Monitoring Stakeholder Behaviour for Adaptive Model Generation and Simulation: A Case Study in Residential Load Forecasting

Simulations are important in the design and development of large-scale infrastructures, such as the smart grid, to explore the impact of design choices on local and overall system behaviour, given reliable information on the behaviour of individual resources and their interaction. This paper proposes a new multi-agent approach to dynamic generation of reliable models leveraging real-time data from network resources in a distributed manner by addressing privacy, scalability and accuracy. The approach was incorporated into an adaptive model generation and simulation system with which models of stakeholder resources are generated, stored in a repository, and updated automatically. The system is illustrated with a use-case on residential load modelling for short-term demand forecasting using neural networks.