises the uncertainty in probability distribution functions or mass functions. The theory and method of construction of an object-oriented BN which encapsulates a number of SEE parameters is described. This is used to model small urban areas as potential adopters of solar PV technology. The BN has been populated with modelled and empirical quantitative data from a variety of disciplines to create an inter-disciplinary knowledge representation of the problem domain. The model has been used to explore a number of scenarios whereby ‘observations’ are made on one or more variables of interest thus altering their prior probability distribution. The updated or posterior distributions of all the other variables are then recalculated using inference algorithms. Results are presented which show the utility of this approach in diagnostic and prognostic inference making. For example it is shown that Solar PV can have a small but significant impact on energy poverty. It is concluded that the adoption of a BN modelling approach that endogenises uncertainty, and reduces investment and policy risks associated with energy technology interventions within communities, can act as a useful due diligence and decision support tool for a number of private, public and community sector stakeholders active in this sector, in particular key decision and policy makers