The challenge is to prevent serious overloading or voltage disturbances using only measurements made within the customers' premises. Frequency will only give information about the overall supply/demand on the wider network, and load current gives no insight into what other customers in the locality are doing. Local voltage, as a single data point, also has limited value. However, if under normal smart grid operations, each customer's energy management system records the local supply voltage and the voltage step change caused when loads are switched, it can build up a pattern identifying the lowest acceptable local supply voltage and the impact of switching each appliance. Hence, in the absence of smart grid control or communications, it would be able to measure the local voltage and take actions based on comparing this with the lowest acceptable voltage. Such a system could provide a level of protection against the loss of communications, or the smart grid controller being compromised.