Abstract A Microgrid (MG) is a promising distributed technology to solve todays energy challenges. They are changing how electricity is produced, transmitted, and distributed, enabling to capture massive amounts of data from sensors, and other electrical infrastructures. However, recent advances in modeling and optimization of MG neither integrate the use of big data technologies aggressively nor focus on developing an optimal operational strategy for a single building. To bridge this gap, this research proposes to use Apache Spark to enhance the performance of a scalable stochastic optimization model for an MG for multiple buildings, and to ensure that a significant portion of the wind power output will be utilized. The decision model is formulated as a chance constraint two-stage optimization problem to obtain operation decisions for a behind-the-meter topology. The comparison between the current practice of using historical data and integrating Apache Spark technologies demonstrates the superiority of the streaming data as energy management strategy. Experiments under different settings show that using big data strategy, the model can (1) achieve more cost savings of the total system, (2) increase resiliency to power disturbances, and (3) build a data analytics framework to enhance the decision-making process.

中文翻译：

---

基于机会约束随机优化和大数据分析的微电网能源管理系统

摘要 微电网（MG）是一种很有前途的分布式技术，可以解决当今的能源挑战。它们正在改变电力的生产、传输和分配方式，从而能够从传感器和其他电气基础设施中捕获大量数据。然而，MG 建模和优化的最新进展既没有积极整合大数据技术的使用，也没有专注于为单个建筑制定最佳运营策略。为了弥合这一差距，本研究建议使用 Apache Spark 来增强多建筑物 MG 的可扩展随机优化模型的性能，并确保将利用大部分风能输出。决策模型被表述为机会约束两阶段优化问题，以获得电表后拓扑的操作决策。当前使用历史数据和集成 Apache Spark 技术的实践之间的比较证明了流数据作为能源管理策略的优越性。不同设置下的实验表明，使用大数据策略，该模型可以（1）实现整个系统的更多成本节约，（2）提高对电力干扰的弹性，以及（3）构建数据分析框架以增强决策过程。