International Core Journal of Engineering 2020-26 | Page 23

2019 International Conference on Artificial Intelligence and Advanced Manufacturing (AIAM) Settlement Data Risk Identification and Processing of Power Market Based on Evidence Theory Minghui Yan Kaisheng Lin Kunming Power Exchange Center Company Limited Kunming 650011, China [email protected] School of Electrical Engineering Wuhan University Wuhan 430072, China [email protected] Xuejin Wang, Ruichen Wang, Weijie Li Liming Ying, Xue Cui Kunming Power Exchange Center Company Limited Kunming 650011, China School of Electrical Engineering Wuhan University Wuhan 430072, China [email protected] Abstract—In recent years, China's power market reform has been continuously promoted, and reasonable settlement of trading behavior is an important cornerstone for ensuring the orderly development of the power market. The paper analyzes the data risks faced in the power market settlement, and proposes the method of using DS evidence theory to identify the data risks in the settlement process through multi-evidence fusion decision. The PCA and ICA construct statistics are used as evidence sources. After multiple evidences are combined, abnormal data is identified through reasonable decision- making methods. In addition, based on the calculation results of evidence theory, the paper puts forward the risk processing method of settlement data, and proposes a reconstruction method based on polynomial fitting and historical correlation for the abnormal data with low reliability. The effectiveness of the proposed method is demonstrated by an example. The reconstruction method can obtain more accurate data and can effectively reduce the settlement risk of the power market. Shafer evidence theory (DS) can fuse multiple evidences, analyze the problems that need to be solved comprehensively, and improve the accuracy of the conclusions. The theory is also widely used in related fields of data processing. In [7], the DS evidence theory is combined with LSSVM, and a power system transient stability assessment method based on data fusion is proposed. Reference [8] proposed a framework for distributed fault detection and recognition based on DS evidence theory. The data samples were divided into different levels as multi-source evidence, and then evidence fusion was performed to obtain more accurate detection results. This paper proposes a method for identifying the data risk of power market settlement based on DS evidence theory. The Principal Component Analysis(PCA) and Independent Component Analysis(ICA) are used to construct the statistic as evidence, and the multiple evidences are combined. The decision rules are used to obtain the more accurate anomaly data identification results, and the corresponding risk identification model is built. The example is proved that the proposed method can effectively identify the abnormal data existing in the data samples. In addition, the paper also analyzes how to deal with the data risk of power market settlement, introduces the concept of reconstruction threshold, and proposes a polynomial fitting and reconstruction method based on historical correlation. The proposed method is verified by an example. Can get more accurate reconstruction data. Keywords—power market, settlement, abnormal data, data risk processing, evidence theory I. I NTRODUCTION Since 2015, all over China have begun to gradually explore the construction of power market [1], and the power market has become a hot spot in domestic research. Power market settlement is an important part of market-oriented transactions and an important cornerstone for ensuring the stable and orderly development of the power market. In the settlement process, data anomalies or even missing cases may occur due to various reasons. This settlement data risk is an important part of the power market settlement research. II. I NTRODUCTION TO DS E VIDENCE T HEORY In the process of data risk identification, DS evidence theory combines multiple evidences to provide decision- making for the identification of abnormal data, which can effectively improve risk identification’s accuracy. It is of great practical significance to study the identification and processing of settlement data risks to improve the accuracy of power market settlement. Multivariate statistical methods [2], cluster analysis methods [3], neural network methods [4], support vector machine methods [5] and other data-driven risk identification methods are dealing with huge amounts of settlement data in the power market with unique advantages. A. Basic of DS Evidence Theory When performing abnormal data identification, evidence theory can integrate the same direction information in multiple evidences [9-10]. When deciding on a specific problem, define all possible outcomes as a recognition framework 4 . An evidence can support one or more subsets, the degree of support for the subset is measured by the As a method of dealing with uncertainty [6], Dempster- 978-1-7281-4691-1/19/$31.00 ©2019 IEEE DOI 10.1109/AIAM48774.2019.00007 1