**6. Conclusions**

High-frequency smart meters have been broadly deployed for collecting electricity data. Our proposed method implements discrete wavelet transform to convert time-domain data to frequency domain. We extracted detailed and approximate signals using a statistical approach, then used Pearson's correlation coefficients to filter the high correlation features. To further reduce dimensionality, we applied PCA to preserve three features. The rest of the three features were used to achieve the clustering algorithm. Our study aimed at obtaining the representative load patterns from high time resolution daily load curves in Manhattan. Our method reduces the large dimensions to increase efficiency in clustering. In addition, it improves the clustering result slightly by estimating the silhouette coefficient, Calinski– Harabasz index, and Davies–Bouldin index, then comparing the clustering without discrete wavelet transform. From representative load patterns, the utility policymaker could design a reasonable demand response scheme to maintain the power system stability and help the utility maximize the profit and even reduce consumers' electricity fees. Based on Figure 14, policymakers could design three different advanced time of use tariffs, according to electricity consumption volume and representitive load curves from the three clusters. To each cluster, the electricity demand increases apparently from 4 pm to 8 pm, which could influence the power system stability. It means the appropriate DR scheme is significant during this period, such as load shifting/shedding. For future work, we suggest exploring the sub-cluster from the previous clusters to get more detailed load patterns based on our method.

**Author Contributions:** Conceptualization, S.C. and C.G.L.; methodology, S.C.; validation, S.C., C.G.L. and J.H.Y.; formal analysis and investigation, S.C.; writing—original draft preparation, S.C.; writing—review and editing, S.C., C.G.L. and J.H.Y.; visualization, S.C. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research is supported by Jeollannam-do (2021 R&D supporting program operated by Jeonnam Technopark) and financially supported by the Ministry of Trade, Industry and Energy (MOTIE) and Korea Institute for Advancement of Technology (KIAT) under the research project: "National Innovation Cluster R&D program" (Grant number: P0016223).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Data is available in a publicly accessible repository. The data used in this study are openly available from the Scientific Data portal in https://www.nature.com/articles/ s41597-020-00721-w, accessed on 13 January 2022.

**Conflicts of Interest:** The authors declare no conflict of interest.
