Clustering of electricity customers supports effective market segmentation and management. The literature suggests the clustering of residential customers by their load characteristics. The key challenge is the application of appropriate processes to reduce the extreme dimensionality of load time series to facilitate unique clusters. Time feature extraction is a potential remedy, however, it is limited by the type of noisy, patchy, and unequal time-series common in residential datasets. In this paper we propose a strategy to alleviate these limitations by converting any types of load time series into map models that can be readily clustered. This also results in higher cluster distinction and robustness against noise compared to a baseline feature-based approach. A large dataset of residential electricity customers is used to confirm the outcomes as measured by a number of analytical and industrial metrics. The experiment with 12 clusters results in around 61% distinction, improved coincidence factor by around 6.75% relative to a random grouping, and robustness of around 59% against the applied noise.