Demand response is widely regarded as a critical enabler of cost-effective low carbon grids. In particular, there is a lot of untapped potential in the 'long tail' of appliances with low cost and moderate power consumption: they can collectively provide a significant contribution to system flexibility, either autonomously or through user actions.

In this talk, I will address two related challenges in unlocking small scale demand response. The first is the quantification and attribution of an observed response - something that is particularly important in dynamic pricing schemes. I will present data, results and methods from the Low Carbon London project, which included the UK’s first trial of residential dynamic time-of-use tariffs. I will describe the observed responsiveness at a high level, before specifically addressing the identification of customers who successfully engaged with the price signal, as evidenced by a noticeable change in their consumption pattern. We developed a nonparametric resampling approach that can be used to identify and rank responsive customers, and (under some assumption) assign a likelihood of responsiveness. In addition, smart appliances are able to provide demand response autonomously, and refrigerators are a particularly appealing choice, due to their always-on character. However, it is not straightforward to design a control strategy that (1) does not require expensive real-time communication, (2) provides accurate control and (3) respects individual fridge temperature limits. I will present a recently developed 'semi-autonomous' control strategy for refrigerators that simultaneously addresses these challenges. By constructing a probabilistic model of population behaviour, fridges can individually take actions in line with desired aggregate behaviour. The capabilities of individual devices can be aggregated into a 'leaky storage' model, which allows them to be dispatched by a service aggregator.

Dr Simon Tindemans is a Research Fellow at Imperial College London, in the Control and Power research group. With a background in biological physics, he has worked on statistical aspects of power system planning and operation since 2010.