电力系统物理信息智能控制与运行优化

Gao QiuSichuan university Groupintroduction Gao Qiu, from Smart Grid Optimal and Power Market OperationResearch Team (Leader: Prof. Junyong Liu) Biography: Gao Qiu received the B.S. degree in college of electricalengineering and information technology, Sichuan University (SCU), Chengdu,China, in 2016, and the PhD dcgree in collegc of clcctrical cngincering, SCU,Chengdu, China, in 2021. He was with the Collcgc of Engincering & ApplicdScicncc, University of Wisconsin-Milwaukce, Milwaukce, USA in 2019 as anacademic visiting scholar. He is currently an Associate Researcher with thecollege of clectrical cngincering, SCU. He won the Nomination Award forExcellent Doctoral Dissertation of the Chinese Electrotechnical Sociely in 2021and has (co-)authored more than 20 papers. His current research interestsincludepower systemknowledge discovery and operation, artificialdynamic transfer limit control, etc. Groupintroduction Prof. Junyong Liu, Ph.D., serves as the head of the research team-Smart Grld optlmal and PowerMarket Operation Research Team, Sichuan University (SCU). He is a Level-2 Professor and doctoralsupervisor, a recipient of the State Council Special Allowance, and holds an honorary doctorate fromBrunel University London, UK. He served as the Chair of the 3rd and 4th Councils of Presidents of ChineseHigher Education Institutions in Electric Power, and is a Board Member of the Chinese Society forElectrical Engineering (CSEE). He is also a Distinguished Expert appointed by IBM, Prof. Liu serves on theeditorlal boards of more than ten academlc Journals, including CSEE Journal of Power and Energy Systemsand Journal of Modern Power Systems and Clean Energy (SCl) The team comprises 4 full professors, 5 associate professors, and 3 associate researchers, annually supervising over 50doctoral and master’ s students. Economic operation of power systems and electricity markets have been core researchand teaching areas. The team has led more than 20 major projects funded by China' s 863 and 973 Programs, theMinistry of Science and Technology’ s Key R&D Program, and the National Natural Science Foundation of China.including key, general, youth, and major intemational cooperation grants, the team has published or translated 7monographs and authored over 400 papers, with more than 250 indexed by SCI and El. Awards include three SichuanProvinclal Sclence and Technology Progress Awards, second-class awards from 5tate Grld and chlna Southern PowerGrid, a first-class award from Guangdong Power Grid, and over 20 granted national invention patents. 1.Backgroundand Motivation2.Learning/surrogate-assisted methodology and its applications3.Physics-driven learning technology and its application4.Prospects BackgroundandMotivation Background: risks in electric power systems BackgroundandMotivation Motivation:Artificial Intelligence(Al) iongleIeepMind'sAlphao developedreinfiorcenent learming strategy conhined withMonte-Carlo tree scarch. AlphiaGo has dcfcatedfamous human GO players, proving its strongreal-timc decision ability leaming dynamic behaviors from massivedata and bypasses complicated physieal modelling.which is a promising solutions for fast and accuratesecurity analysis in large power grids the real-time strategy game StarCraff Il, hasGoogle DeepMind's AlphaStar, developed forsuccessfullydefealedwhich verifics the feasibility of Al in decisionhumanprlessinalsin varying and eomplicated environments BackgroundandMotivation 1.Background and Motivation 2.Learning-assisted methodology and itsapplications 2.1Multi-period Operational Planning with Dynarmic TTC Constraints2.2Real-time Dynamic Total Transfer Capability Control2.3A scenario-classification hybrids-based banding method for power transfer limits2.4Transient stability prevention control using integrated gradient technique 3Physics-driven learning technology and its application Learning-assisted methodologyanditsapplications 2.1 Multi-period Operational Planning with Dynamic TTC Constraints Serurtiy csarinlets be Isikresridr > Challenge I: time-series features bave been widely :introduced into power systcms. It is thus nccessaryconsider multi-period mkiel far TTC regulation, However,:this model is of high dinensionality and complexity 1.).a()) B,w cFiu(r) = PF([x(tL gTLA)W e K Solution:capitalizeonlearning-assistedmcthod,complexity reduetion to original model is enabled Challenge Il: High dimetisiotal leaing-assisted modelis still time-consuming to be solved by gradient-freesolvers. Solution: we dcrive Jacobian and Hessian imairiccs ofthat gradient-based nonlinear solver, i.e_, interior pointmcthod, can be drawn on solving the leaming-assistodmodel in the steepest descendl manmer. 四川大学Learning-assisted methodologyanditsapplications 2.1 Multi-period Operational Planning with Dynamic TTC Constraints > Numerieal study: Tested on thc modificd IEEE 39-bus and 68-bussyslems, including ideal energy skorage syslen andDFIG wind fam4-order generator dy