UNS: Sustainable Energy-Intensive Manufacturing via Demand Response Process Operations

美国和世界各地的能源生产正在发生变化，在过去十年中，可再生能源的贡献增加了两倍。但是，可再生能源本身是可变的，由于风速、日照和云量等因素，它们的发电速率会随时间波动。电网能源消耗也各不相同，通常在下午晚些时候达到每日峰值，在清晨达到最低。为了满足这一高峰需求，电网运营商必须启用额外的发电设施（称为“调峰电厂”），这些设施通常比基本负荷发电机效率低，污染更大。平衡波动的能源资源和消费者是目前开发和部署智能电网的主要动力之一，智能电网的基础是发电和用电的紧密整合和同步。需求响应（DR）策略，由联邦能源管理委员会（FERC）定义为“在批发市场价格高或系统可靠性受到威胁时，需求侧资源从其正常消费模式（对经济激励和动态定价结构的响应）的电力使用变化”，将在协调智能电网的能源生产和消费方面发挥关键作用。DR预计将在未来十年内使美国的净峰值电力需求减少约5%（或约50GW），并具有显著的可持续性效益，因为它消除了对新峰值电厂的需求，减少了现有电厂的使用和排放。该项目旨在将DR策略应用于工业环境。智力优势：能源密集型化学制造过程（例如，空气分离，氨，氧化铝，氯碱）约占工业用电量的10%。由于它们的开关容量和存储能源密集型产品的能力，它们非常适合DR计划。DR要求在原料质量、产品需求和环境条件波动的情况下，在广泛的操作范围内快速过渡并保持高效。虽然与DR操作相关的生产调度问题已经得到了广泛的解决，但实施结果调度的动态和控制方面却很少受到关注。因此，本项目旨在i)建立一个新的框架，用于建模、分析和优化在DR下运行的过程系统的动力学和控制，ii)开发一个计算效率高的算法实现，并在工业相关过程（空气分离、热电联产）上进行验证。这些结果将适用于分析和缓解现有设施的动态瓶颈，也适用于新项目。DR过程的动力学和控制优化将被表述为不确定条件下的动态优化，并利用非线性系统辨识理论的启发进行求解。这种方法避免了传统上使用情景来捕捉不确定性。相反，不确定变量的波动将被描述为具有精确调谐频率内容的伪随机多级信号，这些信号将在动态优化的迭代过程中施加于过程模型。通过这种方式，系统的相关动态模式被选择性地激发（“识别”），并且可以被最优调制以最小化目标函数的期望值。更广泛的影响：通过与德克萨斯大学工程平等机会项目的持续合作，PI将从代表性不足的群体中招募本科研究人员，他们将为未来的研究生学习获得宝贵的经验。在提案中发展的概念将被纳入由PI准备的工程系统数学建模课程。提出了一项新的外展倡议，为来自低收入少数民族家庭的中学生和家长团队提供真实的工程体验，目的是支持学生？努力成为第一代大学毕业生。

Baldea, 1512379Energy generation in the United States and across the world is shifting, with the contribution of renewable sources tripling in the past decade. But, renewable sources are inherently variable and their generation rates fluctuate in time due to factors such as wind speed, insolation and cloud cover. Grid energy consumption varies as well, typically reaching a daily peak in the late hours of the afternoon and a minimum in the early morning. To meet this peak demand, grid operators must turn on additional generation facilities (referred to as "peaking plants"), which are typically less efficient and more polluting than base load generators. Balancing fluctuating energy resources and consumers is one of the central motivators behind current efforts to develop and deploy the smart grid, which is predicated on a close integration and synchronization of electricity generation and electricity use. Demand response (DR) strategies, defined by the Federal Energy Regulatory Commission (FERC) as "Changes in electric usage by demand-side resources from their normal consumption patterns [in response to economic incentives and dynamic pricing structures] at times of high wholesale market prices or when system reliability is jeopardized" will play a key role in coordinating energy generation and consumption in the smart grid. DR is expected to reduce net peak power demand in the US by about 5% (or about 50GW) over the next ten years, with significant sustainability benefits related to eliminating the need for new peaking plants and reducing the use of and emissions from existing ones. This project is aimed at applying DR strategies in an industrial context.Intellectual Merit: Energy-intensive chemical manufacturing processes (e.g., air separation, ammonia, alumina, chlor-alkali) account for about 10% of industrialelectricity consumption. They are ideally suited for DR initiatives due to their turndown capacity and ability to store energy intensive products. DR requires processes to transition rapidly and remain efficient across a broad operating envelope, in the presence of fluctuations in feedstock quality, product demand and ambient conditions. While the production scheduling problem associated with DR operation has been extensively addressed, the dynamic and control aspects of implementing the resulting schedules have received much less attention. Therefore, the present project aims to i) establish a novel framework for modeling, analyzing and optimizing the dynamics and control of process systems operating under DR and, ii) develop a computationally efficient implementation of the proposed algorithms and validate it on industrially relevant processes (air separation, combined heat and power). These results will be applicable to the analysis and mitigation of dynamic bottlenecks in existing facilities, as well as to new projects. The optimization of the dynamics and control of processes operating involved in DR will be formulated as a dynamic optimization under uncertainty, and solved using an approach inspired by nonlinear systems identification theory. THe approach eschews the traditional use of scenarios to capture uncertainty. Rather, the fluctuations of the uncertain variables will be described as pseudo-random multi-level signals with precisely tuned frequency content, which will be imposed on the process model during the iterations of a dynamic optimization. In this manner, the relevant dynamic modes of the system are selectively excited ("identified") and can be optimally modulated to minimize the expected value of the objective function.Broader Impact:Through an ongoing partnership with the University of Texas Equal Opportunity in Engineering program, the PI will recruit undergraduate researchers from underrepresented groups, who will acquire valuable experience towards future graduate studies. The concepts developed in the proposal will be incorporated in a course on Mathematical Modeling of Engineered Systems prepared by the PI. A new outreach initiative to provide real-life engineering experiences for middle-school student and parent teams from low-income, minority families is proposed, with the aim of supporting the students? efforts to become first-generation college graduates.