Modelling voltage-demand relationship on power distribution grid for distributed demand management

Most existing demand response or management algorithms require a dedicated communication infrastructure to coordinate actions of electricity users. However, the necessary communication infrastructures may not be available in many low-voltage (LV) networks around the world. On the other hand, implicit information on the state of the network is readily available at all times via measurements. In this paper we propose a stochastic modelling approach to estimate aggregate network demand from local voltage measurements at each household using a gamma distribution. The model suggests a linear relationship between the expected value of network demand and voltages at households in the network. We propose a set of illustrative distributed demand control algorithms that allow making decisions based on local information only. Depending on the nature of different appliances, the algorithms either shift the entire demand block to another time (for deferable loads such as driers) or alter the consumption rate of an appliance continuously (for granular loads such as electric vehicles). We illustrate via simulations that the stochastic model captures the actual relationship between voltage and demand. The resulting demand management algorithms are efficient in reducing demand peaks without reducing the overall consumption. Moreover, the lack of explicit communication requirements makes the algorithms scalable and readily applicable to most LV networks.