Decentralised Scheduling of Power Consumption in Micro-grids: Optimisation and Security

We consider a micro-grid architecture that is distributed in nature and reliant on renewable energy. In standard grid architectures, demand management is handled via scheduling protocols that are centrally coordinated. Centralised approaches are however computationally intensive, thus not suited to distributed grid architectures with limited computational power. We address this problem with a decentralised scheduling algorithm. In our scheduling algorithm, the alternating direction method of multipliers (ADMM) is used to decompose the scheduling problem into smaller sub problems that are solved in parallel over local computation devices, which yields an optimal solution. We show that ADMM can be used to model a scheduling solution that handles both decentralised and fully decentralised cases. As a further step, we show that false data injection attacks can be provoked by compromising parts of the communication infrastructure or a set of computing devices. In this case, the algorithm fails to converge to an optimum or converges toward a value that lends the attacker an advantage, and impacts the scheduling scheme negatively.