With conventional generation capacity being constrained on environmental grounds and renewable alternatives carrying capacity uncertainties, increasingly accurate forecasts of demand are likely to be required in future power systems: highly distributed renewable generation penetrating low voltage networks must be matched to small dynamic loads, while spinning reserves of conventional generation that are required to maintain security of supply, must be reduced to more efficient margins. Domestic loads, likely to form significant proportions of the loads on islanded power systems such as those in remote rural communities, are currently modeled with homogenous and coarse load profiles developed from aggregated data. An objective of AMR deployment is to clarify the nature and variability of the residential LV customer. In this paper, an algorithm for tracking the consistency of the behavior of small loads is presented. This would allow them to be assessed for their availability to provide demand services to the grid. In the method presented, significant changes in behavior are detected using Bayesian changepoint analysis which tracks a multivariate Gaussian representation of a residential load profile on a day to day basis. A hypothetical single phase feeder, representative of an islanded rural power system, is used to illustrate the detected heterogeneity of load behavior consistency.
- automatic meter reading (AMR)
- LV networks
- demand characterisation
- bayesian statistics