Demand baseline estimation using similarity-based technique for tropical and wet climates
Demand baseline estimation (BE) is key to the impact assessment of a demand response event in a power system. While many BE techniques exist in literature and are implemented by utilities, these are either inaccurate, or computationally intensive, and only provide point estimates of the demand baseline. This study presents a simple, single-stage, similarity-based BE technique. The authors posit a new definition of similarity that includes weather covariates, and therefore eliminate the need for a subsequent adjustment. A novel growth rate assumption for the demand, combined with an optimised exponential smoothing technique results in a higher accuracy for the proposed BE technique. Additionally, an L-order iterated bootstrap is used to generate confidence intervals to account for prediction uncertainties. The proposed BE technique is tested for the Singaporean National Electricity Market, and is shown to be consistently more accurate than other conventional BE techniques.
