Typical load curves identification for practical demand response potential assessment based on fixed interval linear fitting and density-based adaptive clustering

• Fixed-interval fitting retains load curve features while enhancing clustering efficiency. • A-CFSFDP improves cluster accuracy through outlier detection and silhouette coefficients. • Hierarchical target decomposition and dynamic coordination strategy optimizes load adjustment in demand response. • Integrate A-CFSFDP with resource allocation for practical demand response solutions, enhancing target achievement. • Enable data-driven demand response potential analysis via improved clustering method. The identification of power load pattern which is the critical issue for customer-side resource management is studied under the context of the scaled application of demand response technologies in this paper. An industry typical load curve identification method based on improved linear representation strategy and adaptive clustering by fast search and find of density peaks (A-CFSFDP) is proposed. First, a fixed interval linear fitting (FILF) strategy based on controllable dimensions is proposed to achieve the extraction of effective load data features from massive data, which highlights the morphological features of the load curves. This approach effectively retains the load features while reducing the load data to shorten the clustering time. Then, an A-CFSFDP method based on local outlier factor (LOF) and silhouette coefficient (SC) is proposed to classify the load power usage types and summarize the power usage characteristics. Finally, case studies on industrial users in Zhejiang, China verify that the proposed method can effectively identify typical load curves and accurately classify electricity consumption characteristics. The Calinski-Harabasz index (CHI) and Davies-Bouldin index (DBI) of A-CFSFDP improved by an average of 51.13 % and decreased by an average of 34.55 % compared to other methods, respectively. Moreover, users with adjustment potential are selected based on the characteristics of different load patterns and the amount of adjustable load is analyzed. The clustering results exhibit superior performance in comparison to traditional methods in terms of reducing subjective factors and effectiveness indices. The proposed method enables power companies to develop tailored adjustment strategies by classifying load patterns, which can quickly target users with high adjustment capacity and significantly improve demand response execution efficiency.