A Data Alignment and Compression Procedure for Real-world Residual Demand Curves

Data representation is the first and important step for various analysis tasks, and the efficient representation of the residual demand curves is especially important because of their unstructured features. However, there is still a lack of attention on the representation quality. This paper aims to design a flexible and efficient procedure to improve the curve representation. From an application-oriented perspective, the proposed procedure should satisfy some typical requirements, e.g. data alignment and data compression. A curve representation model is then formulated to automatically capture the underlying description pattern. The four-year real-world data in PJM market are fully discussed with statistical results and analysis details. The numerical tests finally verify the applicability and benefits of proposed procedure with the typical application.