Data Valuation From Data-Driven Optimization

作     者：Mieth, Robert Morales, Juan M. Poor, H. Vincent 

作者机构：Rutgers Univ New Brunswick Dept Ind & Syst Engn Piscataway NJ 08854 USA Univ Malaga Dept Math Anal Stat & Operat Res & Appl Math Malaga 29071 Spain Princeton Univ Dept Elect & Comp Engn Princeton NJ 08544 USA 

出 版 物：《IEEE TRANSACTIONS ON CONTROL OF NETWORK SYSTEMS》 (IEEE Trans. Control Netw. Syst.)

年 卷 期：2025年第12卷第1期

页      面：954-966页

核心收录：

学科分类：0808[工学-电气工程] 08[工学] 0811[工学-控制科学与工程] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

基　　金：Spanish Ministry of Science and Innovation (AEI_ [PID2020-115460GB-I00, PID2023-148291NB-I00] European Research Council (ERC) through the European Union C3.ai DigitalTransformation Institute Princeton University School of Engineering and Applied Science 

主　　题：Data integrity Decision making Cost accounting Measurement Power systems Optimization Costs Data-driven modeling differential privacy forecast uncertainty power system analysis computing wind energy integration 

摘      要：With the ongoing investment in data collection and communication technology in power systems, data-driven optimization has been established as a powerful tool for system operators to handle stochastic system states caused by weather-dependent and behavior-dependent resources. However, most methods are ignorant to data quality, which may differ based on measurement and underlying privacy-protection mechanisms. This article addresses this shortcoming by proposing a practical data quality metric based on Wasserstein distance, leveraging a novel modification of distributionally robust optimization using information from multiple datasets with heterogeneous quality to valuate data, applying the proposd optimization framework to an optimal power flow problem, and, finally, showing a direct method to valuate data from the optimal solution. We conduct numerical experiments to analyze and illustrate the proposed model and publish the implementation open source.