Electricity grids are designed to withstand a number of realistic contingencies typically involving only a few pieces of equipment. However, unexpected events can lead to a cascading failure and large-scale blackouts. With EU connecting renewable sources of energy to the grid, there is no time like the present to enhance understanding of vulnerability to cascading outages. With EU support of the project 'Cascading failures in electrical networks: stochastic analysis and distributed prevention methods' (PROFILINGBLACKOUTS), scientists developed a network model of cascading outages triggered by an initial small-scale event. They used it to quantify risks of cascading outages in a number of scenarios. The scenarios differed in the connectivity of the network and the extent of decentralised electricity generation. For each scenario, risk profiles were computed that summarise the expected frequency and magnitude of outages. The profiles were based on large numbers of simulations on randomly generated representative networks. Determining accurate risk profiles for rare-but-large events such as cascading outages necessitated the development of a novel data analysis method with strict robustness properties. In addition, the project investigated whether intelligent appliances in the nodes of the network can provide a robust decentralised service to reduce network stress, and thus reduce the risk of large-scale cascades. Taking domestic refrigerators as a case study, investigators developed and applied a new stochastic control method to balance supply and demand. The controller exploits the intrinsic flexibility of refrigerators’ power consumption, so that the resilience of the grid can be enhanced in a cost-effective way. Public support for renewable energy sources connected to the existing electricity grids will rely largely on ensuring energy security at a reasonable price. PROFILINGBLACKOUTS has delivered important contributions to this goal.
Cascading power failure, electricity grid, large-scale blackouts, vulnerability, network, model, resilience