基于近似动态规划理论的电力系统随机动态经济调度研究

项目名称： 基于近似动态规划理论的电力系统随机动态经济调度研究

项目编号： No.51207056

项目类型： 青年科学基金项目

立项/批准年度： 2013

项目学科： 电气科学与工程学科

项目作者： 林舜江

作者单位： 华南理工大学

项目金额： 24万元

中文摘要： 风光等间歇性新能源发电的出力受风速、光照强度等随机性因素影响，给电力系统动态经济调度的优化带来很大挑战，该问题是属于多阶段随机决策问题，其精确求解难度很大。本项目将近似动态规划理论引入电力系统随机动态经济调度，近似动态规划是为解决动态规划法的"维数灾"问题而提出的，是求解下一阶段状态和决策受随机因素影响的多阶段随机决策问题的有力工具，已在火车头调度、卡车调度等交通工具调度问题上得到了应用。本项目着力解决如下问题：1）建立含新能源接入电力系统动态经济调度的随机型"存储问题(Storage problem)"模型；2）采用近似动态规划中的值函数近似类算法求解该"存储问题"模型；3）基于近似动态规划理论的随机动态经济调度方法在实际电网运行调度中的应用。本项目研究可为电力系统随机动态经济调度问题的建模和求解提供新的思路和方法，并为近似动态规划理论在含随机因素电力系统分析问题中的进一步应用提供借鉴。

中文关键词： 动态经济调度；近似动态规划；存储器模型；值函数近似；多目标优化

英文摘要： The power output of intermittent renewable energy sources such as wind and solar is affected by wind speed and sunlight intensity randomly. And this characteristic brings great challenge to dynamic economic dispatch of power system with renewable energy integration. This is a multistage stochastic decision problem, and it is difficult to slove this problem accurately. In this project, the theory of approximate dynamic programming(ADP) is introduced in power system stochastic dynamic economic dispatch newly. ADP is presented to solve the "curse of dimensionality" problem when dynamic programming method is used. It is an effect tool to solve multistage stochastic decision problem which the state and decision of the next stage is affected by random factors, and it has been applied in the traffic tool dispatch such as locomotive dispatch and truck dispatch successfully. This project focus on solving the prlblems as follows: 1)establish the stochastic "storage problem" model to represent the power system dynamic economic dispatch with renewable energy integration; 2)use the value function approximation algorithm of ADP to solve this "storage problem"; 3)apply the presented model and algorithm to the real large scale power system dynamic economic dispatch. Research of this project can offer a new way to model and sol

英文关键词： Dynamic economic dispatch；approximate dynamic programming；storage model；value function approximation；multi-objective optimization