UNS: A Novel Approach to Multistage Stochastic Programming for Smart Grid Applications

PIs：Chmielewski，Donald and Li，Zuyi Proposal#：1511925机构：伊利诺伊理工学院标题：一种用于智能电网应用的多阶段随机规划的新方法电网必须始终保持发电和用电之间的平衡。由于电力消耗的不确定性(表示为“需求”)，具有调度能力的发电机组处于不断变化的状态。引入可再生能源并不能缓解整体问题，因为可再生能源没有调度能力。因此，能够调度的发电机的峰值功率保持不变，这表明这些(主要是化石)电厂不太可能退役。为了缓解这种调度问题，许多人主张存储能量和/或需求响应。储能单元在电力过剩时(可再生能源发电量高而消费者需求低)收集能量，并在电力短缺时将能量返回系统。需求响应(DR)的概念是招募一部分消费者(需求)对可再生能源和其他消费者的需求做出反应。预期的影响是减少峰值功率(为淘汰一些有调度能力的电厂创造机会)，并减少差异(表明剩余的发电机将经历更少的负担)。虽然能源储存和需求响应在电网层面的好处似乎很明显，但有什么动机可以证明投资于储存设施或需求响应计划是合理的？这些可以在时间相关的电力价格中找到，这是最近放松(或基于市场)的发电商补偿制度的结果。核心问题是，不确定性(价格、需求或可再生能源发电)将迫使净现值(NPV)分析转向随机框架。具体地说，设备设计问题就变成了一个多阶段随机规划问题。在许多情况下，时间尺度的分离将MSP转化为简单的两阶段问题。不幸的是，如果给定大量的能量存储，并且这些系统将需要MSP求解方法，则该近似失败。MSP问题是工程中最具挑战性的优化问题之一。因此，该项目的目标是研究一种新的MSP问题的解决程序。该程序采用了现有和最近开发的解决方案方法的新组合，并针对智能电网应用的特定特征进行了量身定做。该项目本质上是合作的。来自两个研究社区(以PI和co-PI为代表的化学过程控制和电力系统优化)的概念和解决方案技术将需要结合起来，以实现项目目标。这种在两个领域边界进行的智力交叉授粉预计将为两个领域带来意想不到的发现。建议的MSP解决方案程序预计将适用于智能电网以外的领域。一个明显的延伸是对长期存在的集成过程设计和控制问题的扩展。另一种是流程操作调度，它与本项目中研究的整数约束机组组合问题非常相似。计划了三个REU项目的Broader Impact指导，并将继续PI的K-12外联计划。此外，该项目的一个主要目标是通过在化学工程会议、多社区研讨会和小型会议上组织受邀的教程会议以及审查和教程类型的出版物，促进化学工程智能电网社区的发展。开发便于计算的MSP解决方法被认为是转让广泛投资于储能设施的技术和采用需求响应参与的关键步骤。

AbstractPIs: Chmielewski, Donald and Li, ZuyiProposal #: 1511925 Institution: Illinois Institute of TechnologyTitle: A Novel Approach to Multistage Stochastic Programming for Smart Grid ApplicationsThe electric grid must maintain a balance between power generation and consumption at all times. Due to the uncertainty of power consumption (denoted as 'demand'), dispatch capable power generators are in a constant state of flux. Introduction of renewable power does little to alleviate the overall problem because renewable power has no dispatch capability. Thus peak power from the dispatch capable generators remains the same, indicating that retirement of these (mostly fossil) plants is unlikely. To alleviate this dispatch problem, many advocate energy storage and / or demand response. Energy storage units collect energy during periods of excess power (when renewable generation is high and consumer demand is low) and return energy to the system during times of scarcity.Intellectual Merit The notion of Demand Response (DR) is to enlist a subset of consumers (the demand) to respond to renewable sources as well as demand from other consumers. The expected impact is a reduction in peak power (creating an opportunity to retire some dispatch capable plants) and a reduction of the variance (indicating that remaining generators will experience fewer burdens). While the grid-level benefits of energy storage and demand response seem clear, what incentives justify investment in a storage facility or a demand response program? These can be found in the time-dependent electric energy prices, a result of a recently deregulated (or market based) compensation system for power generators. The central issue is that uncertainty (in price, demand or renewable generation) will force the Net Present Value (NPV) analysis to a stochastic framework. Specifically, the equipment design question becomes a Multistage Stochastic Program (MSP). In many cases, a separation of timescales converts the MSP to a simple two-stage problem. Unfortunately, this approximation fails if given significant energy storage and MSP solution methods will be required for these systems. MSP is one of the most challenging optimization problems in all of engineering. Thus, the project objective is to investigate a new solution procedure for MSP problems. This procedure employs a novel combination of existing and recently developed solution methods and is tailored to exploit the specific characteristics of the smart grid application. The project is collaborative in nature. Concepts and solution techniques from two research communities (chemical process control and power system optimization, represented by the PI and co-PI), will need to be combined to achieve project goals. This intellectual cross-pollination at the boundary of two fields is expected to yield unforeseen discoveries for both. The proposed MSP solution procedure is expected to have applicability beyond smart grid. A clear extension is to the longstanding problem of integrated process design and control. Another is process operations scheduling, which is very much similar to the integer constrained unit commitment problem studied in this project.Broader Impacts Mentoring of three REU projects is planned and the PI's K-12 outreach program will be continued. In addition, a major goal of the project is to foster a chemical engineering smart grid community, through the organization of invited tutorial sessions at chemical engineering conferences, multi-community workshops and mini-conferences, and review and tutorial type publications. Development of computationally tractable MSP solution methods is perceived as an essential step to the transfer of technology for widespread investment in energy storage facilities and adoption of demand response participation.